Add benchmark tools & Reorgnazie configs

.dev_scripts/benchmark_test.py

@@ -0,0 +1,250 @@
+import argparse
+import os
+import os.path as osp
+import re
+from collections import OrderedDict
+from pathlib import Path
+
+import mmcv
+import wget
+from modelindex.load_model_index import load
+from rich.console import Console
+from rich.syntax import Syntax
+from rich.table import Table
+
+console = Console()
+MMRAZOR_ROOT = Path(__file__).absolute().parents[1]
+
+METRIC_MAPPINGS = {
+    'accuracy/top1': 'Top 1 Accuracy',
+    'accuracy/top5': 'Top 5 Accuracy'
+}
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(
+        description="Test all models' accuracy in model-index.yml")
+    parser.add_argument(
+        'partition', type=str, help='Cluster partition to use.')
+    parser.add_argument('checkpoint_root', help='Checkpoint file root path.')
+    parser.add_argument(
+        '--job-name',
+        type=str,
+        default='razor-test-benchmark',
+        help='Slurm job name prefix')
+    parser.add_argument('--port', type=int, default=29666, help='dist port')
+    parser.add_argument(
+        '--models', nargs='+', type=str, help='Specify model names to run.')
+    parser.add_argument('--gpus', type=int, default=8, help='num gpus')
+    parser.add_argument(
+        '--work-dir',
+        default='work_dirs/benchmark_test',
+        help='the dir to save metric')
+    parser.add_argument(
+        '--run', action='store_true', help='run script directly')
+    parser.add_argument(
+        '--summary', action='store_true', help='collect results')
+    parser.add_argument(
+        '--local',
+        action='store_true',
+        help='run at local instead of cluster.')
+    parser.add_argument(
+        '--mail', type=str, help='Mail address to watch test status.')
+    parser.add_argument(
+        '--mail-type',
+        nargs='+',
+        default=['BEGIN'],
+        choices=['NONE', 'BEGIN', 'END', 'FAIL', 'REQUEUE', 'ALL'],
+        help='Mail address to watch test status.')
+    parser.add_argument(
+        '--quotatype',
+        default=None,
+        choices=['reserved', 'auto', 'spot'],
+        help='Quota type, only available for phoenix-slurm>=0.2')
+
+    args = parser.parse_args()
+    return args
+
+
+def create_test_job_batch(commands, model_info, args, port):
+
+    fname = model_info.name
+
+    config = Path(model_info.config)
+    # assert config.exists(), f'{fname}: {config} not found.'
+
+    http_prefix = 'https://download.openmmlab.com/mmrazor/'
+    if 's3://' in args.checkpoint_root:
+        from mmcv.fileio import FileClient
+        from petrel_client.common.exception import AccessDeniedError
+        file_client = FileClient.infer_client(uri=args.checkpoint_root)
+        checkpoint = file_client.join_path(
+            args.checkpoint_root, model_info.weights[len(http_prefix):])
+
+        try:
+            exists = file_client.exists(checkpoint)
+        except AccessDeniedError:
+            exists = False
+    else:
+        checkpoint_root = Path(args.checkpoint_root)
+        checkpoint = checkpoint_root / model_info.weights[len(http_prefix):]
+        checkpoint.parent.mkdir(parents=True, exist_ok=True)
+        exists = checkpoint.exists()
+    if exists:
+        print(f'{checkpoint} already exists.')
+    else:
+        wget.download(model_info.weights, str(checkpoint))
+        print(f'\nSaved in {checkpoint}.')
+
+    job_name = f'{args.job_name}_{fname}'
+    work_dir = Path(args.work_dir) / fname
+    work_dir.mkdir(parents=True, exist_ok=True)
+
+    if args.quotatype is not None:
+        quota_cfg = f'#SBATCH --quotatype {args.quotatype}\n'
+    else:
+        quota_cfg = ''
+
+    launcher = 'none' if args.local else 'slurm'
+    runner = 'python' if args.local else 'srun python'
+    master_port = f'NASTER_PORT={port}'
+
+    script_name = osp.join('tools', 'test.py')
+    job_script = (
+        f'#!/bin/bash\n'
+        f'#SBATCH --output {work_dir}/job.%j.out\n'
+        f'#SBATCH --partition={args.partition}\n'
+        f'#SBATCH --job-name {job_name}\n'
+        f'#SBATCH --gres=gpu:{args.gpus}\n'
+        f'{quota_cfg}'
+        f'#SBATCH --ntasks-per-node={args.gpus}\n'
+        f'#SBATCH --ntasks={args.gpus}\n'
+        f'#SBATCH --cpus-per-task=5\n\n'
+        f'{master_port} {runner} -u {script_name} {config} {checkpoint} '
+        f'--work-dir {work_dir} '
+        f'--launcher={launcher}\n')
+
+    with open(work_dir / 'job.sh', 'w') as f:
+        f.write(job_script)
+
+    commands.append(f'echo "{config}"')
+    if args.local:
+        commands.append(f'bash {work_dir}/job.sh')
+    else:
+        commands.append(f'sbatch {work_dir}/job.sh')
+
+    return work_dir / 'job.sh'
+
+
+def summary(args):
+    # parse model-index.yml
+    model_index_file = MMRAZOR_ROOT / 'model-index.yml'
+    model_index = load(str(model_index_file))
+    model_index.build_models_with_collections()
+    models = OrderedDict({model.name: model for model in model_index.models})
+
+    if args.models:
+        patterns = [re.compile(pattern) for pattern in args.models]
+        filter_models = {}
+        for k, v in models.items():
+            if any([re.match(pattern, k) for pattern in patterns]):
+                filter_models[k] = v
+        if len(filter_models) == 0:
+            print('No model found, please specify models in:')
+            print('\n'.join(models.keys()))
+            return
+        models = filter_models
+
+    model_results = dict()
+    for model_info in models.values():
+        model_name = model_info.name
+        work_dir = Path(args.work_dir) / model_name
+        sub_dirs = [p.name for p in work_dir.iterdir() if p.is_dir()]
+
+        if len(sub_dirs) == 0:
+            print(f'{model_name} has no results.')
+            continue
+
+        latest_time = sub_dirs[-1]
+        latest_json = work_dir / latest_time / f'{latest_time}.json'
+
+        if not latest_json.exists():
+            print(f'{model_name} has no results.')
+            continue
+        latest_result = mmcv.load(latest_json, 'json')
+
+        expect_result = model_info.results[0].metrics
+        summary_result = {
+            'expect': expect_result,
+            'actual':
+            {METRIC_MAPPINGS[k]: v
+             for k, v in latest_result.items()}
+        }
+        model_results[model_name] = summary_result
+
+    mmcv.fileio.dump(model_results,
+                     Path(args.work_dir) / 'summary.yml', 'yaml')
+    print(f'Summary results saved in {Path(args.work_dir)}/summary.yml')
+
+
+def test(args):
+    # parse model-index.yml
+    model_index_file = MMRAZOR_ROOT / 'model-index.yml'
+    model_index = load(str(model_index_file))
+    model_index.build_models_with_collections()
+    models = OrderedDict({model.name: model for model in model_index.models})
+
+    commands = []
+    if args.models:
+        patterns = [re.compile(pattern) for pattern in args.models]
+        filter_models = {}
+        for k, v in models.items():
+            if any([re.match(pattern, k) for pattern in patterns]):
+                filter_models[k] = v
+        if len(filter_models) == 0:
+            print('No model found, please specify models in:')
+            print('\n'.join(models.keys()))
+            return
+        models = filter_models
+
+    preview_script = ''
+    port = args.port
+    for model_info in models.values():
+        script_path = create_test_job_batch(commands, model_info, args, port)
+        preview_script = script_path or preview_script
+        port += 1
+    command_str = '\n'.join(commands)
+
+    preview = Table()
+    preview.add_column(str(preview_script))
+    preview.add_column('Shell command preview')
+    preview.add_row(
+        Syntax.from_path(
+            preview_script,
+            background_color='default',
+            line_numbers=True,
+            word_wrap=True),
+        Syntax(
+            command_str,
+            'bash',
+            background_color='default',
+            line_numbers=True,
+            word_wrap=True))
+    console.print(preview)
+
+    if args.run:
+        os.system(command_str)
+    else:
+        console.print('Please set "--run" to start the job')
+
+
+def main():
+    args = parse_args()
+    if args.summary:
+        summary(args)
+    else:
+        test(args)
+
+
+if __name__ == '__main__':
+    main()

.dev_scripts/benchmark_train.py

@@ -0,0 +1,224 @@
+import argparse
+import os
+import os.path as osp
+import re
+from collections import OrderedDict
+from pathlib import Path
+
+import mmcv
+from modelindex.load_model_index import load
+from rich.console import Console
+from rich.syntax import Syntax
+from rich.table import Table
+
+console = Console()
+MMRAZOR_ROOT = Path(__file__).absolute().parents[1]
+
+METRIC_MAPPINGS = {
+    'accuracy/top1': 'Top 1 Accuracy',
+    'accuracy/top5': 'Top 5 Accuracy'
+}
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(
+        description="Test all models' accuracy in model-index.yml")
+    parser.add_argument(
+        'partition', type=str, help='Cluster partition to use.')
+    parser.add_argument(
+        '--job-name',
+        type=str,
+        default='razor-train-benchmark',
+        help='Slurm job name prefix')
+    parser.add_argument('--port', type=int, default=29666, help='dist port')
+    parser.add_argument(
+        '--models', nargs='+', type=str, help='Specify model names to run.')
+    parser.add_argument('--gpus', type=int, default=8, help='num gpus')
+    parser.add_argument(
+        '--work-dir',
+        default='work_dirs/benchmark_test',
+        help='the dir to save metric')
+    parser.add_argument(
+        '--run', action='store_true', help='run script directly')
+    parser.add_argument(
+        '--summary', action='store_true', help='collect results')
+    parser.add_argument(
+        '--local',
+        action='store_true',
+        help='run at local instead of cluster.')
+    parser.add_argument(
+        '--mail', type=str, help='Mail address to watch test status.')
+    parser.add_argument(
+        '--mail-type',
+        nargs='+',
+        default=['BEGIN'],
+        choices=['NONE', 'BEGIN', 'END', 'FAIL', 'REQUEUE', 'ALL'],
+        help='Mail address to watch test status.')
+    parser.add_argument(
+        '--quotatype',
+        default=None,
+        choices=['reserved', 'auto', 'spot'],
+        help='Quota type, only available for phoenix-slurm>=0.2')
+
+    args = parser.parse_args()
+    return args
+
+
+def create_train_job_batch(commands, model_info, args, port):
+
+    fname = model_info.name
+
+    config = Path(model_info.config)
+    # assert config.exists(), f'{fname}: {config} not found.'
+
+    job_name = f'{args.job_name}_{fname}'
+    work_dir = Path(args.work_dir) / fname
+    work_dir.mkdir(parents=True, exist_ok=True)
+
+    if args.quotatype is not None:
+        quota_cfg = f'#SBATCH --quotatype {args.quotatype}\n'
+    else:
+        quota_cfg = ''
+
+    launcher = 'none' if args.local else 'slurm'
+    runner = 'python' if args.local else 'srun python'
+    master_port = f'NASTER_PORT={port}'
+
+    script_name = osp.join('tools', 'train.py')
+    job_script = (f'#!/bin/bash\n'
+                  f'#SBATCH --output {work_dir}/job.%j.out\n'
+                  f'#SBATCH --partition={args.partition}\n'
+                  f'#SBATCH --job-name {job_name}\n'
+                  f'#SBATCH --gres=gpu:{args.gpus}\n'
+                  f'{quota_cfg}'
+                  f'#SBATCH --ntasks-per-node={args.gpus}\n'
+                  f'#SBATCH --ntasks={args.gpus}\n'
+                  f'#SBATCH --cpus-per-task=5\n\n'
+                  f'{master_port} {runner} -u {script_name} {config} '
+                  f'--work-dir {work_dir} '
+                  f'--launcher={launcher}\n')
+
+    with open(work_dir / 'job.sh', 'w') as f:
+        f.write(job_script)
+
+    commands.append(f'echo "{config}"')
+    if args.local:
+        commands.append(f'bash {work_dir}/job.sh')
+    else:
+        commands.append(f'sbatch {work_dir}/job.sh')
+
+    return work_dir / 'job.sh'
+
+
+def summary(args):
+    # parse model-index.yml
+    model_index_file = MMRAZOR_ROOT / 'model-index.yml'
+    model_index = load(str(model_index_file))
+    model_index.build_models_with_collections()
+    models = OrderedDict({model.name: model for model in model_index.models})
+
+    if args.models:
+        patterns = [re.compile(pattern) for pattern in args.models]
+        filter_models = {}
+        for k, v in models.items():
+            if any([re.match(pattern, k) for pattern in patterns]):
+                filter_models[k] = v
+        if len(filter_models) == 0:
+            print('No model found, please specify models in:')
+            print('\n'.join(models.keys()))
+            return
+        models = filter_models
+
+    model_results = dict()
+    for model_info in models.values():
+        model_name = model_info.name
+        work_dir = Path(args.work_dir) / model_name
+        sub_dirs = [p.name for p in work_dir.iterdir() if p.is_dir()]
+
+        if len(sub_dirs) == 0:
+            print(f'{model_name} has no results.')
+            continue
+
+        latest_time = sub_dirs[-1]
+        latest_json = work_dir / latest_time / f'{latest_time}.json'
+
+        if not latest_json.exists():
+            print(f'{model_name} has no results.')
+            continue
+        latest_result = mmcv.load(latest_json, 'json')
+
+        expect_result = model_info.results[0].metrics
+        summary_result = {
+            'expect': expect_result,
+            'actual':
+            {METRIC_MAPPINGS[k]: v
+             for k, v in latest_result.items()}
+        }
+        model_results[model_name] = summary_result
+
+    mmcv.fileio.dump(model_results,
+                     Path(args.work_dir) / 'summary.yml', 'yaml')
+    print(f'Summary results saved in {Path(args.work_dir)}/summary.yml')
+
+
+def train(args):
+    # parse model-index.yml
+    model_index_file = MMRAZOR_ROOT / 'model-index.yml'
+    model_index = load(str(model_index_file))
+    model_index.build_models_with_collections()
+    models = OrderedDict({model.name: model for model in model_index.models})
+
+    commands = []
+    if args.models:
+        patterns = [re.compile(pattern) for pattern in args.models]
+        filter_models = {}
+        for k, v in models.items():
+            if any([re.match(pattern, k) for pattern in patterns]):
+                filter_models[k] = v
+        if len(filter_models) == 0:
+            print('No model found, please specify models in:')
+            print('\n'.join(models.keys()))
+            return
+        models = filter_models
+
+    preview_script = ''
+    port = args.port
+    for model_info in models.values():
+        script_path = create_train_job_batch(commands, model_info, args, port)
+        preview_script = script_path or preview_script
+        port += 1
+    command_str = '\n'.join(commands)
+
+    preview = Table()
+    preview.add_column(str(preview_script))
+    preview.add_column('Shell command preview')
+    preview.add_row(
+        Syntax.from_path(
+            preview_script,
+            background_color='default',
+            line_numbers=True,
+            word_wrap=True),
+        Syntax(
+            command_str,
+            'bash',
+            background_color='default',
+            line_numbers=True,
+            word_wrap=True))
+    console.print(preview)
+
+    if args.run:
+        os.system(command_str)
+    else:
+        console.print('Please set "--run" to start the job')
+
+
+def main():
+    args = parse_args()
+    if args.summary:
+        summary(args)
+    else:
+        train(args)
+
+
+if __name__ == '__main__':
+    main()

configs/_base_/models/arch_settings/mobilenet/cream_114.py

@@ -1,76 +0,0 @@
-se_cfg = dict(
-    ratio=4,
-    divisor=1,
-    act_cfg=(dict(type='HSwish'),
-             dict(
-                 type='HSigmoid', bias=3, divisor=6, min_value=0,
-                 max_value=1)))
-
-_FIRST_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-_OTHER_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k3e6_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e4_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e6_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e4_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e6_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-arch_setting = [
-    # Parameters to build layers. 4 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, stride, mutable cfg.
-    [16, 1, 1, _FIRST_STAGE_MUTABLE],
-    [24, 1, 2, _OTHER_STAGE_MUTABLE],
-    [40, 2, 2, _OTHER_STAGE_MUTABLE],
-    [80, 2, 2, _OTHER_STAGE_MUTABLE],
-    [96, 3, 1, _OTHER_STAGE_MUTABLE],
-    [192, 2, 2, _OTHER_STAGE_MUTABLE],
-    [320, 1, 1, _OTHER_STAGE_MUTABLE]
-]

configs/_base_/models/arch_settings/mobilenet/cream_14.py

@@ -1,76 +0,0 @@
-se_cfg = dict(
-    ratio=4,
-    divisor=1,
-    act_cfg=(dict(type='HSwish'),
-             dict(
-                 type='HSigmoid', bias=3, divisor=6, min_value=0,
-                 max_value=1)))
-
-_FIRST_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-_OTHER_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k3e6_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e4_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e6_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e4_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e6_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-arch_setting = [
-    # Parameters to build layers. 4 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, stride, mutable cfg.
-    [16, 1, 1, _FIRST_STAGE_MUTABLE],
-    [24, 1, 2, _OTHER_STAGE_MUTABLE],
-    [40, 2, 2, _OTHER_STAGE_MUTABLE],
-    [80, 2, 2, _OTHER_STAGE_MUTABLE],
-    [96, 1, 1, _OTHER_STAGE_MUTABLE],
-    [192, 1, 2, _OTHER_STAGE_MUTABLE],
-    [320, 1, 1, _OTHER_STAGE_MUTABLE]
-]

configs/_base_/models/arch_settings/mobilenet/cream_287.py

@@ -1,76 +0,0 @@
-se_cfg = dict(
-    ratio=4,
-    divisor=1,
-    act_cfg=(dict(type='HSwish'),
-             dict(
-                 type='HSigmoid', bias=3, divisor=6, min_value=0,
-                 max_value=1)))
-
-_FIRST_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-_OTHER_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k3e6_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e4_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e6_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e4_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e6_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-arch_setting = [
-    # Parameters to build layers. 4 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, stride, mutable cfg.
-    [16, 1, 1, _FIRST_STAGE_MUTABLE],
-    [24, 1, 2, _OTHER_STAGE_MUTABLE],
-    [40, 2, 2, _OTHER_STAGE_MUTABLE],
-    [80, 3, 2, _OTHER_STAGE_MUTABLE],
-    [96, 4, 1, _OTHER_STAGE_MUTABLE],
-    [192, 3, 2, _OTHER_STAGE_MUTABLE],
-    [320, 1, 1, _OTHER_STAGE_MUTABLE]
-]

configs/_base_/models/arch_settings/mobilenet/cream_43.py

@@ -1,76 +0,0 @@
-se_cfg = dict(
-    ratio=4,
-    divisor=1,
-    act_cfg=(dict(type='HSwish'),
-             dict(
-                 type='HSigmoid', bias=3, divisor=6, min_value=0,
-                 max_value=1)))
-
-_FIRST_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-_OTHER_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k3e6_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e4_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e6_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e4_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e6_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-arch_setting = [
-    # Parameters to build layers. 4 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, stride, mutable cfg.
-    [16, 1, 1, _FIRST_STAGE_MUTABLE],
-    [24, 1, 2, _OTHER_STAGE_MUTABLE],
-    [40, 2, 2, _OTHER_STAGE_MUTABLE],
-    [80, 2, 2, _OTHER_STAGE_MUTABLE],
-    [96, 3, 1, _OTHER_STAGE_MUTABLE],
-    [192, 2, 2, _OTHER_STAGE_MUTABLE],
-    [320, 1, 1, _OTHER_STAGE_MUTABLE]
-]

configs/_base_/models/arch_settings/mobilenet/cream_481.py

@@ -1,76 +0,0 @@
-se_cfg = dict(
-    ratio=4,
-    divisor=1,
-    act_cfg=(dict(type='HSwish'),
-             dict(
-                 type='HSigmoid', bias=3, divisor=6, min_value=0,
-                 max_value=1)))
-
-_FIRST_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-_OTHER_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k3e6_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e4_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e6_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e4_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e6_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-arch_setting = [
-    # Parameters to build layers. 4 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, stride, mutable cfg.
-    [16, 1, 1, _FIRST_STAGE_MUTABLE],
-    [24, 4, 2, _OTHER_STAGE_MUTABLE],
-    [40, 4, 2, _OTHER_STAGE_MUTABLE],
-    [80, 5, 2, _OTHER_STAGE_MUTABLE],
-    [96, 4, 1, _OTHER_STAGE_MUTABLE],
-    [192, 4, 2, _OTHER_STAGE_MUTABLE],
-    [320, 1, 1, _OTHER_STAGE_MUTABLE]
-]

configs/_base_/models/arch_settings/mobilenet/cream_604.py

@@ -1,76 +0,0 @@
-se_cfg = dict(
-    ratio=4,
-    divisor=1,
-    act_cfg=(dict(type='HSwish'),
-             dict(
-                 type='HSigmoid', bias=3, divisor=6, min_value=0,
-                 max_value=1)))
-
-_FIRST_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-_OTHER_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k3e6_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e4_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k5e6_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e4_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish')),
-        mb_k7e6_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='HSwish'))))
-
-arch_setting = [
-    # Parameters to build layers. 4 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, stride, mutable cfg.
-    [16, 1, 1, _FIRST_STAGE_MUTABLE],
-    [24, 5, 2, _OTHER_STAGE_MUTABLE],
-    [40, 5, 2, _OTHER_STAGE_MUTABLE],
-    [80, 5, 2, _OTHER_STAGE_MUTABLE],
-    [96, 6, 1, _OTHER_STAGE_MUTABLE],
-    [192, 6, 2, _OTHER_STAGE_MUTABLE],
-    [320, 1, 1, _OTHER_STAGE_MUTABLE]
-]

configs/_base_/nas_backbones/spos_mobilenet_supernet.py

@@ -0,0 +1,65 @@
+_STAGE_MUTABLE = dict(
+    _scope_='mmrazor',
+    type='OneShotMutableOP',
+    candidates=dict(
+        mb_k3e3=dict(
+            type='MBBlock',
+            kernel_size=3,
+            expand_ratio=3,
+            act_cfg=dict(type='ReLU6')),
+        mb_k5e3=dict(
+            type='MBBlock',
+            kernel_size=5,
+            expand_ratio=3,
+            act_cfg=dict(type='ReLU6')),
+        mb_k7e3=dict(
+            type='MBBlock',
+            kernel_size=7,
+            expand_ratio=3,
+            act_cfg=dict(type='ReLU6')),
+        mb_k3e6=dict(
+            type='MBBlock',
+            kernel_size=3,
+            expand_ratio=6,
+            act_cfg=dict(type='ReLU6')),
+        mb_k5e6=dict(
+            type='MBBlock',
+            kernel_size=5,
+            expand_ratio=6,
+            act_cfg=dict(type='ReLU6')),
+        mb_k7e6=dict(
+            type='MBBlock',
+            kernel_size=7,
+            expand_ratio=6,
+            act_cfg=dict(type='ReLU6')),
+        identity=dict(type='Identity')))
+
+_FIRST_MUTABLE = dict(
+    _scope_='mmrazor',
+    type='OneShotMutableOP',
+    candidates=dict(
+        mb_k3e1=dict(
+            type='MBBlock',
+            kernel_size=3,
+            expand_ratio=1,
+            act_cfg=dict(type='ReLU6'))))
+
+arch_setting = [
+    # Parameters to build layers. 3 parameters are needed to construct a
+    # layer, from left to right: channel, num_blocks, mutable_cfg.
+    [24, 1, 1, _FIRST_MUTABLE],
+    [32, 4, 2, _STAGE_MUTABLE],
+    [56, 4, 2, _STAGE_MUTABLE],
+    [112, 4, 2, _STAGE_MUTABLE],
+    [128, 4, 1, _STAGE_MUTABLE],
+    [256, 4, 2, _STAGE_MUTABLE],
+    [432, 1, 1, _STAGE_MUTABLE]
+]
+
+nas_backbone = dict(
+    _scope_='mmrazor',
+    type='SearchableMobileNet',
+    first_channels=40,
+    last_channels=1728,
+    widen_factor=1.0,
+    arch_setting=arch_setting)

configs/_base_/nas_backbones/spos_shufflenet_supernet.py

@@ -0,0 +1,23 @@
+_STAGE_MUTABLE = dict(
+    _scope_='mmrazor',
+    type='OneShotMutableOP',
+    candidates=dict(
+        shuffle_3x3=dict(type='ShuffleBlock', kernel_size=3),
+        shuffle_5x5=dict(type='ShuffleBlock', kernel_size=5),
+        shuffle_7x7=dict(type='ShuffleBlock', kernel_size=7),
+        shuffle_xception=dict(type='ShuffleXception')))
+
+arch_setting = [
+    # Parameters to build layers. 3 parameters are needed to construct a
+    # layer, from left to right: channel, num_blocks, mutable_cfg.
+    [64, 4, _STAGE_MUTABLE],
+    [160, 4, _STAGE_MUTABLE],
+    [320, 8, _STAGE_MUTABLE],
+    [640, 4, _STAGE_MUTABLE]
+]
+
+nas_backbone = dict(
+    _scope_='mmrazor',
+    type='SearchableShuffleNetV2',
+    widen_factor=1.0,
+    arch_setting=arch_setting)

configs/_base_/settings/imagenet_bs1024_spos.py

@@ -0,0 +1,80 @@
+# dataset settings
+dataset_type = 'mmcls.ImageNet'
+preprocess_cfg = dict(
+    # RGB format normalization parameters
+    mean=[123.675, 116.28, 103.53],
+    std=[58.395, 57.12, 57.375],
+    # convert image from BGR to RGB
+    to_rgb=True,
+)
+
+train_pipeline = [
+    dict(type='mmcls.LoadImageFromFile'),
+    dict(type='mmcls.RandomResizedCrop', scale=224),
+    dict(
+        type='mmcls.ColorJitter', brightness=0.4, contrast=0.4,
+        saturation=0.4),
+    dict(type='mmcls.RandomFlip', prob=0.5, direction='horizontal'),
+    dict(type='mmcls.PackClsInputs'),
+]
+
+test_pipeline = [
+    dict(type='mmcls.LoadImageFromFile'),
+    dict(type='mmcls.ResizeEdge', scale=256, edge='short'),
+    dict(type='mmcls.CenterCrop', crop_size=224),
+    dict(type='mmcls.PackClsInputs'),
+]
+
+train_dataloader = dict(
+    batch_size=128,
+    num_workers=4,
+    dataset=dict(
+        type=dataset_type,
+        data_root='data/imagenet',
+        ann_file='meta/train.txt',
+        data_prefix='train',
+        pipeline=train_pipeline),
+    sampler=dict(type='mmcls.DefaultSampler', shuffle=True),
+    persistent_workers=True,
+)
+
+val_dataloader = dict(
+    batch_size=128,
+    num_workers=4,
+    dataset=dict(
+        type=dataset_type,
+        data_root='data/imagenet',
+        ann_file='meta/val.txt',
+        data_prefix='val',
+        pipeline=test_pipeline),
+    sampler=dict(type='mmcls.DefaultSampler', shuffle=False),
+    persistent_workers=True,
+)
+val_evaluator = dict(type='mmcls.Accuracy', topk=(1, 5))
+
+# If you want standard test, please manually configure the test dataset
+test_dataloader = val_dataloader
+test_evaluator = val_evaluator
+
+# optimizer
+paramwise_cfg = dict(
+    bias_decay_mult=0.0, norm_decay_mult=0.0, dwconv_decay_mult=0.0)
+
+optim_wrapper = dict(
+    optimizer=dict(type='mmcls.SGD', lr=0.5, momentum=0.9, weight_decay=4e-5),
+    paramwise_cfg=paramwise_cfg,
+    clip_grad=None)
+
+# leanring policy
+param_scheduler = dict(
+    type='mmcls.PolyLR',
+    power=1.0,
+    eta_min=0.0,
+    by_epoch=True,
+    end=300,
+    convert_to_iter_based=True)
+
+# train, val, test setting
+train_cfg = dict(by_epoch=True, max_epochs=300)
+val_cfg = dict()
+test_cfg = dict()

configs/_base_/settings/imagenet_bs2048_autoslim.py

@@ -0,0 +1,9 @@
+_base_ = [
+    './imagenet_bs1024_spos.py',
+]
+
+train_dataloader = dict(batch_size=256)
+
+val_dataloader = dict(batch_size=256)
+
+test_dataloader = dict(batch_size=256)

configs/distill/cwd/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco.py

@@ -1,144 +0,0 @@
-_base_ = [
-    '../../_base_/datasets/mmdet/coco_detection.py',
-    '../../_base_/schedules/mmdet/schedule_1x.py',
-    '../../_base_/mmdet_runtime.py'
-]
-
-# model settings
-student = dict(
-    type='mmdet.GFL',
-    backbone=dict(
-        type='ResNet',
-        depth=50,
-        num_stages=4,
-        out_indices=(0, 1, 2, 3),
-        frozen_stages=1,
-        norm_cfg=dict(type='BN', requires_grad=True),
-        norm_eval=True,
-        style='pytorch',
-        init_cfg=dict(type='Pretrained', checkpoint='torchvision://resnet50')),
-    neck=dict(
-        type='FPN',
-        in_channels=[256, 512, 1024, 2048],
-        out_channels=256,
-        start_level=1,
-        add_extra_convs='on_output',
-        num_outs=5),
-    bbox_head=dict(
-        type='GFLHead',
-        num_classes=80,
-        in_channels=256,
-        stacked_convs=4,
-        feat_channels=256,
-        anchor_generator=dict(
-            type='AnchorGenerator',
-            ratios=[1.0],
-            octave_base_scale=8,
-            scales_per_octave=1,
-            strides=[8, 16, 32, 64, 128]),
-        loss_cls=dict(
-            type='QualityFocalLoss',
-            use_sigmoid=True,
-            beta=2.0,
-            loss_weight=1.0),
-        loss_dfl=dict(type='DistributionFocalLoss', loss_weight=0.25),
-        reg_max=16,
-        loss_bbox=dict(type='GIoULoss', loss_weight=2.0)),
-    # training and testing settings
-    train_cfg=dict(
-        assigner=dict(type='ATSSAssigner', topk=9),
-        allowed_border=-1,
-        pos_weight=-1,
-        debug=False),
-    test_cfg=dict(
-        nms_pre=1000,
-        min_bbox_size=0,
-        score_thr=0.05,
-        nms=dict(type='nms', iou_threshold=0.6),
-        max_per_img=100))
-
-checkpoint = 'https://download.openmmlab.com/mmdetection/v2.0/gfl/gfl_r101_fpn_mstrain_2x_coco/gfl_r101_fpn_mstrain_2x_coco_20200629_200126-dd12f847.pth'  # noqa: E501
-
-teacher = dict(
-    type='mmdet.GFL',
-    init_cfg=dict(type='Pretrained', checkpoint=checkpoint),
-    backbone=dict(
-        type='ResNet',
-        depth=101,
-        num_stages=4,
-        out_indices=(0, 1, 2, 3),
-        frozen_stages=1,
-        norm_cfg=dict(type='BN', requires_grad=True),
-        norm_eval=True,
-        style='pytorch',
-        init_cfg=None),
-    neck=dict(
-        type='FPN',
-        in_channels=[256, 512, 1024, 2048],
-        out_channels=256,
-        start_level=1,
-        add_extra_convs='on_output',
-        num_outs=5),
-    bbox_head=dict(
-        type='GFLHead',
-        num_classes=80,
-        in_channels=256,
-        stacked_convs=4,
-        feat_channels=256,
-        anchor_generator=dict(
-            type='AnchorGenerator',
-            ratios=[1.0],
-            octave_base_scale=8,
-            scales_per_octave=1,
-            strides=[8, 16, 32, 64, 128]),
-        loss_cls=dict(
-            type='QualityFocalLoss',
-            use_sigmoid=True,
-            beta=2.0,
-            loss_weight=1.0),
-        loss_dfl=dict(type='DistributionFocalLoss', loss_weight=0.25),
-        reg_max=16,
-        loss_bbox=dict(type='GIoULoss', loss_weight=2.0)),
-    # training and testing settings
-    train_cfg=dict(
-        assigner=dict(type='ATSSAssigner', topk=9),
-        allowed_border=-1,
-        pos_weight=-1,
-        debug=False),
-    test_cfg=dict(
-        nms_pre=1000,
-        min_bbox_size=0,
-        score_thr=0.05,
-        nms=dict(type='nms', iou_threshold=0.6),
-        max_per_img=100))
-
-# algorithm setting
-algorithm = dict(
-    type='GeneralDistill',
-    architecture=dict(
-        type='MMDetArchitecture',
-        model=student,
-    ),
-    distiller=dict(
-        type='SingleTeacherDistiller',
-        teacher=teacher,
-        teacher_trainable=False,
-        components=[
-            dict(
-                student_module='bbox_head.gfl_cls',
-                teacher_module='bbox_head.gfl_cls',
-                losses=[
-                    dict(
-                        type='ChannelWiseDivergence',
-                        name='loss_cwd_cls_head',
-                        tau=1,
-                        loss_weight=5,
-                    )
-                ])
-        ]),
-)
-
-find_unused_parameters = True
-
-# optimizer
-optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0001)

configs/distill/cwd/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k.py

@@ -1,110 +0,0 @@
-_base_ = [
-    '../../_base_/datasets/mmseg/cityscapes.py',
-    '../../_base_/mmseg_runtime.py',
-    '../../_base_/schedules/mmseg/schedule_80k.py'
-]
-
-norm_cfg = dict(type='SyncBN', requires_grad=True)
-
-# pspnet r18
-student = dict(
-    type='mmseg.EncoderDecoder',
-    backbone=dict(
-        type='ResNetV1c',
-        init_cfg=dict(
-            type='Pretrained', checkpoint='open-mmlab://resnet18_v1c'),
-        depth=18,
-        num_stages=4,
-        out_indices=(0, 1, 2, 3),
-        dilations=(1, 1, 2, 4),
-        strides=(1, 2, 1, 1),
-        norm_cfg=norm_cfg,
-        norm_eval=False,
-        style='pytorch',
-        contract_dilation=True),
-    decode_head=dict(
-        type='PSPHead',
-        in_channels=512,
-        in_index=3,
-        channels=128,
-        pool_scales=(1, 2, 3, 6),
-        dropout_ratio=0.1,
-        num_classes=19,
-        norm_cfg=norm_cfg,
-        align_corners=False,
-        loss_decode=dict(
-            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
-    auxiliary_head=dict(
-        type='FCNHead',
-        in_channels=256,
-        in_index=2,
-        channels=64,
-        num_convs=1,
-        concat_input=False,
-        dropout_ratio=0.1,
-        num_classes=19,
-        norm_cfg=norm_cfg,
-        align_corners=False,
-        loss_decode=dict(
-            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=0.4)),
-    train_cfg=dict(),
-    test_cfg=dict(mode='whole'))
-
-checkpoint = 'https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_512x1024_80k_cityscapes/pspnet_r101-d8_512x1024_80k_cityscapes_20200606_112211-e1e1100f.pth'  # noqa: E501
-
-# pspnet r101
-teacher = dict(
-    type='mmseg.EncoderDecoder',
-    init_cfg=dict(type='Pretrained', checkpoint=checkpoint),
-    backbone=dict(
-        type='ResNetV1c',
-        depth=101,
-        num_stages=4,
-        out_indices=(0, 1, 2, 3),
-        dilations=(1, 1, 2, 4),
-        strides=(1, 2, 1, 1),
-        norm_cfg=norm_cfg,
-        norm_eval=False,
-        style='pytorch',
-        contract_dilation=True),
-    decode_head=dict(
-        type='PSPHead',
-        in_channels=2048,
-        in_index=3,
-        channels=512,
-        pool_scales=(1, 2, 3, 6),
-        dropout_ratio=0.1,
-        num_classes=19,
-        norm_cfg=norm_cfg,
-        align_corners=False,
-        loss_decode=dict(
-            type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0)),
-)
-
-# algorithm setting
-algorithm = dict(
-    type='GeneralDistill',
-    architecture=dict(
-        type='MMSegArchitecture',
-        model=student,
-    ),
-    distiller=dict(
-        type='SingleTeacherDistiller',
-        teacher=teacher,
-        teacher_trainable=False,
-        components=[
-            dict(
-                student_module='decode_head.conv_seg',
-                teacher_module='decode_head.conv_seg',
-                losses=[
-                    dict(
-                        type='ChannelWiseDivergence',
-                        name='loss_cwd_logits',
-                        tau=1,
-                        loss_weight=5,
-                    )
-                ])
-        ]),
-)
-
-find_unused_parameters = True

configs/distill/cwd/metafile.yml

@@ -1,50 +0,0 @@
-Collections:
-  - Name: CWD
-    Metadata:
-      Training Data:
-        - Cityscapes
-        - COCO
-    Paper:
-      URL: https://arxiv.org/abs/2011.13256
-      Title: Channel-wise Knowledge Distillation for Dense Prediction
-    README: configs/distill/cwd/README.md
-    Code:
-      URL: https://github.com/open-mmlab/mmrazor/blob/v0.1.0/mmrazor/models/losses/cwd.py#L10
-      Version: v0.1.0
-    Converted From:
-      Code:
-        - https://github.com/pppppM/mmsegmentation-distiller
-        - https://github.com/pppppM/mmdetection-distiller
-Models:
-  - Name: cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k
-    In Collection: CWD
-    Metadata:
-      Location: cls head
-      Student: pspnet-r18-d8
-      Teacher: pspnet-r101-d8
-      Teacher Checkpoint: https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_512x1024_80k_cityscapes/pspnet_r101-d8_512x1024_80k_cityscapes_20200606_112211-e1e1100f.pth
-    Results:
-      - Task: Semantic Segmentation
-        Dataset: Cityscapes
-        Metrics:
-          mIoU: 75.54
-          mIoU(S): 74.87
-          mIoU(T): 79.76
-    Config: configs/distill/cwd/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k.py
-    Weights: https://download.openmmlab.com/mmrazor/v0.1/distill/cwd/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k_mIoU-75.54_20211222-3a26ee1c.pth
-  - Name: cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco
-    In Collection: CWD
-    Metadata:
-      Location: cls head
-      Student: gfl-r50-fpn
-      Teacher: gfl-r101-fpn
-      Teacher Checkpoint: https://download.openmmlab.com/mmdetection/v2.0/gfl/gfl_r101_fpn_mstrain_2x_coco/gfl_r101_fpn_mstrain_2x_coco_20200629_200126-dd12f847.pth
-    Results:
-      - Task: Object Detection
-        Dataset: COCO
-        Metrics:
-          box AP: 41.9
-          box AP(S): 40.2
-          box AP(T): 44.7
-    Config: configs/distill/cwd/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco.py
-    Weights: https://download.openmmlab.com/mmrazor/v0.1/distill/cwd/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco_20211222-655dff39.pth

configs/distill/mmcls/kd/README.md

@@ -0,0 +1,32 @@
+# KD
+
+> [Distilling the Knowledge in a Neural Network](https://arxiv.org/abs/1503.02531)
+
+<!-- [ALGORITHM] -->
+
+## Abstract
+
+A very simple way to improve the performance of almost any machine learning algorithm is to train many different models on the same data and then to average their predictions. Unfortunately, making predictions using a whole ensemble of models is cumbersome and may be too computationally expensive to allow deployment to a large number of users, especially if the individual models are large neural nets. Caruana and his collaborators have shown that it is possible to compress the knowledge in an ensemble into a single model which is much easier to deploy and we develop this approach further using a different compression technique. We achieve some surprising results on MNIST and we show that we can significantly improve the acoustic model of a heavily used commercial system by distilling the knowledge in an ensemble of models into a single model. We also introduce a new type of ensemble composed of one or more full models and many specialist models which learn to distinguish fine-grained classes that the full models confuse. Unlike a mixture of experts, these specialist models can be trained rapidly and in parallel.
+
+![pipeline](/docs/en/imgs/model_zoo/kd/pipeline.png)
+
+## Results and models
+
+### Classification
+
+| Location | Dataset  |                                                   Teacher                                                    |                                                   Student                                                    |  Acc  | Acc(T) | Acc(S) |                          Config                           | Download                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     |
+| :------: | :------: | :----------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------: | :---: | :----: | :----: | :-------------------------------------------------------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+|  logits  | ImageNet | [resnet34](https://github.com/open-mmlab/mmclassification/blob/master/configs/resnet/resnet34_8xb32_in1k.py) | [resnet18](https://github.com/open-mmlab/mmclassification/blob/master/configs/resnet/resnet18_8xb32_in1k.py) | 71.54 | 73.62  | 69.90  | [config](./wsld_cls_head_resnet34_resnet18_8xb32_in1k.py) | [teacher](https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_8xb32_in1k_20210831-ea4938fc.pth) \|[model](https://openmmlab-share.oss-cn-hangzhou.aliyuncs.com/mmrazor/v0.1/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k/wsld_cls_head_resnet34_resnet18_8xb32_in1k_acc-71.54_20211222-91f28cf6.pth?versionId=CAEQHxiBgMC6memK7xciIGMzMDFlYTA4YzhlYTRiMTNiZWU0YTVhY2I5NjVkMjY2) \| [log](https://openmmlab-share.oss-cn-hangzhou.aliyuncs.com/mmrazor/v0.1/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k/wsld_cls_head_resnet34_resnet18_8xb32_in1k_20211221_181516.log.json?versionId=CAEQHxiBgIDLmemK7xciIGNkM2FiN2Y4N2E5YjRhNDE4NDVlNmExNDczZDIxN2E5) |
+
+## Citation
+
+```latex
+@article{hinton2015distilling,
+  title={Distilling the knowledge in a neural network},
+  author={Hinton, Geoffrey and Vinyals, Oriol and Dean, Jeff and others},
+  journal={arXiv preprint arXiv:1503.02531},
+  volume={2},
+  number={7},
+  year={2015}
+}
+```

configs/distill/mmcls/kd/kd_logits_r34_r18_8xb32_in1k.py

@@ -19,20 +19,18 @@ model = dict(
     teacher=dict(
         cfg_path='mmcls::resnet/resnet34_8xb32_in1k.py', pretrained=True),
     teacher_ckpt='resnet34_8xb32_in1k_20210831-f257d4e6.pth',
-    student_recorders=dict(fc=dict(type='ModuleOutputs', source='head.fc')),
-    teacher_recorders=dict(fc=dict(type='ModuleOutputs', source='head.fc')),
-    distill_losses=dict(
-        loss_kl=dict(type='KLDivergence', tau=1, loss_weight=5)),
-    loss_forward_mappings=dict(
-        loss_kl=dict(
-            preds_S=dict(
-                from_student=True,
-                recorder='fc',
-            ),
-            preds_T=dict(
-                from_student=False,
-                recorder='fc',
-            ))))
+    distiller=dict(
+        type='ConfigurableDistiller',
+        student_recorders=dict(
+            fc=dict(type='ModuleOutputs', source='head.fc')),
+        teacher_recorders=dict(
+            fc=dict(type='ModuleOutputs', source='head.fc')),
+        distill_losses=dict(
+            loss_kl=dict(type='KLDivergence', tau=1, loss_weight=5)),
+        loss_forward_mappings=dict(
+            loss_kl=dict(
+                preds_S=dict(from_student=True, recorder='fc'),
+                preds_T=dict(from_student=False, recorder='fc')))))
 
 find_unused_parameters = True
 

configs/distill/mmcls/kd/metafile.yml

@@ -0,0 +1,36 @@
+Collections:
+  - Name: KD
+    Metadata:
+      Training Data:
+        - ImageNet-1k
+    Paper:
+      URL: https://arxiv.org/abs/1503.02531
+      Title: Distilling the Knowledge in a Neural Network
+    README: configs/distill/mmcls/kd/README.md
+    Code:
+      URL: https://github.com/open-mmlab/mmrazor/blob/v0.1.0/mmrazor/models/losses/weighted_soft_label_distillation.py
+      Version: v0.1.0
+Models:
+  - Name: kd_logits_resnet34_resnet18_8xb32_in1k
+    In Collection: KD
+    Metadata:
+      Location: logits
+      Student:
+        Config: mmcls::resnet/resnet18_8xb32_in1k.py
+        Weights: https://download.openmmlab.com/mmclassification/v0/resnet/resnet18_8xb32_in1k_20210831-fbbb1da6.pth
+        Metrics:
+          Top 1 Accuracy: 69.90
+          Top 5 Accuracy: 89.43
+      Teacher:
+        Config: mmcls::resnet/resnet34_8xb32_in1k.py
+        Weights: https://download.openmmlab.com/mmclassification/v0/resnet/resnet34_8xb32_in1k_20210831-f257d4e6.pth
+        Metrics:
+          Top 1 Accuracy: 73.62
+          Top 5 Accuracy: 91.59
+    Results:
+      - Task: Image Classification
+        Dataset: ImageNet-1k
+        Metrics:
+          Top 1 Accuracy: 71.54
+    Config: configs/distill/mmcls/kd/kd_logits_resnet34_resnet18_8xb32_in1k.py
+    Weights: https://download.openmmlab.com/mmrazor/v0.1/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k/wsld_cls_head_resnet34_resnet18_8xb32_in1k_acc-71.54_20211222-91f28cf6.pth

configs/distill/mmcls/rkd/metafile.yml

@@ -6,7 +6,7 @@ Collections:
     Paper:
       URL: https://arxiv.org/abs/1904.05068
       Title: Relational Knowledge Distillation
-    README: configs/distill/rkd/README.md
+    README: configs/distill/mmcls/rkd/README.md
     Code:
       URL: https://github.com/open-mmlab/mmrazor/blob/v0.3.0/mmrazor/models/losses/relation_kd.py
       Version: v0.3.0
@@ -17,15 +17,22 @@ Models:
     In Collection: RKD
     Metadata:
       Location: neck
-      Student: R-18
-      Teacher: R-34
-      Teacher Checkpoint: https://download.openmmlab.com/mmclassification/v0/resnet/resnet34_8xb32_in1k_20210831-f257d4e6.pth
+      Student:
+        Config: mmcls::resnet/resnet18_8xb32_in1k.py
+        Weights: https://download.openmmlab.com/mmclassification/v0/resnet/resnet18_8xb32_in1k_20210831-fbbb1da6.pth
+        Metrics:
+          Top 1 Accuracy: 69.90
+          Top 5 Accuracy: 89.43
+      Teacher:
+        Config: mmcls::resnet/resnet34_8xb32_in1k.py
+        Weights: https://download.openmmlab.com/mmclassification/v0/resnet/resnet34_8xb32_in1k_20210831-f257d4e6.pth
+        Metrics:
+          Top 1 Accuracy: 73.62
+          Top 5 Accuracy: 91.59
     Results:
       - Task: Image Classification
         Dataset: ImageNet-1k
         Metrics:
           Top 1 Accuracy: 70.23
-          Top 1 Accuracy:(S): 69.90
-          Top 1 Accuracy:(T): 73.62
-    Config: configs/distill/rkd/rkd_neck_resnet34_resnet18_8xb32_in1k.py
+    Config: configs/distill/mmcls/rkd/rkd_neck_resnet34_resnet18_8xb32_in1k.py
     Weights: https://download.openmmlab.com/mmrazor/v0.3/distill/rkd/rkd_neck_resnet34_resnet18_8xb32_in1k_acc-70.23_20220401-f25700ac.pth

configs/distill/mmcls/rkd/rkd_neck_resnet34_resnet18_8xb32_in1k.py

@@ -0,0 +1,43 @@
+_base_ = [
+    'mmcls::_base_/datasets/imagenet_bs32.py',
+    'mmcls::_base_/schedules/imagenet_bs256.py',
+    'mmcls::_base_/default_runtime.py'
+]
+
+model = dict(
+    _scope_='mmrazor',
+    type='SingleTeacherDistill',
+    data_preprocessor=dict(
+        type='ImgDataPreprocessor',
+        # RGB format normalization parameters
+        mean=[123.675, 116.28, 103.53],
+        std=[58.395, 57.12, 57.375],
+        # convert image from BGR to RGB
+        bgr_to_rgb=True),
+    architecture=dict(
+        cfg_path='mmcls::resnet/resnet18_8xb32_in1k.py', pretrained=False),
+    teacher=dict(
+        cfg_path='mmcls::resnet/resnet34_8xb32_in1k.py', pretrained=True),
+    teacher_ckpt='resnet34_8xb32_in1k_20210831-f257d4e6.pth',
+    distiller=dict(
+        type='ConfigurableDistiller',
+        student_recorders=dict(
+            feat=dict(type='ModuleOutputs', source='neck.gap')),
+        teacher_recorders=dict(
+            feat=dict(type='ModuleOutputs', source='neck.gap')),
+        distill_losses=dict(
+            loss_dw=dict(
+                type='DistanceWiseRKD', with_l2_norm=True, loss_weight=25),
+            loss_aw=dict(
+                type='AngleWiseRKD', with_l2_norm=True, loss_weight=50)),
+        loss_forward_mappings=dict(
+            loss_dw=dict(
+                preds_S=dict(from_student=True, recorder='feat'),
+                preds_T=dict(from_student=False, recorder='feat')),
+            loss_aw=dict(
+                preds_S=dict(from_student=True, recorder='feat'),
+                preds_T=dict(from_student=False, recorder='feat')))))
+
+find_unused_parameters = True
+
+val_cfg = dict(_delete_=True, type='mmrazor.SingleTeacherDistillValLoop')

configs/distill/mmcls/wsld/README.md

@@ -27,9 +27,9 @@ effectiveness of our method.
 
 ### Classification
 
-| Location | Dataset  |                                                   Teacher                                                    |                                                   Student                                                    |  Acc  | Acc(T) | Acc(S) |                          Config                           | Download                                                                                                                                                                                                                                                                                                                                                                                                                                                                                   |
-| :------: | :------: | :----------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------: | :---: | :----: | :----: | :-------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
-| cls head | ImageNet | [resnet34](https://github.com/open-mmlab/mmclassification/blob/master/configs/resnet/resnet34_8xb32_in1k.py) | [resnet18](https://github.com/open-mmlab/mmclassification/blob/master/configs/resnet/resnet18_8xb32_in1k.py) | 71.54 | 73.62  | 69.90  | [config](./wsld_cls_head_resnet34_resnet18_8xb32_in1k.py) | [teacher](https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_8xb32_in1k_20210831-ea4938fc.pth) \|[model](https://download.openmmlab.com/mmrazor/v0.1/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k/wsld_cls_head_resnet34_resnet18_8xb32_in1k_acc-71.54_20211222-91f28cf6.pth) \| [log](https://download.openmmlab.com/mmrazor/v0.1/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k/wsld_cls_head_resnet34_resnet18_8xb32_in1k_20211221_181516.log.json) |
+| Location | Dataset  |                                                   Teacher                                                    |                                                   Student                                                    |  Acc  | Acc(T) | Acc(S) |                          Config                           | Download                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                     |
+| :------: | :------: | :----------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------: | :---: | :----: | :----: | :-------------------------------------------------------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| cls head | ImageNet | [resnet34](https://github.com/open-mmlab/mmclassification/blob/master/configs/resnet/resnet34_8xb32_in1k.py) | [resnet18](https://github.com/open-mmlab/mmclassification/blob/master/configs/resnet/resnet18_8xb32_in1k.py) | 71.54 | 73.62  | 69.90  | [config](./wsld_cls_head_resnet34_resnet18_8xb32_in1k.py) | [teacher](https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_8xb32_in1k_20210831-ea4938fc.pth) \|[model](https://openmmlab-share.oss-cn-hangzhou.aliyuncs.com/mmrazor/v0.1/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k/wsld_cls_head_resnet34_resnet18_8xb32_in1k_acc-71.54_20211222-91f28cf6.pth?versionId=CAEQHxiBgMC6memK7xciIGMzMDFlYTA4YzhlYTRiMTNiZWU0YTVhY2I5NjVkMjY2) \| [log](https://openmmlab-share.oss-cn-hangzhou.aliyuncs.com/mmrazor/v0.1/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k/wsld_cls_head_resnet34_resnet18_8xb32_in1k_20211221_181516.log.json?versionId=CAEQHxiBgIDLmemK7xciIGNkM2FiN2Y4N2E5YjRhNDE4NDVlNmExNDczZDIxN2E5) |
 
 ## Citation
 

configs/distill/mmcls/wsld/metafile.yml

@@ -6,26 +6,33 @@ Collections:
     Paper:
       URL: https://arxiv.org/abs/2102.00650
       Title: Rethinking Soft Labels for Knowledge Distillation:A Bias-Variance Tradeoff Perspective
-    README: configs/distill/wsld/README.md
+    README: configs/distill/mmcls/wsld/README.md
     Code:
       URL: https://github.com/open-mmlab/mmrazor/blob/v0.1.0/mmrazor/models/losses/weighted_soft_label_distillation.py
       Version: v0.1.0
     Converted From:
       Code: https://github.com/bellymonster/Weighted-Soft-Label-Distillation
 Models:
-  - Name: wsld_cls_head_resnet34_resnet18_8xb32_in1k
+  - Name: wsld_logits_resnet34_resnet18_8xb32_in1k
     In Collection: WSLD
     Metadata:
-      Location: cls head
-      Student: R-18
-      Teacher: R-34
-      Teacher Checkpoint: https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_8xb32_in1k_20210831-ea4938fc.pth
+      Location: logits
+      Student:
+        Config: mmcls::resnet/resnet18_8xb32_in1k.py
+        Weights: https://download.openmmlab.com/mmclassification/v0/resnet/resnet18_8xb32_in1k_20210831-fbbb1da6.pth
+        Metrics:
+          Top 1 Accuracy: 69.90
+          Top 5 Accuracy: 89.43
+      Teacher:
+        Config: mmcls::resnet/resnet34_8xb32_in1k.py
+        Weights: https://download.openmmlab.com/mmclassification/v0/resnet/resnet34_8xb32_in1k_20210831-f257d4e6.pth
+        Metrics:
+          Top 1 Accuracy: 73.62
+          Top 5 Accuracy: 91.59
     Results:
       - Task: Image Classification
         Dataset: ImageNet-1k
         Metrics:
           Top 1 Accuracy: 71.54
-          Top 1 Accuracy:(S): 69.90
-          Top 1 Accuracy:(T): 73.62
-    Config: configs/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k.py
+    Config: configs/distill/mmcls/wsld/wsld_logits_resnet34_resnet18_8xb32_in1k.py
     Weights: https://download.openmmlab.com/mmrazor/v0.1/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k/wsld_cls_head_resnet34_resnet18_8xb32_in1k_acc-71.54_20211222-91f28cf6.pth

configs/distill/mmcls/wsld/wsld_logits_resnet34_resnet18_8xb32_in1k.py

@@ -19,17 +19,20 @@ model = dict(
     teacher=dict(
         cfg_path='mmcls::resnet/resnet34_8xb32_in1k.py', pretrained=True),
     teacher_ckpt='resnet34_8xb32_in1k_20210831-f257d4e6.pth',
-    student_recorders=dict(
-        fc=dict(type='ModuleOutputs', source='head.fc'),
-        data_samples=dict(type='ModuleInputs', source='')),
-    teacher_recorders=dict(fc=dict(type='ModuleOutputs', source='head.fc')),
-    distill_losses=dict(loss_wsld=dict(type='WSLD', tau=2, loss_weight=2.5)),
-    loss_forward_mappings=dict(
-        loss_wsld=dict(
-            student=dict(recorder='fc', from_student=True),
-            teacher=dict(recorder='fc', from_student=False),
-            data_samples=dict(
-                recorder='data_samples', from_student=True, data_idx=1))))
+    distiller=dict(
+        student_recorders=dict(
+            fc=dict(type='ModuleOutputs', source='head.fc'),
+            data_samples=dict(type='ModuleInputs', source='')),
+        teacher_recorders=dict(
+            fc=dict(type='ModuleOutputs', source='head.fc')),
+        distill_losses=dict(
+            loss_wsld=dict(type='WSLD', tau=2, loss_weight=2.5)),
+        loss_forward_mappings=dict(
+            loss_wsld=dict(
+                student=dict(recorder='fc', from_student=True),
+                teacher=dict(recorder='fc', from_student=False),
+                data_samples=dict(
+                    recorder='data_samples', from_student=True, data_idx=1)))))
 
 find_unused_parameters = True
 

configs/distill/mmdet/cwd/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco.py

@@ -1,4 +1,4 @@
-_base_ = ['./cwd_fpn_gfl_r101_gfl_r50_1x_coco.py']
+_base_ = ['./cwd_fpn_retina_r101_retina_r50_1x_coco.py']
 
 model = dict(
     architecture=dict(

configs/distill/mmdet/cwd/cwd_fpn_frcnn_r101_frcnn_r50_1x_coco.py

@@ -16,42 +16,38 @@ model = dict(
         cfg_path='mmdet::faster_rcnn/faster_rcnn_r101_fpn_2x_coco.py',
         pretrained=False),
     teacher_ckpt=teacher_ckpt,
-    distill_losses=dict(
-        loss_cwd_fpn0=dict(
-            type='ChannelWiseDivergence', tau=1, loss_weight=10),
-        loss_cwd_fpn1=dict(
-            type='ChannelWiseDivergence', tau=1, loss_weight=10),
-        loss_cwd_fpn2=dict(
-            type='ChannelWiseDivergence', tau=1, loss_weight=10),
-        loss_cwd_fpn3=dict(
-            type='ChannelWiseDivergence', tau=1, loss_weight=10),
-        loss_cwd_fpn4=dict(
-            type='ChannelWiseDivergence', tau=1, loss_weight=10)),
-    student_recorders=dict(fpn=dict(type='ModuleOutputs', source='neck')),
-    teacher_recorders=dict(fpn=dict(type='ModuleOutputs', source='neck')),
-    loss_forward_mappings=dict(
-        loss_cwd_fpn0=dict(
-            preds_S=dict(from_student=True, recorder='fpn', data_idx=0),
-            preds_T=dict(from_student=False, recorder='fpn', data_idx=0),
-        ),
-        loss_cwd_fpn1=dict(
-            preds_S=dict(from_student=True, recorder='fpn', data_idx=1),
-            preds_T=dict(from_student=False, recorder='fpn', data_idx=1),
-        ),
-        loss_cwd_fpn2=dict(
-            preds_S=dict(from_student=True, recorder='fpn', data_idx=2),
-            preds_T=dict(from_student=False, recorder='fpn', data_idx=2),
-        ),
-        loss_cwd_fpn3=dict(
-            preds_S=dict(from_student=True, recorder='fpn', data_idx=3),
-            preds_T=dict(from_student=False, recorder='fpn', data_idx=3),
-        ),
-        loss_cwd_fpn4=dict(
-            preds_S=dict(from_student=True, recorder='fpn', data_idx=4),
-            preds_T=dict(from_student=False, recorder='fpn', data_idx=4),
-        ),
-    ),
-)
+    distiller=dict(
+        type='ConfigurableDistiller',
+        student_recorders=dict(fpn=dict(type='ModuleOutputs', source='neck')),
+        teacher_recorders=dict(fpn=dict(type='ModuleOutputs', source='neck')),
+        distill_losses=dict(
+            loss_cwd_fpn0=dict(
+                type='ChannelWiseDivergence', tau=1, loss_weight=10),
+            loss_cwd_fpn1=dict(
+                type='ChannelWiseDivergence', tau=1, loss_weight=10),
+            loss_cwd_fpn2=dict(
+                type='ChannelWiseDivergence', tau=1, loss_weight=10),
+            loss_cwd_fpn3=dict(
+                type='ChannelWiseDivergence', tau=1, loss_weight=10),
+            loss_cwd_fpn4=dict(
+                type='ChannelWiseDivergence', tau=1, loss_weight=10)),
+        loss_forward_mappings=dict(
+            loss_cwd_fpn0=dict(
+                preds_S=dict(from_student=True, recorder='fpn', data_idx=0),
+                preds_T=dict(from_student=False, recorder='fpn', data_idx=0)),
+            loss_cwd_fpn1=dict(
+                preds_S=dict(from_student=True, recorder='fpn', data_idx=1),
+                preds_T=dict(from_student=False, recorder='fpn', data_idx=1)),
+            loss_cwd_fpn2=dict(
+                preds_S=dict(from_student=True, recorder='fpn', data_idx=2),
+                preds_T=dict(from_student=False, recorder='fpn', data_idx=2)),
+            loss_cwd_fpn3=dict(
+                preds_S=dict(from_student=True, recorder='fpn', data_idx=3),
+                preds_T=dict(from_student=False, recorder='fpn', data_idx=3)),
+            loss_cwd_fpn4=dict(
+                preds_S=dict(from_student=True, recorder='fpn', data_idx=4),
+                preds_T=dict(from_student=False, recorder='fpn',
+                             data_idx=4)))))
 
 find_unused_parameters = True
 

configs/distill/mmdet/cwd/cwd_fpn_retina_r101_retina_r50_1x_coco.py

@@ -7,3 +7,7 @@ model = dict(
     teacher=dict(
         cfg_path='mmdet::retinanet/retinanet_r101_fpn_2x_coco.py',
         pretrained=True))
+
+# optimizer
+optim_wrapper = dict(
+    optimizer=dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0001))

configs/distill/mmdet/cwd/metafile.yml

@@ -0,0 +1,23 @@
+
+Models:
+  - Name: cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco
+    In Collection: CWD
+    Metadata:
+      Location: cls head
+      Student:
+        Metrics:
+          box AP: 40.2
+        Config: mmdet::gfl/gfl_r50_fpn_1x_coco.py
+        Weights: https://download.openmmlab.com/mmdetection/v2.0/gfl/gfl_r50_fpn_1x_coco/gfl_r50_fpn_1x_coco_20200629_121244-25944287.pth
+      Teacher:
+        Metrics:
+          box AP: 44.7
+        Config: mmdet::gfl/gfl_r50_fpn_mstrain_2x_coco.py
+        Weights: https://download.openmmlab.com/mmdetection/v2.0/gfl/gfl_r101_fpn_mstrain_2x_coco/gfl_r101_fpn_mstrain_2x_coco_20200629_200126-dd12f847.pth
+    Results:
+      - Task: Object Detection
+        Dataset: COCO
+        Metrics:
+          box AP: 41.9
+    Config: configs/distill/mmdet/cwd/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco.py
+    Weights: https://download.openmmlab.com/mmrazor/v0.1/distill/cwd/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco_20211222-655dff39.pth

configs/distill/mmseg/cwd/README.md

@@ -0,0 +1,37 @@
+# CWD
+
+> [Channel-wise Knowledge Distillation for Dense Prediction](https://arxiv.org/abs/2011.13256)
+
+<!-- [ALGORITHM] -->
+
+## Abstract
+
+Knowledge distillation (KD) has been proven to be a simple and effective tool for training compact models. Almost all KD variants for dense prediction tasks align the student and teacher networks' feature maps in the spatial domain, typically by minimizing point-wise and/or pair-wise discrepancy. Observing that in semantic segmentation, some layers' feature activations of each channel tend to encode saliency of scene categories (analogue to class activation mapping), we propose to align features channel-wise between the student and teacher networks. To this end, we first transform the feature map of each channel into a probability map using softmax normalization, and then minimize the Kullback-Leibler (KL) divergence of the corresponding channels of the two networks. By doing so, our method focuses on mimicking the soft distributions of channels between networks. In particular, the KL divergence enables learning to pay more attention to the most salient regions of the channel-wise maps, presumably corresponding to the most useful signals for semantic segmentation. Experiments demonstrate that our channel-wise distillation outperforms almost all existing spatial distillation methods for semantic segmentation considerably, and requires less computational cost during training. We consistently achieve superior performance on three benchmarks with various network structures.
+
+![pipeline](/docs/en/imgs/model_zoo/cwd/pipeline.png)
+
+## Results and models
+
+### Segmentation
+
+| Location |  Dataset   |                                                             Teacher                                                              |                                                            Student                                                             | mIoU  | mIoU(T) | mIou(S) |    Config    | Download                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                |
+| :------: | :--------: | :------------------------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------------------------------: | :---: | :-----: | :-----: | :----------: | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+|  logits  | cityscapes | [pspnet_r101](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/pspnet/pspnet_r101-d8_512x1024_80k_cityscapes.py) | [pspnet_r18](https://github.com/open-mmlab/mmsegmentation/blob/master/configs/pspnet/pspnet_r18-d8_512x1024_80k_cityscapes.py) | 75.54 |  79.76  |  74.87  | [config](<>) | [teacher](https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_512x1024_80k_cityscapes/pspnet_r101-d8_512x1024_80k_cityscapes_20200606_112211-e1e1100f.pth) \|[model](https://download.openmmlab.com/mmrazor/v0.1/distill/cwd/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k_mIoU-75.54_20211222-3a26ee1c.pth) \| [log](https://download.openmmlab.com/mmrazor/v0.1/distill/cwd/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k_20211212_205711.log.json?) |
+
+### Detection
+
+| Location | Dataset |                                                     Teacher                                                      |                                                Student                                                 | mAP  | mAP(T) | mAP(S) |    Config    | Download                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                          |
+| :------: | :-----: | :--------------------------------------------------------------------------------------------------------------: | :----------------------------------------------------------------------------------------------------: | :--: | :----: | :----: | :----------: | :-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| cls head |  COCO   | [gfl_r101_2x](https://github.com/open-mmlab/mmdetection/tree/master/configs/gfl/gfl_r101_fpn_mstrain_2x_coco.py) | [gfl_r50_1x](https://github.com/open-mmlab/mmdetection/tree/master/configs/gfl/gfl_r50_fpn_1x_coco.py) | 41.9 |  44.7  |  40.2  | [config](<>) | [teacher](https://download.openmmlab.com/mmdetection/v2.0/gfl/gfl_r101_fpn_mstrain_2x_coco/gfl_r101_fpn_mstrain_2x_coco_20200629_200126-dd12f847.pth) \|[model](https://download.openmmlab.com/mmrazor/v0.1/distill/cwd/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco_20211222-655dff39.pth) \| [log](https://download.openmmlab.com/mmrazor/v0.1/distill/cwd/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco/cwd_cls_head_gfl_r101_fpn_gfl_r50_fpn_1x_coco_20211212_205444.log.json) |
+
+## Citation
+
+```latex
+@inproceedings{shu2021channel,
+  title={Channel-Wise Knowledge Distillation for Dense Prediction},
+  author={Shu, Changyong and Liu, Yifan and Gao, Jianfei and Yan, Zheng and Shen, Chunhua},
+  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
+  pages={5311--5320},
+  year={2021}
+}
+```

configs/distill/mmseg/cwd/cwd_logits_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k.py

@@ -0,0 +1,33 @@
+_base_ = [
+    'mmseg::_base_/datasets/cityscapes.py',
+    'mmseg::_base_/schedules/schedule_80k.py',
+    'mmseg::_base_/default_runtime.py'
+]
+
+teacher_ckpt = 'https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_512x1024_80k_cityscapes/pspnet_r101-d8_512x1024_80k_cityscapes_20200606_112211-e1e1100f.pth'  # noqa: E501
+model = dict(
+    _scope_='mmrazor',
+    type='SingleTeacherDistill',
+    architecture=dict(
+        cfg_path='mmseg::pspnet/pspnet_r18-d8_512x1024_80k_cityscapes.py',
+        pretrained=False),
+    teacher=dict(
+        cfg_path='mmseg::pspnet/pspnet_r101-d8_512x1024_80k_cityscapes.py',
+        pretrained=False),
+    teacher_ckpt=teacher_ckpt,
+    distiller=dict(
+        type='ConfigurableDistiller',
+        distill_losses=dict(
+            loss_cwd=dict(type='ChannelWiseDivergence', tau=1, loss_weight=5)),
+        student_recorders=dict(
+            logits=dict(type='ModuleOutputs', source='decode_head.conv_seg')),
+        teacher_recorders=dict(
+            logits=dict(type='ModuleOutputs', source='decode_head.conv_seg')),
+        loss_forward_mappings=dict(
+            loss_cwd=dict(
+                preds_S=dict(from_student=True, recorder='logits'),
+                preds_T=dict(from_student=False, recorder='logits')))))
+
+find_unused_parameters = True
+
+val_cfg = dict(_delete_=True, type='mmrazor.SingleTeacherDistillValLoop')

configs/distill/mmseg/cwd/metafile.yml

@@ -0,0 +1,41 @@
+Collections:
+  - Name: CWD
+    Metadata:
+      Training Data:
+        - Cityscapes
+        - COCO
+    Paper:
+      URL: https://arxiv.org/abs/2011.13256
+      Title: Channel-wise Knowledge Distillation for Dense Prediction
+    README: configs/distill/mmseg/cwd/README.md
+    Code:
+      URL: https://github.com/open-mmlab/mmrazor/blob/v0.1.0/mmrazor/models/losses/cwd.py#L10
+      Version: v0.1.0
+    Converted From:
+      Code:
+        - https://github.com/pppppM/mmsegmentation-distiller
+        - https://github.com/pppppM/mmdetection-distiller
+Models:
+  - Name: cwd_logits_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k
+    In Collection: CWD
+    Metadata:
+      Location: logits
+      Student:
+        Metrics:
+          mIoU: 74.87
+          mIoU(ms+flip): 76.04
+        Config: mmseg::pspnet/pspnet_r18-d8_512x1024_80k_cityscapes.py
+        Weights: https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r18-d8_512x1024_80k_cityscapes/pspnet_r18-d8_512x1024_80k_cityscapes_20201225_021458-09ffa746.pth
+      Teacher:
+        Metrics:
+          mIoU: 79.76
+          mIoU(ms+flip): 81.01
+        Config: mmseg::pspnet/pspnet_r101-d8_512x1024_80k_cityscapes.py
+        Weights: https://download.openmmlab.com/mmsegmentation/v0.5/pspnet/pspnet_r101-d8_512x1024_80k_cityscapes/pspnet_r101-d8_512x1024_80k_cityscapes_20200606_112211-e1e1100f.pth
+    Results:
+      - Task: Semantic Segmentation
+        Dataset: Cityscapes
+        Metrics:
+          mIoU: 75.54
+    Config: configs/distill/mmseg/cwd/cwd_logits_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k.py
+    Weights: https://download.openmmlab.com/mmrazor/v0.1/distill/cwd/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k/cwd_cls_head_pspnet_r101_d8_pspnet_r18_d8_512x1024_cityscapes_80k_mIoU-75.54_20211222-3a26ee1c.pth

configs/distill/rkd/rkd_neck_resnet34_resnet18_8xb32_in1k.py

@@ -1,79 +0,0 @@
-_base_ = [
-    '../../_base_/datasets/mmcls/imagenet_bs32.py',
-    '../../_base_/schedules/mmcls/imagenet_bs256.py',
-    '../../_base_/mmcls_runtime.py'
-]
-
-# model settings
-student = dict(
-    type='mmcls.ImageClassifier',
-    backbone=dict(
-        type='ResNet',
-        depth=18,
-        num_stages=4,
-        out_indices=(3, ),
-        style='pytorch'),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=512,
-        loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
-        topk=(1, 5),
-    ))
-
-# teacher settings
-teacher_ckpt = 'https://download.openmmlab.com/mmclassification/v0/resnet/resnet34_8xb32_in1k_20210831-f257d4e6.pth'  # noqa: E501
-
-teacher = dict(
-    type='mmcls.ImageClassifier',
-    init_cfg=dict(type='Pretrained', checkpoint=teacher_ckpt),
-    backbone=dict(
-        type='ResNet',
-        depth=34,
-        num_stages=4,
-        out_indices=(3, ),
-        style='pytorch'),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=512,
-        loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
-        topk=(1, 5),
-    ))
-
-# algorithm setting
-algorithm = dict(
-    type='GeneralDistill',
-    architecture=dict(
-        type='MMClsArchitecture',
-        model=student,
-    ),
-    with_student_loss=True,
-    with_teacher_loss=False,
-    distiller=dict(
-        type='SingleTeacherDistiller',
-        teacher=teacher,
-        teacher_trainable=False,
-        teacher_norm_eval=True,
-        components=[
-            dict(
-                student_module='neck.gap',
-                teacher_module='neck.gap',
-                losses=[
-                    dict(
-                        type='DistanceWiseRKD',
-                        name='distance_wise_loss',
-                        loss_weight=25.0,
-                        with_l2_norm=True),
-                    dict(
-                        type='AngleWiseRKD',
-                        name='angle_wise_loss',
-                        loss_weight=50.0,
-                        with_l2_norm=True),
-                ])
-        ]),
-)
-
-find_unused_parameters = True

configs/distill/wsld/wsld_cls_head_resnet34_resnet18_8xb32_in1k.py

@@ -1,78 +0,0 @@
-_base_ = [
-    '../../_base_/datasets/mmcls/imagenet_bs32.py',
-    '../../_base_/schedules/mmcls/imagenet_bs256.py',
-    '../../_base_/mmcls_runtime.py'
-]
-
-# model settings
-student = dict(
-    type='mmcls.ImageClassifier',
-    backbone=dict(
-        type='ResNet',
-        depth=18,
-        num_stages=4,
-        out_indices=(3, ),
-        style='pytorch'),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=512,
-        loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
-        topk=(1, 5),
-    ))
-
-checkpoint = 'https://download.openmmlab.com/mmclassification/v0/resnet/resnet50_8xb32_in1k_20210831-ea4938fc.pth'  # noqa: E501
-
-# teacher settings
-teacher = dict(
-    type='mmcls.ImageClassifier',
-    init_cfg=dict(type='Pretrained', checkpoint=checkpoint),
-    backbone=dict(
-        type='ResNet',
-        depth=34,
-        num_stages=4,
-        out_indices=(3, ),
-        style='pytorch'),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=512,
-        loss=dict(type='CrossEntropyLoss', loss_weight=1.0),
-        topk=(1, 5),
-    ))
-
-# algorithm setting
-algorithm = dict(
-    type='GeneralDistill',
-    architecture=dict(
-        type='MMClsArchitecture',
-        model=student,
-    ),
-    with_student_loss=True,
-    # teacher_trainable and with_teacher_loss have a dependency
-    # relationship, if teacher_trainable is false, then
-    # with_teacher_loss must be false.
-    with_teacher_loss=False,
-    distiller=dict(
-        type='SingleTeacherDistiller',
-        teacher=teacher,
-        teacher_trainable=False,
-        teacher_norm_eval=True,
-        components=[
-            dict(
-                student_module='head.fc',
-                teacher_module='head.fc',
-                losses=[
-                    dict(
-                        type='WSLD',
-                        name='loss_wsld',
-                        tau=2,
-                        loss_weight=2.5,
-                        num_classes=1000)
-                ])
-        ]),
-)
-
-find_unused_parameters = True

configs/nas/cream/CREAM_14_MOBILENET_IN1k_2.0.yaml

@@ -1,11 +0,0 @@
-modules:
-  backbone.layer1.0: depthsepconv
-  backbone.layer2.0: mb_k3e4_se
-  backbone.layer3.0: mb_k5e6_se
-  backbone.layer3.1: mb_k5e6_se
-  backbone.layer4.0: mb_k5e6_se
-  backbone.layer4.1: mb_k5e6_se
-  backbone.layer5.0: mb_k3e6_se
-  backbone.layer6.0: mb_k5e6_se
-  backbone.layer7.0: convbnact
-channels:

configs/nas/cream/cream_14_subnet_mobilenet.py

@@ -1,8 +0,0 @@
-_base_ = ['./cream_14_supernet_mobilenet.py']
-
-# FIXME: you may replace this with the mutable_cfg searched by yourself
-fix_subnet = 'configs/nas/cream/CREAM_14_MOBILENET_IN1k_2.0.yaml'  # noqa: E501
-
-model = dict(fix_subnet=fix_subnet)
-
-find_unused_parameters = False

configs/nas/cream/cream_14_supernet_mobilenet.py

@@ -1,241 +0,0 @@
-# dataset settings
-dataset_type = 'ImageNet'
-
-preprocess_cfg = dict(
-    # RGB format normalization parameters
-    mean=[123.675, 116.28, 103.53],
-    std=[58.395, 57.12, 57.375],
-    # convert image from BGR to RGB
-    to_rgb=True,
-)
-
-# file_client_args = dict(
-#     backend='petrel',
-#     path_mapping=dict({
-#         './data/imagenet': 's3://openmmlab/datasets/classification/imagenet',
-#         'data/imagenet': 's3://openmmlab/datasets/classification/imagenet'
-#     }))
-
-train_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(type='RandomResizedCrop', scale=224),
-    dict(type='ColorJitter', brightness=0.4, contrast=0.4, saturation=0.4),
-    dict(type='RandomFlip', prob=0.5, direction='horizontal'),
-    dict(type='PackClsInputs'),
-]
-
-test_pipeline = [
-    dict(type='LoadImageFromFile'),
-    dict(
-        type='ResizeEdge',
-        scale=73,
-        edge='short',
-        backend='pillow',
-        interpolation='bicubic'),
-    dict(type='CenterCrop', crop_size=64),
-    dict(type='PackClsInputs'),
-]
-
-train_dataloader = dict(
-    batch_size=128,
-    num_workers=5,
-    dataset=dict(
-        type=dataset_type,
-        data_root='/mnt/cache/share/images',
-        ann_file='meta/train.txt',
-        data_prefix='train',
-        pipeline=train_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=True),
-    persistent_workers=True,
-)
-
-# /mnt/lustre/share_data/wangjiaqi/data/imagenet',
-
-val_dataloader = dict(
-    batch_size=128,
-    num_workers=5,
-    dataset=dict(
-        type=dataset_type,
-        data_root='/mnt/cache/share/images',
-        ann_file='meta/val.txt',
-        data_prefix='val',
-        pipeline=test_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    persistent_workers=True,
-)
-val_evaluator = dict(type='Accuracy', topk=(1, 5))
-
-# If you want standard test, please manually configure the test dataset
-test_dataloader = val_dataloader
-test_evaluator = val_evaluator
-
-# scheduler
-
-# optimizer
-optim_wrapper = dict(
-    optimizer=dict(type='SGD', lr=0.5, momentum=0.9, weight_decay=4e-5),
-    clip_grad=None)
-
-# leanring policy
-param_scheduler = [
-    dict(type='PolyLR', power=1.0, eta_min=0.0, by_epoch=False),
-]
-
-# train, val, test setting
-train_cfg = dict(by_epoch=False, max_iters=300000)
-val_cfg = dict()
-test_cfg = dict()
-
-# runtime
-
-# defaults to use registries in mmrazor
-default_scope = 'mmcls'
-
-# configure default hooks
-default_hooks = dict(
-    timer=dict(type='IterTimerHook'),
-    logger=dict(type='LoggerHook', interval=100),
-    param_scheduler=dict(type='ParamSchedulerHook'),
-    checkpoint=dict(type='CheckpointHook', by_epoch=False, interval=1000),
-    sampler_seed=dict(type='DistSamplerSeedHook'),
-    visualization=dict(type='VisualizationHook', enable=False),
-)
-
-# configure environment
-env_cfg = dict(
-    cudnn_benchmark=False,
-    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
-    dist_cfg=dict(backend='nccl'),
-)
-
-# set visualizer
-vis_backends = [dict(type='LocalVisBackend')]
-visualizer = dict(
-    type='ClsVisualizer', vis_backends=vis_backends, name='visualizer')
-
-# set log level
-log_level = 'INFO'
-
-# load from which checkpoint
-load_from = None
-
-# whether to resume training from the loaded checkpoint
-resume = False
-
-se_cfg = dict(
-    ratio=4,
-    divisor=8,
-    act_cfg=(dict(type='ReLU'),
-             dict(
-                 type='HSigmoid', bias=3, divisor=6, min_value=0,
-                 max_value=1)))
-
-_FIRST_STAGE_MUTABLE = dict(  # DepthwiseSep
-    type='OneShotMutableOP',
-    candidates=dict(
-        depthsepconv=dict(
-            type='DepthwiseSeparableConv',
-            dw_kernel_size=3,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='Swish'))))
-
-_MIDDLE_STAGE_MUTABLE = dict(
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e4_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='Swish')),
-        mb_k3e6_se=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='Swish')),
-        mb_k5e4_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='Swish')),
-        mb_k5e6_se=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='Swish')),
-        mb_k7e4_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=4,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='Swish')),
-        mb_k7e6_se=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=6,
-            se_cfg=se_cfg,
-            norm_cfg=dict(type='BN'),
-            act_cfg=dict(type='Swish'))))
-
-arch_setting = [
-    # Parameters to build layers. 4 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, stride, mutable cfg.
-    [16, 1, 1, _FIRST_STAGE_MUTABLE],
-    [24, 1, 2, _MIDDLE_STAGE_MUTABLE],
-    [40, 2, 2, _MIDDLE_STAGE_MUTABLE],
-    [80, 2, 2, _MIDDLE_STAGE_MUTABLE],
-    [96, 1, 1, _MIDDLE_STAGE_MUTABLE],
-    [192, 1, 2, _MIDDLE_STAGE_MUTABLE],
-]
-
-norm_cfg = dict(type='BN')
-supernet = dict(
-    _scope_='mmcls',
-    type='ImageClassifier',
-    data_preprocessor=preprocess_cfg,
-    backbone=dict(
-        _scope_='mmrazor',
-        type='SearchableMobileNet',
-        arch_setting=arch_setting,
-        first_channels=16,
-        last_channels=320,
-        widen_factor=1.0,
-        norm_cfg=norm_cfg,
-        act_cfg=dict(type='Swish'),
-        out_indices=(6, ),
-    ),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='mmrazor.CreamClsHead',
-        num_classes=1000,
-        in_channels=320,
-        num_features=1280,
-        act_cfg=dict(type='Swish'),
-        loss=dict(
-            type='LabelSmoothLoss',
-            num_classes=1000,
-            label_smooth_val=0.1,
-            mode='original',
-            loss_weight=1.0),
-        topk=(1, 5),
-    ),
-)
-
-mutator = dict(type='mmrazor.OneShotModuleMutator')
-
-model = dict(
-    type='mmrazor.SPOS',
-    architecture=supernet,
-    mutator=mutator,
-)
-
-find_unused_parameters = True

configs/nas/detnas/DETNAS_FRCNN_SHUFFLENETV2_340M_COCO_MMRAZOR_2.0.yaml

@@ -1,22 +0,0 @@
-modules:
-  backbone.layers.0.0: shuffle_5x5
-  backbone.layers.0.1: shuffle_3x3
-  backbone.layers.0.2: shuffle_3x3
-  backbone.layers.0.3: shuffle_3x3
-  backbone.layers.1.0: shuffle_xception
-  backbone.layers.1.1: shuffle_3x3
-  backbone.layers.1.2: shuffle_xception
-  backbone.layers.1.3: shuffle_7x7
-  backbone.layers.2.0: shuffle_7x7
-  backbone.layers.2.1: shuffle_7x7
-  backbone.layers.2.2: shuffle_xception
-  backbone.layers.2.3: shuffle_xception
-  backbone.layers.2.4: shuffle_3x3
-  backbone.layers.2.5: shuffle_7x7
-  backbone.layers.2.6: shuffle_5x5
-  backbone.layers.2.7: shuffle_xception
-  backbone.layers.3.0: shuffle_7x7
-  backbone.layers.3.1: shuffle_7x7
-  backbone.layers.3.2: shuffle_7x7
-  backbone.layers.3.3: shuffle_5x5
-channels:

configs/nas/detnas/DetNAS_SPOS_SHUFFLENETV2_330M_IN1k_PAPER_2.0.yaml

@@ -1,22 +0,0 @@
-modules:
-  backbone.layers.0.0: shuffle_5x5
-  backbone.layers.0.1: shuffle_3x3
-  backbone.layers.0.2: shuffle_3x3
-  backbone.layers.0.3: shuffle_3x3
-  backbone.layers.1.0: shuffle_xception
-  backbone.layers.1.1: shuffle_3x3
-  backbone.layers.1.2: shuffle_xception
-  backbone.layers.1.3: shuffle_7x7
-  backbone.layers.2.0: shuffle_7x7
-  backbone.layers.2.1: shuffle_7x7
-  backbone.layers.2.2: shuffle_xception
-  backbone.layers.2.3: shuffle_xception
-  backbone.layers.2.4: shuffle_3x3
-  backbone.layers.2.5: shuffle_7x7
-  backbone.layers.2.6: shuffle_5x5
-  backbone.layers.2.7: shuffle_xception
-  backbone.layers.3.0: shuffle_7x7
-  backbone.layers.3.1: shuffle_7x7
-  backbone.layers.3.2: shuffle_7x7
-  backbone.layers.3.3: shuffle_5x5
-channels:

configs/nas/detnas/detnas_subnet_shufflenetv2_8xb128_in1k_2.0_frcnn.py

@@ -1,8 +0,0 @@
-_base_ = ['./detnas_supernet_shufflenetv2_coco_1x_2.0_frcnn.py']
-
-# FIXME: you may replace this with the mutable_cfg searched by yourself
-fix_subnet = 'configs/nas/detnas/DETNAS_FRCNN_SHUFFLENETV2_340M_COCO_MMRAZOR_2.0.yaml'  # noqa: E501
-
-model = dict(fix_subnet=fix_subnet)
-
-find_unused_parameters = False

configs/nas/detnas/detnas_supernet_shufflenetv2_coco_1x_2.0_frcnn.py

@@ -1,87 +0,0 @@
-_base_ = [
-    'mmdet::_base_/models/faster_rcnn_r50_fpn.py',
-    'mmdet::_base_/datasets/coco_detection.py',
-    'mmdet::_base_/schedules/schedule_1x.py',
-    'mmdet::_base_/default_runtime.py'
-]
-
-data_root = '/mnt/lustre/share_data/zhangwenwei/data/coco/'
-
-_base_.train_dataloader.dataset.data_root = data_root
-
-visualizer = None
-
-log_level = 'INFO'
-load_from = '/mnt/lustre/dongpeijie/detnas_subnet_shufflenetv2_8xb128_in1k_acc-74.08_20211223-92e9b66a_2.0.pth'  # noqa: E501
-resume = False
-
-norm_cfg = dict(type='SyncBN', requires_grad=True)
-# model settings
-_STAGE_MUTABLE = dict(
-    _scope_='mmrazor',
-    type='mmrazor.OneShotMutableOP',
-    candidates=dict(
-        shuffle_3x3=dict(
-            type='mmrazor.ShuffleBlock', kernel_size=3, norm_cfg=norm_cfg),
-        shuffle_5x5=dict(
-            type='mmrazor.ShuffleBlock', kernel_size=5, norm_cfg=norm_cfg),
-        shuffle_7x7=dict(
-            type='mmrazor.ShuffleBlock', kernel_size=7, norm_cfg=norm_cfg),
-        shuffle_xception=dict(
-            type='mmrazor.ShuffleXception', norm_cfg=norm_cfg),
-    ))
-
-arch_setting = [
-    # Parameters to build layers. 3 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, mutable_cfg.
-    [64, 4, _STAGE_MUTABLE],
-    [160, 4, _STAGE_MUTABLE],
-    [320, 8, _STAGE_MUTABLE],
-    [640, 4, _STAGE_MUTABLE],
-]
-
-supernet = _base_.model
-
-supernet.backbone = dict(
-    type='mmrazor.SearchableShuffleNetV2',
-    arch_setting=arch_setting,
-    norm_cfg=norm_cfg,
-    out_indices=(0, 1, 2, 3),
-    widen_factor=1.0,
-    with_last_layer=False)
-
-supernet.neck = dict(
-    type='FPN',
-    norm_cfg=norm_cfg,
-    in_channels=[64, 160, 320, 640],
-    out_channels=256,
-    num_outs=5)
-
-supernet.roi_head.bbox_head = dict(
-    type='Shared4Conv1FCBBoxHead',
-    norm_cfg=norm_cfg,
-    in_channels=256,
-    fc_out_channels=1024,
-    roi_feat_size=7,
-    num_classes=80,
-    bbox_coder=dict(
-        type='DeltaXYWHBBoxCoder',
-        target_means=[0., 0., 0., 0.],
-        target_stds=[0.1, 0.1, 0.2, 0.2]),
-    reg_class_agnostic=False,
-    loss_cls=dict(type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0),
-    loss_bbox=dict(type='L1Loss', loss_weight=1.0))
-
-mutator = dict(type='mmrazor.OneShotModuleMutator')
-
-fix_subnet = 'configs/nas/detnas/DETNAS_FRCNN_SHUFFLENETV2_340M_COCO_MMRAZOR_2.0.yaml'  # noqa: E501
-
-model = dict(
-    _delete_=True,
-    type='mmrazor.SPOS',
-    architecture=supernet,
-    mutator=mutator,
-    fix_subnet=fix_subnet,
-)
-
-find_unused_parameters = True

configs/nas/detnas/detnas_supernet_shufflenetv2_coco_1x_2.0_retinanet.py

@@ -1,114 +0,0 @@
-_base_ = [
-    'mmdet::faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py',
-    'mmdet::datasets/coco_detection.py', 'mmdet::schedules/schedule_1x.py',
-    'mmdet::default_runtime.py'
-]
-
-data_root = '/mnt/lustre/share_data/zhangwenwei/data/coco/'
-
-train_dataloader = dict(dataset=dict(data_root=data_root, ))
-
-visualizer = None
-# custom_hooks = [dict(type='DetVisualizationHook', interval=10)]
-
-log_level = 'INFO'
-load_from = None
-resume = False
-
-# TODO: support auto scaling lr
-
-norm_cfg = dict(type='SyncBN', requires_grad=True)
-# model settings
-_STAGE_MUTABLE = dict(
-    _scope_='mmrazor',
-    type='mmrazor.OneShotMutableOP',
-    candidates=dict(
-        shuffle_3x3=dict(
-            type='mmrazor.ShuffleBlock', kernel_size=3, norm_cfg=norm_cfg),
-        shuffle_5x5=dict(
-            type='mmrazor.ShuffleBlock', kernel_size=5, norm_cfg=norm_cfg),
-        shuffle_7x7=dict(
-            type='mmrazor.ShuffleBlock', kernel_size=7, norm_cfg=norm_cfg),
-        shuffle_xception=dict(
-            type='mmrazor.ShuffleXception', norm_cfg=norm_cfg),
-    ))
-
-arch_setting = [
-    # Parameters to build layers. 3 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, mutable_cfg.
-    [64, 4, _STAGE_MUTABLE],
-    [160, 4, _STAGE_MUTABLE],
-    [320, 8, _STAGE_MUTABLE],
-    [640, 4, _STAGE_MUTABLE],
-]
-
-supernet = dict(
-    type='RetinaNet',
-    data_preprocessor=dict(
-        type='DetDataPreprocessor',
-        mean=[123.675, 116.28, 103.53],
-        std=[58.395, 57.12, 57.375],
-        bgr_to_rgb=True,
-        pad_size_divisor=32),
-    backbone=dict(
-        type='mmrazor.SearchableShuffleNetV2',
-        arch_setting=arch_setting,
-        norm_cfg=norm_cfg,
-        out_indices=(0, 1, 2, 3),
-        widen_factor=1.0,
-        with_last_layer=False),
-    neck=dict(
-        type='FPN',
-        in_channels=[64, 160, 320, 640],
-        out_channels=256,
-        num_outs=5),
-    bbox_head=dict(
-        type='RetinaHead',
-        num_classes=80,
-        in_channels=256,
-        stacked_convs=4,
-        feat_channels=256,
-        anchor_generator=dict(
-            type='AnchorGenerator',
-            octave_base_scale=4,
-            scales_per_octave=3,
-            ratios=[0.5, 1.0, 2.0],
-            strides=[8, 16, 32, 64, 128]),
-        bbox_coder=dict(
-            type='DeltaXYWHBBoxCoder',
-            target_means=[.0, .0, .0, .0],
-            target_stds=[1.0, 1.0, 1.0, 1.0]),
-        loss_cls=dict(
-            type='FocalLoss',
-            use_sigmoid=True,
-            gamma=2.0,
-            alpha=0.25,
-            loss_weight=1.0),
-        loss_bbox=dict(type='L1Loss', loss_weight=1.0)),
-    # model training and testing settings
-    train_cfg=dict(
-        assigner=dict(
-            type='MaxIoUAssigner',
-            pos_iou_thr=0.5,
-            neg_iou_thr=0.4,
-            min_pos_iou=0,
-            ignore_iof_thr=-1),
-        allowed_border=-1,
-        pos_weight=-1,
-        debug=False),
-    test_cfg=dict(
-        nms_pre=1000,
-        min_bbox_size=0,
-        score_thr=0.05,
-        nms=dict(type='nms', iou_threshold=0.5),
-        max_per_img=100))
-
-mutator = dict(type='mmrazor.OneShotModuleMutator')
-
-model = dict(
-    type='mmrazor.SPOS',
-    architecture=supernet,
-    mutator=mutator,
-)
-
-find_unused_parameters = True

configs/nas/mmcls/spos/metafile.yml

@@ -6,24 +6,23 @@ Collections:
     Paper:
       URL: https://arxiv.org/abs/1904.00420
       Title: Single Path One-Shot Neural Architecture Search with Uniform Sampling
-    README: configs/nas/spos/README.md
+    README: configs/nas/mmcls/spos/README.md
     Code:
       URL: https://github.com/open-mmlab/mmrazor/blob/v0.1.0/mmrazor/models/algorithms/spos.py
       Version: v0.1.0
     Converted From:
       Code: https://github.com/megvii-model/SinglePathOneShot
 Models:
-  - Name: spos_subnet_shufflenetv2_8xb128_in1k
+  - Name: spos_shufflenet_subnet_8xb128_in1k
     In Collection: SPOS
     Metadata:
       FLOPs: 330 MB
-      Supernet: ShuffleNetV2
-      Mutable: https://download.openmmlab.com/mmrazor/v0.1/nas/spos/spos_shufflenetv2_subnet_8xb128_in1k/spos_shufflenetv2_subnet_8xb128_in1k_flops_0.33M_acc_73.87_20211222-454627be_mutable_cfg.yaml
+      Subnet: https://download.openmmlab.com/mmrazor/v0.1/nas/spos/spos_shufflenetv2_subnet_8xb128_in1k/spos_shufflenetv2_subnet_8xb128_in1k_flops_0.33M_acc_73.87_20211222-454627be_mutable_cfg.yaml
     Results:
       - Task: Image Classification
         Dataset: ImageNet-1k
         Metrics:
           Top 1 Accuracy: 73.87
           Top 5 Accuracy: 91.60
-    Config: configs/nas/spos/spos_subnet_shufflenetv2_8xb128_in1k.py
+    Config: configs/nas/mmcls/spos/spos_shufflenet_subnet_8xb128_in1k.py
     Weights: https://download.openmmlab.com/mmrazor/v0.1/nas/spos/spos_shufflenetv2_subnet_8xb128_in1k/spos_shufflenetv2_subnet_8xb128_in1k_flops_0.33M_acc_73.87_20211222-1f0a0b4d.pth

configs/nas/mmcls/spos/spos_mobilenet_supernet_8xb128_in1k.py

@@ -0,0 +1,30 @@
+_base_ = [
+    'mmrazor::_base_/settings/imagenet_bs1024_spos.py',
+    'mmrazor::_base_/nas_backbones/spos_mobilenet_supernet.py',
+    'mmcls::_base_/default_runtime.py',
+]
+
+# model
+supernet = dict(
+    type='ImageClassifier',
+    # data_preprocessor=_base_.preprocess_cfg,
+    backbone=_base_.nas_backbone,
+    neck=dict(type='GlobalAveragePooling'),
+    head=dict(
+        type='LinearClsHead',
+        num_classes=1000,
+        in_channels=1024,
+        loss=dict(
+            type='LabelSmoothLoss',
+            num_classes=1000,
+            label_smooth_val=0.1,
+            mode='original',
+            loss_weight=1.0),
+        topk=(1, 5)))
+
+model = dict(
+    type='mmrazor.SPOS',
+    architecture=supernet,
+    mutator=dict(type='mmrazor.OneShotModuleMutator'))
+
+find_unused_parameters = True

configs/nas/mmcls/spos/spos_shufflenet_supernet_8xb128_in1k.py

@@ -0,0 +1,30 @@
+_base_ = [
+    'mmrazor::_base_/settings/imagenet_bs1024_spos.py',
+    'mmrazor::_base_/nas_backbones/spos_shufflenet_supernet.py',
+    'mmcls::_base_/default_runtime.py',
+]
+
+# model
+supernet = dict(
+    type='ImageClassifier',
+    # data_preprocessor=_base_.preprocess_cfg,
+    backbone=_base_.nas_backbone,
+    neck=dict(type='GlobalAveragePooling'),
+    head=dict(
+        type='LinearClsHead',
+        num_classes=1000,
+        in_channels=1024,
+        loss=dict(
+            type='LabelSmoothLoss',
+            num_classes=1000,
+            label_smooth_val=0.1,
+            mode='original',
+            loss_weight=1.0),
+        topk=(1, 5)))
+
+model = dict(
+    type='mmrazor.SPOS',
+    architecture=supernet,
+    mutator=dict(type='mmrazor.OneShotModuleMutator'))
+
+find_unused_parameters = True

configs/nas/mmdet/detnas/detnas_frcnn_shufflenet_subnet_coco_1x.py

@@ -0,0 +1,9 @@
+_base_ = ['./spos_supernet_frcnn_shufflenet_coco_1x.py']
+
+# FIXME: you may replace this with the mutable_cfg searched by yourself
+# fix_subnet = 'configs/nas/spos/SPOS_SHUFFLENETV2_330M_IN1k_PAPER_2.0.yaml'  # noqa: E501
+fix_subnet = 'configs/nas/detnas/DetNAS_SPOS_SHUFFLENETV2_330M_IN1k_PAPER_2.0.yaml'  # noqa: E501
+
+model = dict(fix_subnet=fix_subnet)
+
+find_unused_parameters = False

configs/nas/mmdet/detnas/detnas_frcnn_shufflenet_supernet_coco_1x.py

@@ -0,0 +1,29 @@
+_base_ = [
+    'mmdet::_base_/models/faster_rcnn_r50_fpn.py',
+    'mmdet::_base_/datasets/coco_detection.py',
+    'mmdet::_base_/schedules/schedule_1x.py',
+    'mmdet::_base_/default_runtime.py',
+    'mmrazor::_base_/nas_backbones/spos_shufflenet_supernet.py'
+]
+
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+
+supernet = _base_.model
+
+supernet.backbone = _base_.nas_backbone
+supernet.backbone.norm_cfg = norm_cfg
+supernet.backbone.out_indices = (0, 1, 2, 3)
+supernet.backbone.with_last_layer = False
+
+supernet.neck.norm_cfg = norm_cfg
+supernet.neck.in_channels = [64, 160, 320, 640]
+
+supernet.roi_head.bbox_head.norm_cfg = norm_cfg
+
+model = dict(
+    _delete_=True,
+    type='mmrazor.SPOS',
+    architecture=supernet,
+    mutator=dict(type='mmrazor.OneShotModuleMutator'))
+
+find_unused_parameters = True

configs/nas/mmdet/detnas/detnas_retina_shufflenet_supernet_coco_1x.py

@@ -0,0 +1,27 @@
+_base_ = [
+    'mmdet::_base_/models/retinanet_r50_fpn.py',
+    'mmdet::_base_/datasets/coco_detection.py',
+    'mmdet::_base_/schedules/schedule_1x.py',
+    'mmdet::_base_/default_runtime.py',
+    'mmrazor::_base_/nas_backbones/spos_shufflenet_supernet.py'
+]
+
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+
+supernet = _base_.model
+
+supernet.backbone = _base_.nas_backbone
+supernet.backbone.norm_cfg = norm_cfg
+supernet.backbone.out_indices = (0, 1, 2, 3)
+supernet.backbone.with_last_layer = False
+
+supernet.neck.norm_cfg = norm_cfg
+supernet.neck.in_channels = [64, 160, 320, 640]
+
+model = dict(
+    _delete_=True,
+    type='mmrazor.SPOS',
+    architecture=supernet,
+    mutator=dict(type='mmrazor.OneShotModuleMutator'))
+
+find_unused_parameters = True

configs/nas/mmdet/detnas/detnas_shufflenet_subnet_8xb128_in1k.py

@@ -0,0 +1,9 @@
+_base_ = './detnas_shufflenet_supernet_8xb128_in1k.py'
+
+# FIXME: you may replace this with the mutable_cfg searched by yourself
+# fix_subnet = 'configs/nas/spos/SPOS_SHUFFLENETV2_330M_IN1k_PAPER_2.0.yaml'  # noqa: E501
+fix_subnet = 'configs/nas/detnas/DetNAS_SPOS_SHUFFLENETV2_330M_IN1k_PAPER_2.0.yaml'  # noqa: E501
+
+model = dict(fix_subnet=fix_subnet)
+
+find_unused_parameters = False

configs/nas/mmdet/detnas/detnas_shufflenet_supernet_8xb128_in1k.py

@@ -0,0 +1 @@
+_base_ = 'mmrazor::nas/mmcls/spos/shufflenet/spos_shufflenet_supernet_8xb128_in1k.py'  # noqa: E501

configs/nas/mmdet/detnas/metafile.yml

@@ -7,14 +7,14 @@ Collections:
     Paper:
       URL: https://arxiv.org/abs/1903.10979
       Title: DetNAS:Backbone Search for Object Detection
-    README: configs/nas/detnas/README.md
+    README: configs/nas/mmdet/detnas/README.md
     Code:
       URL: https://github.com/open-mmlab/mmrazor/blob/v0.1.0/mmrazor/models/algorithms/detnas.py
       Version: v0.1.0
     Converted From:
       Code: https://github.com/megvii-model/DetNAS
 Models:
-  - Name: detnas_subnet_frcnn_shufflenetv2_fpn_1x_coco
+  - Name: detnas_frcnn_shufflenet_subnet_coco_1x
     In Collection: DetNAS
     Metadata:
       FLOPs(Backbone): 340 MB
@@ -26,5 +26,5 @@ Models:
         Dataset: COCO
         Metrics:
           box AP: 37.5
-    Config: configs/nas/detnas/detnas_subnet_frcnn_shufflenetv2_fpn_1x_coco.py
+    Config: configs/nas/mmdet/detnas/detnas_frcnn_shufflenet_subnet_coco_1x.py
     Weights: https://download.openmmlab.com/mmrazor/v0.1/nas/detnas/detnas_subnet_frcnn_shufflenetv2_fpn_1x_coco/detnas_subnet_frcnn_shufflenetv2_fpn_1x_coco_bbox_backbone_flops-0.34M_mAP-37.5_20211222-67fea61f_mutable_cfg.yaml

configs/nas/spos/AngleNAS_SHUFFLENETV2_IN1k_2.0.yaml

@@ -1,24 +0,0 @@
-modules:
-  backbone.layer1.0: mb_k3e1
-  backbone.layer2.0: mb_k5e3
-  backbone.layer2.1: mb_k5e3
-  backbone.layer2.2: identity
-  backbone.layer2.3: mb_k3e3
-  backbone.layer3.0: mb_k3e3
-  backbone.layer3.1: identity
-  backbone.layer3.2: identity
-  backbone.layer3.3: mb_k3e3
-  backbone.layer4.0: mb_k7e6
-  backbone.layer4.1: identity
-  backbone.layer4.2: mb_k7e3
-  backbone.layer4.3: mb_k7e3
-  backbone.layer5.0: mb_k3e3
-  backbone.layer5.1: mb_k3e3
-  backbone.layer5.2: mb_k7e3
-  backbone.layer5.3: mb_k5e3
-  backbone.layer6.0: mb_k5e6
-  backbone.layer6.1: mb_k7e3
-  backbone.layer6.2: mb_k7e3
-  backbone.layer6.3: mb_k7e3
-  backbone.layer7.0: mb_k5e6
-channels:

configs/nas/spos/SPOS_SHUFFLENETV2_330M_IN1k_PAPER_2.0.yaml

@@ -1,22 +0,0 @@
-modules:
-  backbone.layers.0.0: shuffle_7x7
-  backbone.layers.0.1: shuffle_3x3
-  backbone.layers.0.2: shuffle_7x7
-  backbone.layers.0.3: shuffle_3x3
-  backbone.layers.1.0: shuffle_xception
-  backbone.layers.1.1: shuffle_5x5
-  backbone.layers.1.2: shuffle_5x5
-  backbone.layers.1.3: shuffle_3x3
-  backbone.layers.2.0: shuffle_3x3
-  backbone.layers.2.1: shuffle_5x5
-  backbone.layers.2.2: shuffle_3x3
-  backbone.layers.2.3: shuffle_5x5
-  backbone.layers.2.4: shuffle_3x3
-  backbone.layers.2.5: shuffle_xception
-  backbone.layers.2.6: shuffle_5x5
-  backbone.layers.2.7: shuffle_7x7
-  backbone.layers.3.0: shuffle_7x7
-  backbone.layers.3.1: shuffle_3x3
-  backbone.layers.3.2: shuffle_5x5
-  backbone.layers.3.3: shuffle_xception
-channels:

configs/nas/spos/spos_supernet_mobilenet_proxyless_gpu_8xb128_in1k_2.0.py

@@ -1,245 +0,0 @@
-# dataset settings
-dataset_type = 'ImageNet'
-preprocess_cfg = dict(
-    # RGB format normalization parameters
-    mean=[0., 0., 0.],
-    std=[1., 1., 1.],
-    # convert image from BGR to RGB
-    to_rgb=False,
-)
-
-file_client_args = dict(
-    backend='petrel',
-    path_mapping=dict({
-        './data/imagenet':
-        'sproject:s3://openmmlab/datasets/classification/imagenet',
-        'data/imagenet':
-        'sproject:s3://openmmlab/datasets/classification/imagenet'
-    }))
-
-train_pipeline = [
-    dict(type='LoadImageFromFile', file_client_args=file_client_args),
-    dict(type='RandomResizedCrop', scale=224),
-    dict(type='ColorJitter', brightness=0.4, contrast=0.4, saturation=0.4),
-    dict(type='RandomFlip', prob=0.5, direction='horizontal'),
-    dict(type='PackClsInputs'),
-]
-
-test_pipeline = [
-    dict(type='LoadImageFromFile', file_client_args=file_client_args),
-    dict(
-        type='ResizeEdge',
-        scale=256,
-        edge='short',
-        backend='pillow',
-        interpolation='bicubic'),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='PackClsInputs'),
-]
-
-train_dataloader = dict(
-    batch_size=128,
-    num_workers=8,
-    dataset=dict(
-        type=dataset_type,
-        data_root='/mnt/cache/share/images',
-        ann_file='meta/train.txt',
-        data_prefix='train',
-        pipeline=train_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=True),
-    persistent_workers=True,
-)
-
-# /mnt/lustre/share_data/wangjiaqi/data/imagenet',
-
-val_dataloader = dict(
-    batch_size=128,
-    num_workers=8,
-    dataset=dict(
-        type=dataset_type,
-        data_root='/mnt/cache/share/images',
-        ann_file='meta/val.txt',
-        data_prefix='val',
-        pipeline=test_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    persistent_workers=True,
-)
-val_evaluator = dict(type='Accuracy', topk=(1, 5))
-
-# If you want standard test, please manually configure the test dataset
-test_dataloader = val_dataloader
-test_evaluator = val_evaluator
-
-# scheduler
-
-# optimizer
-optim_wrapper = dict(
-    optimizer=dict(type='SGD', lr=0.5, momentum=0.9, weight_decay=4e-5),
-    clip_grad=None)
-
-# leanring policy
-param_scheduler = [
-    dict(type='PolyLR', power=1.0, eta_min=0.0, by_epoch=False, end=300000),
-]
-
-# train, val, test setting
-train_cfg = dict(by_epoch=False, max_iters=300000)
-val_cfg = dict()
-test_cfg = dict()
-
-# runtime
-
-# defaults to use registries in mmrazor
-default_scope = 'mmcls'
-
-log_processor = dict(
-    window_size=100,
-    by_epoch=False,
-    custom_cfg=[
-        dict(
-            data_src='loss',
-            log_name='loss_large_window',
-            method_name='mean',
-            window_size=100)
-    ])
-
-# configure default hooks
-default_hooks = dict(
-    timer=dict(type='IterTimerHook'),
-    logger=dict(type='LoggerHook', interval=100),
-    param_scheduler=dict(type='ParamSchedulerHook'),
-    checkpoint=dict(
-        type='CheckpointHook',
-        by_epoch=False,
-        interval=10000,
-        save_last=True,
-        max_keep_ckpts=3),
-    sampler_seed=dict(type='DistSamplerSeedHook'),
-    visualization=dict(type='VisualizationHook', enable=False),
-)
-
-# configure environment
-env_cfg = dict(
-    cudnn_benchmark=False,
-    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
-    dist_cfg=dict(backend='nccl'),
-)
-
-# set visualizer
-visualizer = None
-# dict(type='ClsVisualizer', vis_backends=vis_backends, name='visualizer')
-# vis_backends = [dict(type='LocalVisBackend')]
-
-# set log level
-log_level = 'INFO'
-
-# load from which checkpoint
-load_from = None
-
-# whether to resume training from the loaded checkpoint
-resume = False
-
-# model
-norm_cfg = dict(type='BN')
-_STAGE_MUTABLE = dict(
-    _scope_='mmrazor',
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e3=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=3,
-            norm_cfg=norm_cfg,
-            act_cfg=dict(type='ReLU6')),
-        mb_k5e3=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=3,
-            norm_cfg=norm_cfg,
-            act_cfg=dict(type='ReLU6')),
-        mb_k7e3=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=3,
-            norm_cfg=norm_cfg,
-            act_cfg=dict(type='ReLU6')),
-        mb_k3e6=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=6,
-            norm_cfg=norm_cfg,
-            act_cfg=dict(type='ReLU6')),
-        mb_k5e6=dict(
-            type='MBBlock',
-            kernel_size=5,
-            expand_ratio=6,
-            norm_cfg=norm_cfg,
-            act_cfg=dict(type='ReLU6')),
-        mb_k7e6=dict(
-            type='MBBlock',
-            kernel_size=7,
-            expand_ratio=6,
-            norm_cfg=norm_cfg,
-            act_cfg=dict(type='ReLU6')),
-        identity=dict(type='Identity'),
-    ))
-
-_FIRST_MUTABLE = dict(
-    _scope_='mmrazor',
-    type='OneShotMutableOP',
-    candidates=dict(
-        mb_k3e1=dict(
-            type='MBBlock',
-            kernel_size=3,
-            expand_ratio=1,
-            norm_cfg=norm_cfg,
-            act_cfg=dict(type='ReLU6')), ))
-
-arch_setting = [
-    # Parameters to build layers. 3 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, mutable_cfg.
-    [24, 1, 1, _FIRST_MUTABLE],
-    [32, 4, 2, _STAGE_MUTABLE],
-    [56, 4, 2, _STAGE_MUTABLE],
-    [112, 4, 2, _STAGE_MUTABLE],
-    [128, 4, 1, _STAGE_MUTABLE],
-    [256, 4, 2, _STAGE_MUTABLE],
-    [432, 1, 1, _STAGE_MUTABLE]
-]
-
-norm_cfg = dict(type='BN')
-supernet = dict(
-    type='ImageClassifier',
-    data_preprocessor=preprocess_cfg,
-    backbone=dict(
-        _scope_='mmrazor',
-        type='SearchableMobileNet',
-        first_channels=40,
-        last_channels=1728,
-        widen_factor=1.0,
-        norm_cfg=norm_cfg,
-        arch_setting=arch_setting),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=1728,
-        loss=dict(
-            type='LabelSmoothLoss',
-            num_classes=1000,
-            label_smooth_val=0.1,
-            mode='original',
-            loss_weight=1.0),
-        topk=(1, 5),
-    ),
-)
-
-mutator = dict(type='mmrazor.OneShotModuleMutator')
-
-model = dict(
-    type='mmrazor.SPOS',
-    architecture=supernet,
-    mutator=mutator,
-)
-
-find_unused_parameters = True

configs/nas/spos/spos_supernet_shufflenetv2_8xb128_in1k_2.0_example.py

@@ -1,214 +0,0 @@
-# dataset settings
-dataset_type = 'ImageNet'
-preprocess_cfg = dict(
-    # RGB format normalization parameters
-    mean=[123.675, 116.28, 103.53],
-    std=[58.395, 57.12, 57.375],
-    # convert image from BGR to RGB
-    to_rgb=True,
-)
-
-file_client_args = dict(
-    backend='petrel',
-    path_mapping=dict({
-        './data/imagenet':
-        'sproject:s3://openmmlab/datasets/classification/imagenet',
-        'data/imagenet':
-        'sproject:s3://openmmlab/datasets/classification/imagenet'
-    }))
-
-train_pipeline = [
-    dict(type='LoadImageFromFile', file_client_args=file_client_args),
-    dict(type='RandomResizedCrop', scale=224),
-    dict(type='ColorJitter', brightness=0.4, contrast=0.4, saturation=0.4),
-    dict(type='RandomFlip', prob=0.5, direction='horizontal'),
-    dict(type='PackClsInputs'),
-]
-
-test_pipeline = [
-    dict(type='LoadImageFromFile', file_client_args=file_client_args),
-    dict(type='ResizeEdge', scale=256, edge='short', backend='cv2'),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='PackClsInputs'),
-]
-
-train_dataloader = dict(
-    batch_size=128,
-    num_workers=5,
-    dataset=dict(
-        type=dataset_type,
-        data_root='/mnt/cache/share/images',
-        ann_file='meta/train.txt',
-        data_prefix='train',
-        pipeline=train_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=True),
-    persistent_workers=True,
-)
-
-# /mnt/lustre/share_data/wangjiaqi/data/imagenet',
-
-val_dataloader = dict(
-    batch_size=128,
-    num_workers=5,
-    dataset=dict(
-        type=dataset_type,
-        data_root='/mnt/cache/share/images',
-        ann_file='meta/val.txt',
-        data_prefix='val',
-        pipeline=test_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    persistent_workers=True,
-)
-val_evaluator = dict(type='Accuracy', topk=(1, 5))
-
-# If you want standard test, please manually configure the test dataset
-test_dataloader = val_dataloader
-test_evaluator = val_evaluator
-
-# scheduler
-
-# optimizer
-optim_wrapper = dict(
-    optimizer=dict(type='SGD', lr=0.5, momentum=0.9, weight_decay=4e-5),
-    clip_grad=None)
-
-# leanring policy
-param_scheduler = [
-    dict(type='PolyLR', power=1.0, eta_min=0.0, by_epoch=False, end=300000),
-]
-
-# train, val, test setting
-train_cfg = dict(by_epoch=False, max_iters=300000)
-val_cfg = dict()
-test_cfg = dict()
-
-# runtime
-
-# defaults to use registries in mmrazor
-default_scope = 'mmcls'
-
-log_processor = dict(
-    window_size=100,
-    by_epoch=False,
-    custom_cfg=[
-        dict(
-            data_src='loss',
-            log_name='loss_large_window',
-            method_name='mean',
-            window_size=100)
-    ])
-
-# configure default hooks
-default_hooks = dict(
-    # record the time of every iteration.
-    timer=dict(type='IterTimerHook'),
-
-    # print log every 100 iterations.
-    logger=dict(type='LoggerHook', interval=100),
-
-    # enable the parameter scheduler.
-    param_scheduler=dict(type='ParamSchedulerHook'),
-
-    # save checkpoint per epoch.
-    checkpoint=dict(
-        type='CheckpointHook',
-        by_epoch=False,
-        interval=10000,
-        save_last=True,
-        max_keep_ckpts=3),
-
-    # set sampler seed in distributed evrionment.
-    sampler_seed=dict(type='DistSamplerSeedHook'),
-
-    # validation results visualization, set True to enable it.
-    visualization=dict(type='VisualizationHook', enable=False),
-)
-
-# configure environment
-env_cfg = dict(
-    # whether to enable cudnn benchmark
-    cudnn_benchmark=False,
-
-    # set multi process parameters
-    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
-
-    # set distributed parameters
-    dist_cfg=dict(backend='nccl'),
-)
-
-# set visualizer
-visualizer = None
-# dict(type='ClsVisualizer', vis_backends=vis_backends, name='visualizer')
-# vis_backends = [dict(type='LocalVisBackend')]
-
-# set log level
-log_level = 'INFO'
-
-# load from which checkpoint
-load_from = None
-
-# "/mnt/lustre/dongpeijie/spos_shufflenetv2_subnet_8xb128_in1k_flops_0.33M_acc_73.87_20211222-1f0a0b4d.pth"
-
-# whether to resume training from the loaded checkpoint
-resume = False
-
-# model
-
-_STAGE_MUTABLE = dict(
-    _scope_='mmrazor',
-    type='OneShotMutableOP',
-    candidates=dict(
-        shuffle_3x3=dict(
-            type='ShuffleBlock', kernel_size=3, norm_cfg=dict(type='BN')),
-        shuffle_5x5=dict(
-            type='ShuffleBlock', kernel_size=5, norm_cfg=dict(type='BN')),
-        shuffle_7x7=dict(
-            type='ShuffleBlock', kernel_size=7, norm_cfg=dict(type='BN')),
-        shuffle_xception=dict(
-            type='ShuffleXception', norm_cfg=dict(type='BN')),
-    ))
-
-arch_setting = [
-    # Parameters to build layers. 3 parameters are needed to construct a
-    # layer, from left to right: channel, num_blocks, mutable_cfg.
-    [64, 4, _STAGE_MUTABLE],
-    [160, 4, _STAGE_MUTABLE],
-    [320, 8, _STAGE_MUTABLE],
-    [640, 4, _STAGE_MUTABLE],
-]
-
-norm_cfg = dict(type='BN')
-supernet = dict(
-    type='ImageClassifier',
-    data_preprocessor=preprocess_cfg,
-    backbone=dict(
-        _scope_='mmrazor',
-        type='SearchableShuffleNetV2',
-        widen_factor=1.0,
-        norm_cfg=norm_cfg,
-        arch_setting=arch_setting),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=1024,
-        loss=dict(
-            type='LabelSmoothLoss',
-            num_classes=1000,
-            label_smooth_val=0.1,
-            mode='original',
-            loss_weight=1.0),
-        topk=(1, 5),
-    ),
-)
-
-mutator = dict(type='mmrazor.OneShotModuleMutator')
-
-model = dict(
-    type='mmrazor.SPOS',
-    architecture=supernet,
-    mutator=mutator,
-    # fix_subnet='configs/nas/spos/SPOS_SHUFFLENETV2_330M_IN1k_PAPER_2.0.yaml'
-)
-
-find_unused_parameters = True

configs/pruning/autoslim/AUTOSLIM_MBV2_220M_OFFICIAL.yaml

@@ -1,421 +0,0 @@
-backbone.conv1.bn:
-  out_channels: 8
-  raw_out_channels: 48
-backbone.conv1.conv:
-  in_channels: 3
-  out_channels: 8
-  raw_in_channels: 3
-  raw_out_channels: 48
-backbone.conv2.bn:
-  out_channels: 1920
-  raw_out_channels: 1920
-backbone.conv2.conv:
-  in_channels: 280
-  out_channels: 1920
-  raw_in_channels: 480
-  raw_out_channels: 1920
-backbone.layer1.0.conv.0.bn:
-  out_channels: 8
-  raw_out_channels: 48
-backbone.layer1.0.conv.0.conv:
-  in_channels: 8
-  out_channels: 8
-  raw_in_channels: 48
-  raw_out_channels: 48
-backbone.layer1.0.conv.1.bn:
-  out_channels: 8
-  raw_out_channels: 24
-backbone.layer1.0.conv.1.conv:
-  in_channels: 8
-  out_channels: 8
-  raw_in_channels: 48
-  raw_out_channels: 24
-backbone.layer2.0.conv.0.bn:
-  out_channels: 96
-  raw_out_channels: 144
-backbone.layer2.0.conv.0.conv:
-  in_channels: 8
-  out_channels: 96
-  raw_in_channels: 24
-  raw_out_channels: 144
-backbone.layer2.0.conv.1.bn:
-  out_channels: 96
-  raw_out_channels: 144
-backbone.layer2.0.conv.1.conv:
-  in_channels: 96
-  out_channels: 96
-  raw_in_channels: 144
-  raw_out_channels: 144
-backbone.layer2.0.conv.2.bn:
-  out_channels: 16
-  raw_out_channels: 40
-backbone.layer2.0.conv.2.conv:
-  in_channels: 96
-  out_channels: 16
-  raw_in_channels: 144
-  raw_out_channels: 40
-backbone.layer2.1.conv.0.bn:
-  out_channels: 96
-  raw_out_channels: 240
-backbone.layer2.1.conv.0.conv:
-  in_channels: 16
-  out_channels: 96
-  raw_in_channels: 40
-  raw_out_channels: 240
-backbone.layer2.1.conv.1.bn:
-  out_channels: 96
-  raw_out_channels: 240
-backbone.layer2.1.conv.1.conv:
-  in_channels: 96
-  out_channels: 96
-  raw_in_channels: 240
-  raw_out_channels: 240
-backbone.layer2.1.conv.2.bn:
-  out_channels: 16
-  raw_out_channels: 40
-backbone.layer2.1.conv.2.conv:
-  in_channels: 96
-  out_channels: 16
-  raw_in_channels: 240
-  raw_out_channels: 40
-backbone.layer3.0.conv.0.bn:
-  out_channels: 96
-  raw_out_channels: 240
-backbone.layer3.0.conv.0.conv:
-  in_channels: 16
-  out_channels: 96
-  raw_in_channels: 40
-  raw_out_channels: 240
-backbone.layer3.0.conv.1.bn:
-  out_channels: 96
-  raw_out_channels: 240
-backbone.layer3.0.conv.1.conv:
-  in_channels: 96
-  out_channels: 96
-  raw_in_channels: 240
-  raw_out_channels: 240
-backbone.layer3.0.conv.2.bn:
-  out_channels: 24
-  raw_out_channels: 48
-backbone.layer3.0.conv.2.conv:
-  in_channels: 96
-  out_channels: 24
-  raw_in_channels: 240
-  raw_out_channels: 48
-backbone.layer3.1.conv.0.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.1.conv.0.conv:
-  in_channels: 24
-  out_channels: 144
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer3.1.conv.1.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.1.conv.1.conv:
-  in_channels: 144
-  out_channels: 144
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer3.1.conv.2.bn:
-  out_channels: 24
-  raw_out_channels: 48
-backbone.layer3.1.conv.2.conv:
-  in_channels: 144
-  out_channels: 24
-  raw_in_channels: 288
-  raw_out_channels: 48
-backbone.layer3.2.conv.0.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.2.conv.0.conv:
-  in_channels: 24
-  out_channels: 144
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer3.2.conv.1.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.2.conv.1.conv:
-  in_channels: 144
-  out_channels: 144
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer3.2.conv.2.bn:
-  out_channels: 24
-  raw_out_channels: 48
-backbone.layer3.2.conv.2.conv:
-  in_channels: 144
-  out_channels: 24
-  raw_in_channels: 288
-  raw_out_channels: 48
-backbone.layer4.0.conv.0.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer4.0.conv.0.conv:
-  in_channels: 24
-  out_channels: 144
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer4.0.conv.1.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer4.0.conv.1.conv:
-  in_channels: 144
-  out_channels: 144
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer4.0.conv.2.bn:
-  out_channels: 48
-  raw_out_channels: 96
-backbone.layer4.0.conv.2.conv:
-  in_channels: 144
-  out_channels: 48
-  raw_in_channels: 288
-  raw_out_channels: 96
-backbone.layer4.1.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.1.conv.0.conv:
-  in_channels: 48
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.1.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.1.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.1.conv.2.bn:
-  out_channels: 48
-  raw_out_channels: 96
-backbone.layer4.1.conv.2.conv:
-  in_channels: 288
-  out_channels: 48
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer4.2.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.2.conv.0.conv:
-  in_channels: 48
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.2.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.2.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.2.conv.2.bn:
-  out_channels: 48
-  raw_out_channels: 96
-backbone.layer4.2.conv.2.conv:
-  in_channels: 288
-  out_channels: 48
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer4.3.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.3.conv.0.conv:
-  in_channels: 48
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.3.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.3.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.3.conv.2.bn:
-  out_channels: 48
-  raw_out_channels: 96
-backbone.layer4.3.conv.2.conv:
-  in_channels: 288
-  out_channels: 48
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer5.0.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer5.0.conv.0.conv:
-  in_channels: 48
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer5.0.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer5.0.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer5.0.conv.2.bn:
-  out_channels: 64
-  raw_out_channels: 144
-backbone.layer5.0.conv.2.conv:
-  in_channels: 288
-  out_channels: 64
-  raw_in_channels: 576
-  raw_out_channels: 144
-backbone.layer5.1.conv.0.bn:
-  out_channels: 432
-  raw_out_channels: 864
-backbone.layer5.1.conv.0.conv:
-  in_channels: 64
-  out_channels: 432
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer5.1.conv.1.bn:
-  out_channels: 432
-  raw_out_channels: 864
-backbone.layer5.1.conv.1.conv:
-  in_channels: 432
-  out_channels: 432
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer5.1.conv.2.bn:
-  out_channels: 64
-  raw_out_channels: 144
-backbone.layer5.1.conv.2.conv:
-  in_channels: 432
-  out_channels: 64
-  raw_in_channels: 864
-  raw_out_channels: 144
-backbone.layer5.2.conv.0.bn:
-  out_channels: 432
-  raw_out_channels: 864
-backbone.layer5.2.conv.0.conv:
-  in_channels: 64
-  out_channels: 432
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer5.2.conv.1.bn:
-  out_channels: 432
-  raw_out_channels: 864
-backbone.layer5.2.conv.1.conv:
-  in_channels: 432
-  out_channels: 432
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer5.2.conv.2.bn:
-  out_channels: 64
-  raw_out_channels: 144
-backbone.layer5.2.conv.2.conv:
-  in_channels: 432
-  out_channels: 64
-  raw_in_channels: 864
-  raw_out_channels: 144
-backbone.layer6.0.conv.0.bn:
-  out_channels: 648
-  raw_out_channels: 864
-backbone.layer6.0.conv.0.conv:
-  in_channels: 64
-  out_channels: 648
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer6.0.conv.1.bn:
-  out_channels: 648
-  raw_out_channels: 864
-backbone.layer6.0.conv.1.conv:
-  in_channels: 648
-  out_channels: 648
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer6.0.conv.2.bn:
-  out_channels: 176
-  raw_out_channels: 240
-backbone.layer6.0.conv.2.conv:
-  in_channels: 648
-  out_channels: 176
-  raw_in_channels: 864
-  raw_out_channels: 240
-backbone.layer6.1.conv.0.bn:
-  out_channels: 720
-  raw_out_channels: 1440
-backbone.layer6.1.conv.0.conv:
-  in_channels: 176
-  out_channels: 720
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer6.1.conv.1.bn:
-  out_channels: 720
-  raw_out_channels: 1440
-backbone.layer6.1.conv.1.conv:
-  in_channels: 720
-  out_channels: 720
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.1.conv.2.bn:
-  out_channels: 176
-  raw_out_channels: 240
-backbone.layer6.1.conv.2.conv:
-  in_channels: 720
-  out_channels: 176
-  raw_in_channels: 1440
-  raw_out_channels: 240
-backbone.layer6.2.conv.0.bn:
-  out_channels: 720
-  raw_out_channels: 1440
-backbone.layer6.2.conv.0.conv:
-  in_channels: 176
-  out_channels: 720
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer6.2.conv.1.bn:
-  out_channels: 720
-  raw_out_channels: 1440
-backbone.layer6.2.conv.1.conv:
-  in_channels: 720
-  out_channels: 720
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.2.conv.2.bn:
-  out_channels: 176
-  raw_out_channels: 240
-backbone.layer6.2.conv.2.conv:
-  in_channels: 720
-  out_channels: 176
-  raw_in_channels: 1440
-  raw_out_channels: 240
-backbone.layer7.0.conv.0.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.0.conv:
-  in_channels: 176
-  out_channels: 1440
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer7.0.conv.1.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.1.conv:
-  in_channels: 1440
-  out_channels: 1440
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.2.bn:
-  out_channels: 280
-  raw_out_channels: 480
-backbone.layer7.0.conv.2.conv:
-  in_channels: 1440
-  out_channels: 280
-  raw_in_channels: 1440
-  raw_out_channels: 480
-head.fc:
-  in_channels: 1920
-  out_channels: 1000
-  raw_in_channels: 1920
-  raw_out_channels: 1000

configs/pruning/autoslim/AUTOSLIM_MBV2_320M_OFFICIAL.yaml

@@ -1,421 +0,0 @@
-backbone.conv1.bn:
-  out_channels: 8
-  raw_out_channels: 48
-backbone.conv1.conv:
-  in_channels: 3
-  out_channels: 8
-  raw_in_channels: 3
-  raw_out_channels: 48
-backbone.conv2.bn:
-  out_channels: 1920
-  raw_out_channels: 1920
-backbone.conv2.conv:
-  in_channels: 480
-  out_channels: 1920
-  raw_in_channels: 480
-  raw_out_channels: 1920
-backbone.layer1.0.conv.0.bn:
-  out_channels: 8
-  raw_out_channels: 48
-backbone.layer1.0.conv.0.conv:
-  in_channels: 8
-  out_channels: 8
-  raw_in_channels: 48
-  raw_out_channels: 48
-backbone.layer1.0.conv.1.bn:
-  out_channels: 8
-  raw_out_channels: 24
-backbone.layer1.0.conv.1.conv:
-  in_channels: 8
-  out_channels: 8
-  raw_in_channels: 48
-  raw_out_channels: 24
-backbone.layer2.0.conv.0.bn:
-  out_channels: 96
-  raw_out_channels: 144
-backbone.layer2.0.conv.0.conv:
-  in_channels: 8
-  out_channels: 96
-  raw_in_channels: 24
-  raw_out_channels: 144
-backbone.layer2.0.conv.1.bn:
-  out_channels: 96
-  raw_out_channels: 144
-backbone.layer2.0.conv.1.conv:
-  in_channels: 96
-  out_channels: 96
-  raw_in_channels: 144
-  raw_out_channels: 144
-backbone.layer2.0.conv.2.bn:
-  out_channels: 16
-  raw_out_channels: 40
-backbone.layer2.0.conv.2.conv:
-  in_channels: 96
-  out_channels: 16
-  raw_in_channels: 144
-  raw_out_channels: 40
-backbone.layer2.1.conv.0.bn:
-  out_channels: 96
-  raw_out_channels: 240
-backbone.layer2.1.conv.0.conv:
-  in_channels: 16
-  out_channels: 96
-  raw_in_channels: 40
-  raw_out_channels: 240
-backbone.layer2.1.conv.1.bn:
-  out_channels: 96
-  raw_out_channels: 240
-backbone.layer2.1.conv.1.conv:
-  in_channels: 96
-  out_channels: 96
-  raw_in_channels: 240
-  raw_out_channels: 240
-backbone.layer2.1.conv.2.bn:
-  out_channels: 16
-  raw_out_channels: 40
-backbone.layer2.1.conv.2.conv:
-  in_channels: 96
-  out_channels: 16
-  raw_in_channels: 240
-  raw_out_channels: 40
-backbone.layer3.0.conv.0.bn:
-  out_channels: 96
-  raw_out_channels: 240
-backbone.layer3.0.conv.0.conv:
-  in_channels: 16
-  out_channels: 96
-  raw_in_channels: 40
-  raw_out_channels: 240
-backbone.layer3.0.conv.1.bn:
-  out_channels: 96
-  raw_out_channels: 240
-backbone.layer3.0.conv.1.conv:
-  in_channels: 96
-  out_channels: 96
-  raw_in_channels: 240
-  raw_out_channels: 240
-backbone.layer3.0.conv.2.bn:
-  out_channels: 24
-  raw_out_channels: 48
-backbone.layer3.0.conv.2.conv:
-  in_channels: 96
-  out_channels: 24
-  raw_in_channels: 240
-  raw_out_channels: 48
-backbone.layer3.1.conv.0.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.1.conv.0.conv:
-  in_channels: 24
-  out_channels: 144
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer3.1.conv.1.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.1.conv.1.conv:
-  in_channels: 144
-  out_channels: 144
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer3.1.conv.2.bn:
-  out_channels: 24
-  raw_out_channels: 48
-backbone.layer3.1.conv.2.conv:
-  in_channels: 144
-  out_channels: 24
-  raw_in_channels: 288
-  raw_out_channels: 48
-backbone.layer3.2.conv.0.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.2.conv.0.conv:
-  in_channels: 24
-  out_channels: 144
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer3.2.conv.1.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.2.conv.1.conv:
-  in_channels: 144
-  out_channels: 144
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer3.2.conv.2.bn:
-  out_channels: 24
-  raw_out_channels: 48
-backbone.layer3.2.conv.2.conv:
-  in_channels: 144
-  out_channels: 24
-  raw_in_channels: 288
-  raw_out_channels: 48
-backbone.layer4.0.conv.0.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer4.0.conv.0.conv:
-  in_channels: 24
-  out_channels: 144
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer4.0.conv.1.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer4.0.conv.1.conv:
-  in_channels: 144
-  out_channels: 144
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer4.0.conv.2.bn:
-  out_channels: 56
-  raw_out_channels: 96
-backbone.layer4.0.conv.2.conv:
-  in_channels: 144
-  out_channels: 56
-  raw_in_channels: 288
-  raw_out_channels: 96
-backbone.layer4.1.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.1.conv.0.conv:
-  in_channels: 56
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.1.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.1.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.1.conv.2.bn:
-  out_channels: 56
-  raw_out_channels: 96
-backbone.layer4.1.conv.2.conv:
-  in_channels: 288
-  out_channels: 56
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer4.2.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.2.conv.0.conv:
-  in_channels: 56
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.2.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.2.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.2.conv.2.bn:
-  out_channels: 56
-  raw_out_channels: 96
-backbone.layer4.2.conv.2.conv:
-  in_channels: 288
-  out_channels: 56
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer4.3.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.3.conv.0.conv:
-  in_channels: 56
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.3.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.3.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.3.conv.2.bn:
-  out_channels: 56
-  raw_out_channels: 96
-backbone.layer4.3.conv.2.conv:
-  in_channels: 288
-  out_channels: 56
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer5.0.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer5.0.conv.0.conv:
-  in_channels: 56
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer5.0.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer5.0.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer5.0.conv.2.bn:
-  out_channels: 96
-  raw_out_channels: 144
-backbone.layer5.0.conv.2.conv:
-  in_channels: 288
-  out_channels: 96
-  raw_in_channels: 576
-  raw_out_channels: 144
-backbone.layer5.1.conv.0.bn:
-  out_channels: 432
-  raw_out_channels: 864
-backbone.layer5.1.conv.0.conv:
-  in_channels: 96
-  out_channels: 432
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer5.1.conv.1.bn:
-  out_channels: 432
-  raw_out_channels: 864
-backbone.layer5.1.conv.1.conv:
-  in_channels: 432
-  out_channels: 432
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer5.1.conv.2.bn:
-  out_channels: 96
-  raw_out_channels: 144
-backbone.layer5.1.conv.2.conv:
-  in_channels: 432
-  out_channels: 96
-  raw_in_channels: 864
-  raw_out_channels: 144
-backbone.layer5.2.conv.0.bn:
-  out_channels: 432
-  raw_out_channels: 864
-backbone.layer5.2.conv.0.conv:
-  in_channels: 96
-  out_channels: 432
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer5.2.conv.1.bn:
-  out_channels: 432
-  raw_out_channels: 864
-backbone.layer5.2.conv.1.conv:
-  in_channels: 432
-  out_channels: 432
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer5.2.conv.2.bn:
-  out_channels: 96
-  raw_out_channels: 144
-backbone.layer5.2.conv.2.conv:
-  in_channels: 432
-  out_channels: 96
-  raw_in_channels: 864
-  raw_out_channels: 144
-backbone.layer6.0.conv.0.bn:
-  out_channels: 864
-  raw_out_channels: 864
-backbone.layer6.0.conv.0.conv:
-  in_channels: 96
-  out_channels: 864
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer6.0.conv.1.bn:
-  out_channels: 864
-  raw_out_channels: 864
-backbone.layer6.0.conv.1.conv:
-  in_channels: 864
-  out_channels: 864
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer6.0.conv.2.bn:
-  out_channels: 240
-  raw_out_channels: 240
-backbone.layer6.0.conv.2.conv:
-  in_channels: 864
-  out_channels: 240
-  raw_in_channels: 864
-  raw_out_channels: 240
-backbone.layer6.1.conv.0.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.1.conv.0.conv:
-  in_channels: 240
-  out_channels: 1440
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer6.1.conv.1.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.1.conv.1.conv:
-  in_channels: 1440
-  out_channels: 1440
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.1.conv.2.bn:
-  out_channels: 240
-  raw_out_channels: 240
-backbone.layer6.1.conv.2.conv:
-  in_channels: 1440
-  out_channels: 240
-  raw_in_channels: 1440
-  raw_out_channels: 240
-backbone.layer6.2.conv.0.bn:
-  out_channels: 960
-  raw_out_channels: 1440
-backbone.layer6.2.conv.0.conv:
-  in_channels: 240
-  out_channels: 960
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer6.2.conv.1.bn:
-  out_channels: 960
-  raw_out_channels: 1440
-backbone.layer6.2.conv.1.conv:
-  in_channels: 960
-  out_channels: 960
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.2.conv.2.bn:
-  out_channels: 240
-  raw_out_channels: 240
-backbone.layer6.2.conv.2.conv:
-  in_channels: 960
-  out_channels: 240
-  raw_in_channels: 1440
-  raw_out_channels: 240
-backbone.layer7.0.conv.0.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.0.conv:
-  in_channels: 240
-  out_channels: 1440
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer7.0.conv.1.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.1.conv:
-  in_channels: 1440
-  out_channels: 1440
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.2.bn:
-  out_channels: 480
-  raw_out_channels: 480
-backbone.layer7.0.conv.2.conv:
-  in_channels: 1440
-  out_channels: 480
-  raw_in_channels: 1440
-  raw_out_channels: 480
-head.fc:
-  in_channels: 1920
-  out_channels: 1000
-  raw_in_channels: 1920
-  raw_out_channels: 1000

configs/pruning/autoslim/AUTOSLIM_MBV2_530M_OFFICIAL.yaml

@@ -1,421 +0,0 @@
-backbone.conv1.bn:
-  out_channels: 32
-  raw_out_channels: 48
-backbone.conv1.conv:
-  in_channels: 3
-  out_channels: 32
-  raw_in_channels: 3
-  raw_out_channels: 48
-backbone.conv2.bn:
-  out_channels: 1920
-  raw_out_channels: 1920
-backbone.conv2.conv:
-  in_channels: 480
-  out_channels: 1920
-  raw_in_channels: 480
-  raw_out_channels: 1920
-backbone.layer1.0.conv.0.bn:
-  out_channels: 32
-  raw_out_channels: 48
-backbone.layer1.0.conv.0.conv:
-  in_channels: 32
-  out_channels: 32
-  raw_in_channels: 48
-  raw_out_channels: 48
-backbone.layer1.0.conv.1.bn:
-  out_channels: 16
-  raw_out_channels: 24
-backbone.layer1.0.conv.1.conv:
-  in_channels: 32
-  out_channels: 16
-  raw_in_channels: 48
-  raw_out_channels: 24
-backbone.layer2.0.conv.0.bn:
-  out_channels: 144
-  raw_out_channels: 144
-backbone.layer2.0.conv.0.conv:
-  in_channels: 16
-  out_channels: 144
-  raw_in_channels: 24
-  raw_out_channels: 144
-backbone.layer2.0.conv.1.bn:
-  out_channels: 144
-  raw_out_channels: 144
-backbone.layer2.0.conv.1.conv:
-  in_channels: 144
-  out_channels: 144
-  raw_in_channels: 144
-  raw_out_channels: 144
-backbone.layer2.0.conv.2.bn:
-  out_channels: 24
-  raw_out_channels: 40
-backbone.layer2.0.conv.2.conv:
-  in_channels: 144
-  out_channels: 24
-  raw_in_channels: 144
-  raw_out_channels: 40
-backbone.layer2.1.conv.0.bn:
-  out_channels: 176
-  raw_out_channels: 240
-backbone.layer2.1.conv.0.conv:
-  in_channels: 24
-  out_channels: 176
-  raw_in_channels: 40
-  raw_out_channels: 240
-backbone.layer2.1.conv.1.bn:
-  out_channels: 176
-  raw_out_channels: 240
-backbone.layer2.1.conv.1.conv:
-  in_channels: 176
-  out_channels: 176
-  raw_in_channels: 240
-  raw_out_channels: 240
-backbone.layer2.1.conv.2.bn:
-  out_channels: 24
-  raw_out_channels: 40
-backbone.layer2.1.conv.2.conv:
-  in_channels: 176
-  out_channels: 24
-  raw_in_channels: 240
-  raw_out_channels: 40
-backbone.layer3.0.conv.0.bn:
-  out_channels: 192
-  raw_out_channels: 240
-backbone.layer3.0.conv.0.conv:
-  in_channels: 24
-  out_channels: 192
-  raw_in_channels: 40
-  raw_out_channels: 240
-backbone.layer3.0.conv.1.bn:
-  out_channels: 192
-  raw_out_channels: 240
-backbone.layer3.0.conv.1.conv:
-  in_channels: 192
-  out_channels: 192
-  raw_in_channels: 240
-  raw_out_channels: 240
-backbone.layer3.0.conv.2.bn:
-  out_channels: 48
-  raw_out_channels: 48
-backbone.layer3.0.conv.2.conv:
-  in_channels: 192
-  out_channels: 48
-  raw_in_channels: 240
-  raw_out_channels: 48
-backbone.layer3.1.conv.0.bn:
-  out_channels: 240
-  raw_out_channels: 288
-backbone.layer3.1.conv.0.conv:
-  in_channels: 48
-  out_channels: 240
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer3.1.conv.1.bn:
-  out_channels: 240
-  raw_out_channels: 288
-backbone.layer3.1.conv.1.conv:
-  in_channels: 240
-  out_channels: 240
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer3.1.conv.2.bn:
-  out_channels: 48
-  raw_out_channels: 48
-backbone.layer3.1.conv.2.conv:
-  in_channels: 240
-  out_channels: 48
-  raw_in_channels: 288
-  raw_out_channels: 48
-backbone.layer3.2.conv.0.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.2.conv.0.conv:
-  in_channels: 48
-  out_channels: 144
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer3.2.conv.1.bn:
-  out_channels: 144
-  raw_out_channels: 288
-backbone.layer3.2.conv.1.conv:
-  in_channels: 144
-  out_channels: 144
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer3.2.conv.2.bn:
-  out_channels: 48
-  raw_out_channels: 48
-backbone.layer3.2.conv.2.conv:
-  in_channels: 144
-  out_channels: 48
-  raw_in_channels: 288
-  raw_out_channels: 48
-backbone.layer4.0.conv.0.bn:
-  out_channels: 264
-  raw_out_channels: 288
-backbone.layer4.0.conv.0.conv:
-  in_channels: 48
-  out_channels: 264
-  raw_in_channels: 48
-  raw_out_channels: 288
-backbone.layer4.0.conv.1.bn:
-  out_channels: 264
-  raw_out_channels: 288
-backbone.layer4.0.conv.1.conv:
-  in_channels: 264
-  out_channels: 264
-  raw_in_channels: 288
-  raw_out_channels: 288
-backbone.layer4.0.conv.2.bn:
-  out_channels: 88
-  raw_out_channels: 96
-backbone.layer4.0.conv.2.conv:
-  in_channels: 264
-  out_channels: 88
-  raw_in_channels: 288
-  raw_out_channels: 96
-backbone.layer4.1.conv.0.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.1.conv.0.conv:
-  in_channels: 88
-  out_channels: 288
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.1.conv.1.bn:
-  out_channels: 288
-  raw_out_channels: 576
-backbone.layer4.1.conv.1.conv:
-  in_channels: 288
-  out_channels: 288
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.1.conv.2.bn:
-  out_channels: 88
-  raw_out_channels: 96
-backbone.layer4.1.conv.2.conv:
-  in_channels: 288
-  out_channels: 88
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer4.2.conv.0.bn:
-  out_channels: 336
-  raw_out_channels: 576
-backbone.layer4.2.conv.0.conv:
-  in_channels: 88
-  out_channels: 336
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.2.conv.1.bn:
-  out_channels: 336
-  raw_out_channels: 576
-backbone.layer4.2.conv.1.conv:
-  in_channels: 336
-  out_channels: 336
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.2.conv.2.bn:
-  out_channels: 88
-  raw_out_channels: 96
-backbone.layer4.2.conv.2.conv:
-  in_channels: 336
-  out_channels: 88
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer4.3.conv.0.bn:
-  out_channels: 432
-  raw_out_channels: 576
-backbone.layer4.3.conv.0.conv:
-  in_channels: 88
-  out_channels: 432
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer4.3.conv.1.bn:
-  out_channels: 432
-  raw_out_channels: 576
-backbone.layer4.3.conv.1.conv:
-  in_channels: 432
-  out_channels: 432
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer4.3.conv.2.bn:
-  out_channels: 88
-  raw_out_channels: 96
-backbone.layer4.3.conv.2.conv:
-  in_channels: 432
-  out_channels: 88
-  raw_in_channels: 576
-  raw_out_channels: 96
-backbone.layer5.0.conv.0.bn:
-  out_channels: 576
-  raw_out_channels: 576
-backbone.layer5.0.conv.0.conv:
-  in_channels: 88
-  out_channels: 576
-  raw_in_channels: 96
-  raw_out_channels: 576
-backbone.layer5.0.conv.1.bn:
-  out_channels: 576
-  raw_out_channels: 576
-backbone.layer5.0.conv.1.conv:
-  in_channels: 576
-  out_channels: 576
-  raw_in_channels: 576
-  raw_out_channels: 576
-backbone.layer5.0.conv.2.bn:
-  out_channels: 144
-  raw_out_channels: 144
-backbone.layer5.0.conv.2.conv:
-  in_channels: 576
-  out_channels: 144
-  raw_in_channels: 576
-  raw_out_channels: 144
-backbone.layer5.1.conv.0.bn:
-  out_channels: 576
-  raw_out_channels: 864
-backbone.layer5.1.conv.0.conv:
-  in_channels: 144
-  out_channels: 576
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer5.1.conv.1.bn:
-  out_channels: 576
-  raw_out_channels: 864
-backbone.layer5.1.conv.1.conv:
-  in_channels: 576
-  out_channels: 576
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer5.1.conv.2.bn:
-  out_channels: 144
-  raw_out_channels: 144
-backbone.layer5.1.conv.2.conv:
-  in_channels: 576
-  out_channels: 144
-  raw_in_channels: 864
-  raw_out_channels: 144
-backbone.layer5.2.conv.0.bn:
-  out_channels: 648
-  raw_out_channels: 864
-backbone.layer5.2.conv.0.conv:
-  in_channels: 144
-  out_channels: 648
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer5.2.conv.1.bn:
-  out_channels: 648
-  raw_out_channels: 864
-backbone.layer5.2.conv.1.conv:
-  in_channels: 648
-  out_channels: 648
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer5.2.conv.2.bn:
-  out_channels: 144
-  raw_out_channels: 144
-backbone.layer5.2.conv.2.conv:
-  in_channels: 648
-  out_channels: 144
-  raw_in_channels: 864
-  raw_out_channels: 144
-backbone.layer6.0.conv.0.bn:
-  out_channels: 864
-  raw_out_channels: 864
-backbone.layer6.0.conv.0.conv:
-  in_channels: 144
-  out_channels: 864
-  raw_in_channels: 144
-  raw_out_channels: 864
-backbone.layer6.0.conv.1.bn:
-  out_channels: 864
-  raw_out_channels: 864
-backbone.layer6.0.conv.1.conv:
-  in_channels: 864
-  out_channels: 864
-  raw_in_channels: 864
-  raw_out_channels: 864
-backbone.layer6.0.conv.2.bn:
-  out_channels: 240
-  raw_out_channels: 240
-backbone.layer6.0.conv.2.conv:
-  in_channels: 864
-  out_channels: 240
-  raw_in_channels: 864
-  raw_out_channels: 240
-backbone.layer6.1.conv.0.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.1.conv.0.conv:
-  in_channels: 240
-  out_channels: 1440
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer6.1.conv.1.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.1.conv.1.conv:
-  in_channels: 1440
-  out_channels: 1440
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.1.conv.2.bn:
-  out_channels: 240
-  raw_out_channels: 240
-backbone.layer6.1.conv.2.conv:
-  in_channels: 1440
-  out_channels: 240
-  raw_in_channels: 1440
-  raw_out_channels: 240
-backbone.layer6.2.conv.0.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.2.conv.0.conv:
-  in_channels: 240
-  out_channels: 1440
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer6.2.conv.1.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.2.conv.1.conv:
-  in_channels: 1440
-  out_channels: 1440
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer6.2.conv.2.bn:
-  out_channels: 240
-  raw_out_channels: 240
-backbone.layer6.2.conv.2.conv:
-  in_channels: 1440
-  out_channels: 240
-  raw_in_channels: 1440
-  raw_out_channels: 240
-backbone.layer7.0.conv.0.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.0.conv:
-  in_channels: 240
-  out_channels: 1440
-  raw_in_channels: 240
-  raw_out_channels: 1440
-backbone.layer7.0.conv.1.bn:
-  out_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.1.conv:
-  in_channels: 1440
-  out_channels: 1440
-  raw_in_channels: 1440
-  raw_out_channels: 1440
-backbone.layer7.0.conv.2.bn:
-  out_channels: 480
-  raw_out_channels: 480
-backbone.layer7.0.conv.2.conv:
-  in_channels: 1440
-  out_channels: 480
-  raw_in_channels: 1440
-  raw_out_channels: 480
-head.fc:
-  in_channels: 1920
-  out_channels: 1000
-  raw_in_channels: 1920
-  raw_out_channels: 1000

configs/pruning/autoslim/autoslim_mbv2_search_8xb1024_in1k.py

@@ -1,13 +0,0 @@
-_base_ = [
-    './autoslim_mbv2_supernet_8xb256_in1k.py',
-]
-
-algorithm = dict(distiller=None, input_shape=(3, 224, 224))
-
-searcher = dict(
-    type='GreedySearcher',
-    target_flops=[500000000, 300000000, 200000000],
-    max_channel_bins=12,
-    metrics='accuracy')
-
-data = dict(samples_per_gpu=1024, workers_per_gpu=4)

configs/pruning/autoslim/autoslim_mbv2_subnet_8xb256_in1k.py

@@ -1,29 +0,0 @@
-_base_ = [
-    './autoslim_mbv2_supernet_8xb256_in1k.py',
-]
-
-model = dict(
-    head=dict(
-        loss=dict(
-            type='LabelSmoothLoss',
-            mode='original',
-            label_smooth_val=0.1,
-            loss_weight=1.0)))
-
-# FIXME: you may replace this with the channel_cfg searched by yourself
-channel_cfg = [
-    'https://download.openmmlab.com/mmrazor/v0.1/pruning/autoslim/autoslim_mbv2_subnet_8xb256_in1k/autoslim_mbv2_subnet_8xb256_in1k_flops-0.53M_acc-74.23_20211222-e5208bbd_channel_cfg.yaml',  # noqa: E501
-    'https://download.openmmlab.com/mmrazor/v0.1/pruning/autoslim/autoslim_mbv2_subnet_8xb256_in1k/autoslim_mbv2_subnet_8xb256_in1k_flops-0.32M_acc-72.73_20211222-b5b0b33c_channel_cfg.yaml',  # noqa: E501
-    'https://download.openmmlab.com/mmrazor/v0.1/pruning/autoslim/autoslim_mbv2_subnet_8xb256_in1k/autoslim_mbv2_subnet_8xb256_in1k_flops-0.22M_acc-71.39_20211222-43117c7b_channel_cfg.yaml'  # noqa: E501
-]
-
-algorithm = dict(
-    architecture=dict(type='MMClsArchitecture', model=model),
-    distiller=None,
-    retraining=True,
-    bn_training_mode=False,
-    channel_cfg=channel_cfg)
-
-runner = dict(type='EpochBasedRunner', max_epochs=300)
-
-find_unused_parameters = True

configs/pruning/autoslim/autoslim_mbv2_supernet_8xb256_in1k.py

@@ -1,51 +0,0 @@
-_base_ = [
-    '../../_base_/datasets/mmcls/imagenet_bs256_autoslim.py',
-    '../../_base_/schedules/mmcls/imagenet_bs2048_autoslim.py',
-    '../../_base_/mmcls_runtime.py'
-]
-
-model = dict(
-    type='mmcls.ImageClassifier',
-    backbone=dict(type='MobileNetV2', widen_factor=1.5),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=1920,
-        loss=dict(
-            type='LabelSmoothLoss',
-            mode='original',
-            label_smooth_val=0.1,
-            loss_weight=1.0),
-        topk=(1, 5),
-    ))
-
-algorithm = dict(
-    type='AutoSlim',
-    architecture=dict(type='MMClsArchitecture', model=model),
-    distiller=dict(
-        type='SelfDistiller',
-        components=[
-            dict(
-                student_module='head.fc',
-                teacher_module='head.fc',
-                losses=[
-                    dict(
-                        type='KLDivergence',
-                        name='loss_kd',
-                        tau=1,
-                        loss_weight=1,
-                    )
-                ]),
-        ]),
-    pruner=dict(
-        type='RatioPruner',
-        ratios=(2 / 12, 3 / 12, 4 / 12, 5 / 12, 6 / 12, 7 / 12, 8 / 12, 9 / 12,
-                10 / 12, 11 / 12, 1.0)),
-    retraining=False,
-    bn_training_mode=True,
-    input_shape=None)
-
-runner = dict(type='EpochBasedRunner', max_epochs=50)
-
-use_ddp_wrapper = True

configs/pruning/autoslim/slimmable_network_retrain_mbv2_8xb256_in1k.py

@@ -1,216 +0,0 @@
-# defaults to use registries in mmcls
-default_scope = 'mmcls'
-
-# !architecture config
-# ==========================================================================
-architecture = dict(
-    _scope_='mmcls',
-    type='ImageClassifier',
-    backbone=dict(type='MobileNetV2', widen_factor=1.5),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=1920,
-        loss=dict(
-            type='LabelSmoothLoss',
-            mode='original',
-            label_smooth_val=0.1,
-            loss_weight=1.0),
-        topk=(1, 5),
-    ))
-# ==========================================================================
-
-# !dataset config
-# ==========================================================================
-# data preprocessor
-data_preprocessor = dict(
-    type='ImgDataPreprocessor',
-    # RGB format normalization parameters
-    mean=[123.675, 116.28, 103.53],
-    std=[58.395, 57.12, 57.375],
-    # convert image from BGR to RGB
-    bgr_to_rgb=True,
-)
-
-dataset_type = 'ImageNet'
-
-# ceph config
-use_ceph = True
-
-ceph_file_client_args = dict(
-    backend='petrel',
-    path_mapping=dict({
-        './data/imagenet':
-        'sproject:s3://openmmlab/datasets/classification/imagenet',
-        'data/imagenet':
-        'sproject:s3://openmmlab/datasets/classification/imagenet'
-    }))
-disk_file_client_args = dict(backend='disk')
-
-if use_ceph:
-    file_client_args = ceph_file_client_args
-else:
-    file_client_args = disk_file_client_args
-
-train_pipeline = [
-    dict(type='LoadImageFromFile', file_client_args=file_client_args),
-    dict(
-        type='RandomResizedCrop',
-        scale=224,
-        crop_ratio_range=(0.25, 1.0),
-        backend='pillow'),
-    dict(type='ColorJitter', brightness=0.4, contrast=0.4, saturation=0.4),
-    dict(type='RandomFlip', prob=0.5, direction='horizontal'),
-    dict(type='PackClsInputs'),
-]
-
-test_pipeline = [
-    dict(type='LoadImageFromFile', file_client_args=file_client_args),
-    dict(type='ResizeEdge', scale=256, edge='short', backend='pillow'),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='PackClsInputs'),
-]
-
-_batch_size_per_gpu = 256
-
-train_dataloader = dict(
-    batch_size=_batch_size_per_gpu,
-    num_workers=16,
-    dataset=dict(
-        type=dataset_type,
-        data_root='data/imagenet',
-        ann_file='meta/train.txt',
-        data_prefix='train',
-        pipeline=train_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=True),
-    persistent_workers=True,
-)
-
-val_dataloader = dict(
-    batch_size=_batch_size_per_gpu,
-    num_workers=16,
-    dataset=dict(
-        type=dataset_type,
-        data_root='data/imagenet',
-        ann_file='meta/val.txt',
-        data_prefix='val',
-        pipeline=test_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    persistent_workers=True,
-)
-val_evaluator = dict(type='Accuracy', topk=(1, 5))
-
-# If you want standard test, please manually configure the test dataset
-test_dataloader = val_dataloader
-test_evaluator = val_evaluator
-
-# !runtime config
-# ==========================================================================
-# configure log processor
-log_processor = dict(window_size=100)
-
-# configure default hooks
-default_hooks = dict(
-    # record the time of every iteration.
-    timer=dict(type='IterTimerHook'),
-
-    # print log every 100 iterations.
-    logger=dict(type='LoggerHook', interval=100),
-
-    # enable the parameter scheduler.
-    param_scheduler=dict(type='ParamSchedulerHook'),
-
-    # save checkpoint per epoch.
-    checkpoint=dict(
-        type='CheckpointHook', max_keep_ckpts=50, save_best='auto',
-        interval=1),
-
-    # set sampler seed in distributed evrionment.
-    sampler_seed=dict(type='DistSamplerSeedHook'),
-
-    # validation results visualization, set True to enable it.
-    visualization=dict(type='VisualizationHook', enable=False),
-)
-
-# configure environment
-env_cfg = dict(
-    # whether to enable cudnn benchmark
-    cudnn_benchmark=False,
-
-    # set multi process parameters
-    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
-
-    # set distributed parameters
-    dist_cfg=dict(backend='nccl'),
-)
-
-# set visualizer
-vis_backends = [dict(type='LocalVisBackend')]
-visualizer = dict(type='ClsVisualizer', vis_backends=vis_backends)
-
-# set log level
-log_level = 'INFO'
-
-# load from which checkpoint
-load_from = None
-
-# whether to resume training from the loaded checkpoint
-resume = False
-# ==========================================================================
-
-# !autoslim algorithm config
-# ==========================================================================
-channel_cfg_paths = [
-    'tests/data/MBV2_220M.yaml', 'tests/data/MBV2_320M.yaml',
-    'tests/data/MBV2_530M.yaml'
-]
-model = dict(
-    _scope_='mmrazor',
-    type='SlimmableNetwork',
-    architecture=architecture,
-    data_preprocessor=data_preprocessor,
-    channel_cfg_paths=channel_cfg_paths,
-    mutator=dict(
-        type='SlimmableChannelMutator',
-        mutable_cfg=dict(type='SlimmableMutableChannel'),
-        tracer_cfg=dict(
-            type='BackwardTracer',
-            loss_calculator=dict(type='ImageClassifierPseudoLoss'))))
-# ==========================================================================
-
-# !model wrapper config
-# ==========================================================================
-model_wrapper_cfg = dict(
-    type='mmrazor.SlimmableNetworkDDP',
-    broadcast_buffers=False,
-    find_unused_parameters=True)
-# ==========================================================================
-
-# !scheduler config
-# ==========================================================================
-paramwise_cfg = dict(
-    bias_decay_mult=0.0, norm_decay_mult=0.0, dwconv_decay_mult=0.0)
-optimizer = dict(
-    type='SGD', lr=0.5, momentum=0.9, nesterov=True, weight_decay=0.0001)
-optim_wrapper = dict(
-    optimizer=optimizer,
-    paramwise_cfg=paramwise_cfg,
-    accumulative_counts=len(channel_cfg_paths))
-
-# learning policy
-max_epochs = 300
-
-param_scheduler = dict(
-    type='PolyLR',
-    power=1.0,
-    eta_min=0.0,
-    by_epoch=True,
-    end=max_epochs,
-    convert_to_iter_based=True)
-
-# train, val, test setting
-train_cfg = dict(by_epoch=True, max_epochs=max_epochs, val_interval=1)
-val_cfg = dict(type='mmrazor.SlimmableValLoop')
-test_cfg = dict()
-# ==========================================================================

configs/pruning/autoslim/slimmable_network_supernet_mbv2_8xb256_in1k.py

@@ -1,228 +0,0 @@
-# defaults to use registries in mmcls
-default_scope = 'mmcls'
-
-# !architecture config
-# ==========================================================================
-architecture = dict(
-    _scope_='mmcls',
-    type='ImageClassifier',
-    backbone=dict(type='MobileNetV2', widen_factor=1.5),
-    neck=dict(type='GlobalAveragePooling'),
-    head=dict(
-        type='LinearClsHead',
-        num_classes=1000,
-        in_channels=1920,
-        loss=dict(
-            type='LabelSmoothLoss',
-            mode='original',
-            label_smooth_val=0.1,
-            loss_weight=1.0),
-        topk=(1, 5),
-    ))
-# ==========================================================================
-
-# !dataset config
-# ==========================================================================
-# data preprocessor
-data_preprocessor = dict(
-    type='ImgDataPreprocessor',
-    # RGB format normalization parameters
-    mean=[123.675, 116.28, 103.53],
-    std=[58.395, 57.12, 57.375],
-    # convert image from BGR to RGB
-    bgr_to_rgb=True,
-)
-
-dataset_type = 'ImageNet'
-
-# ceph config
-use_ceph = True
-
-ceph_file_client_args = dict(
-    backend='petrel',
-    path_mapping=dict({
-        './data/imagenet':
-        'sproject:s3://openmmlab/datasets/classification/imagenet',
-        'data/imagenet':
-        'sproject:s3://openmmlab/datasets/classification/imagenet'
-    }))
-disk_file_client_args = dict(backend='disk')
-
-if use_ceph:
-    file_client_args = ceph_file_client_args
-else:
-    file_client_args = disk_file_client_args
-
-train_pipeline = [
-    dict(type='LoadImageFromFile', file_client_args=file_client_args),
-    dict(
-        type='RandomResizedCrop',
-        scale=224,
-        crop_ratio_range=(0.25, 1.0),
-        backend='pillow'),
-    dict(type='ColorJitter', brightness=0.4, contrast=0.4, saturation=0.4),
-    dict(type='RandomFlip', prob=0.5, direction='horizontal'),
-    dict(type='PackClsInputs'),
-]
-
-test_pipeline = [
-    dict(type='LoadImageFromFile', file_client_args=file_client_args),
-    dict(type='ResizeEdge', scale=256, edge='short', backend='pillow'),
-    dict(type='CenterCrop', crop_size=224),
-    dict(type='PackClsInputs'),
-]
-
-_batch_size_per_gpu = 256
-
-train_dataloader = dict(
-    batch_size=_batch_size_per_gpu,
-    num_workers=16,
-    dataset=dict(
-        type=dataset_type,
-        data_root='data/imagenet',
-        ann_file='meta/train.txt',
-        data_prefix='train',
-        pipeline=train_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=True),
-    persistent_workers=True,
-)
-
-val_dataloader = dict(
-    batch_size=_batch_size_per_gpu,
-    num_workers=16,
-    dataset=dict(
-        type=dataset_type,
-        data_root='data/imagenet',
-        ann_file='meta/val.txt',
-        data_prefix='val',
-        pipeline=test_pipeline),
-    sampler=dict(type='DefaultSampler', shuffle=False),
-    persistent_workers=True,
-)
-val_evaluator = dict(type='Accuracy', topk=(1, 5))
-
-# If you want standard test, please manually configure the test dataset
-test_dataloader = val_dataloader
-test_evaluator = val_evaluator
-
-# !runtime config
-# ==========================================================================
-# configure log processor
-log_processor = dict(window_size=100)
-
-# configure default hooks
-default_hooks = dict(
-    # record the time of every iteration.
-    timer=dict(type='IterTimerHook'),
-
-    # print log every 100 iterations.
-    logger=dict(type='LoggerHook', interval=100),
-
-    # enable the parameter scheduler.
-    param_scheduler=dict(type='ParamSchedulerHook'),
-
-    # save checkpoint per epoch.
-    checkpoint=dict(
-        type='CheckpointHook', max_keep_ckpts=50, save_best='auto',
-        interval=1),
-
-    # set sampler seed in distributed evrionment.
-    sampler_seed=dict(type='DistSamplerSeedHook'),
-
-    # validation results visualization, set True to enable it.
-    visualization=dict(type='VisualizationHook', enable=False),
-)
-
-# configure environment
-env_cfg = dict(
-    # whether to enable cudnn benchmark
-    cudnn_benchmark=False,
-
-    # set multi process parameters
-    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
-
-    # set distributed parameters
-    dist_cfg=dict(backend='nccl'),
-)
-
-# set visualizer
-vis_backends = [dict(type='LocalVisBackend')]
-visualizer = dict(type='ClsVisualizer', vis_backends=vis_backends)
-
-# set log level
-log_level = 'INFO'
-
-# load from which checkpoint
-load_from = None
-
-# whether to resume training from the loaded checkpoint
-resume = False
-# ==========================================================================
-
-# !autoslim algorithm config
-# ==========================================================================
-num_samples = 2
-model = dict(
-    _scope_='mmrazor',
-    type='AutoSlim',
-    num_samples=num_samples,
-    architecture=architecture,
-    data_preprocessor=data_preprocessor,
-    distiller=dict(
-        type='ConfigurableDistiller',
-        teacher_recorders=dict(
-            fc=dict(type='ModuleOutputs', source='head.fc')),
-        student_recorders=dict(
-            fc=dict(type='ModuleOutputs', source='head.fc')),
-        distill_losses=dict(
-            loss_kl=dict(type='KLDivergence', tau=1, loss_weight=1)),
-        loss_forward_mappings=dict(
-            loss_kl=dict(
-                preds_S=dict(recorder='fc', from_student=True),
-                preds_T=dict(recorder='fc', from_student=False)))),
-    mutator=dict(
-        type='OneShotChannelMutator',
-        mutable_cfg=dict(
-            type='OneShotMutableChannel',
-            candidate_choices=list(i / 12 for i in range(2, 13)),
-            candidate_mode='ratio'),
-        tracer_cfg=dict(
-            type='BackwardTracer',
-            loss_calculator=dict(type='ImageClassifierPseudoLoss'))))
-# ==========================================================================
-
-# !model wrapper config
-# ==========================================================================
-model_wrapper_cfg = dict(
-    type='mmrazor.AutoSlimDDP',
-    broadcast_buffers=False,
-    find_unused_parameters=False)
-# ==========================================================================
-
-# !scheduler config
-# ==========================================================================
-paramwise_cfg = dict(
-    bias_decay_mult=0.0, norm_decay_mult=0.0, dwconv_decay_mult=0.0)
-optimizer = dict(
-    type='SGD', lr=0.5, momentum=0.9, nesterov=True, weight_decay=0.0001)
-optim_wrapper = dict(
-    optimizer=optimizer,
-    paramwise_cfg=paramwise_cfg,
-    accumulative_counts=num_samples + 2)
-
-# learning policy
-max_epochs = 50
-
-param_scheduler = dict(
-    type='PolyLR',
-    power=1.0,
-    eta_min=0.0,
-    by_epoch=True,
-    end=max_epochs,
-    convert_to_iter_based=True)
-
-# train, val, test setting
-train_cfg = dict(by_epoch=True, max_epochs=max_epochs, val_interval=1)
-val_cfg = dict(type='mmrazor.AutoSlimValLoop')
-test_cfg = dict()
-# ==========================================================================

configs/pruning/mmcls/autoslim/autoslim_mbv2_1.5x_slimmable_subnet_8xb256_in1k.py

@@ -0,0 +1,50 @@
+_base_ = [
+    'mmrazor::_base_/settings/imagenet_bs2048_autoslim.py',
+    'mmcls::_base_/models/mobilenet_v2_1x.py',
+    'mmcls::_base_/default_runtime.py',
+]
+
+supernet = _base_.model
+supernet.backbone.widen_factor = 1.5
+supernet.head.in_channels = 1920
+
+# !dataset config
+# ==========================================================================
+# data preprocessor
+data_preprocessor = dict(
+    type='ImgDataPreprocessor',
+    # RGB format normalization parameters
+    mean=[123.675, 116.28, 103.53],
+    std=[58.395, 57.12, 57.375],
+    # convert image from BGR to RGB
+    bgr_to_rgb=True)
+
+# !autoslim algorithm config
+# ==========================================================================
+channel_cfg_paths = [
+    'tests/data/MBV2_220M.yaml', 'tests/data/MBV2_320M.yaml',
+    'tests/data/MBV2_530M.yaml'
+]
+
+model = dict(
+    _delete_=True,
+    _scope_='mmrazor',
+    type='SlimmableNetwork',
+    architecture=supernet,
+    data_preprocessor=data_preprocessor,
+    channel_cfg_paths=channel_cfg_paths,
+    mutator=dict(
+        type='SlimmableChannelMutator',
+        mutable_cfg=dict(type='SlimmableMutableChannel'),
+        tracer_cfg=dict(
+            type='BackwardTracer',
+            loss_calculator=dict(type='ImageClassifierPseudoLoss'))))
+
+model_wrapper_cfg = dict(
+    type='mmrazor.SlimmableNetworkDDP',
+    broadcast_buffers=False,
+    find_unused_parameters=True)
+
+optim_wrapper = dict(accumulative_counts=3)
+
+val_cfg = dict(type='mmrazor.SlimmableValLoop')

configs/pruning/mmcls/autoslim/autoslim_mbv2_1.5x_subnet_8xb256_in1k.py

@@ -1,3 +1,3 @@
-_base_ = 'autoslim_slimmable.py'
+_base_ = 'autoslim_mbv2_1.5x_supernet_8xb256_in1k.py'
 
 model = dict(channel_cfg_paths='tests/data/MBV2_530M.yaml')

configs/pruning/mmcls/autoslim/autoslim_mbv2_1.5x_supernet_8xb256_in1k.py

@@ -0,0 +1,66 @@
+_base_ = [
+    'mmrazor::_base_/settings/imagenet_bs2048_autoslim.py',
+    'mmcls::_base_/models/mobilenet_v2_1x.py',
+    'mmcls::_base_/default_runtime.py',
+]
+
+supernet = _base_.model
+supernet.backbone.widen_factor = 1.5
+supernet.head.in_channels = 1920
+
+# !dataset config
+# ==========================================================================
+# data preprocessor
+data_preprocessor = dict(
+    type='ImgDataPreprocessor',
+    # RGB format normalization parameters
+    mean=[123.675, 116.28, 103.53],
+    std=[58.395, 57.12, 57.375],
+    # convert image from BGR to RGB
+    bgr_to_rgb=True,
+)
+
+# !autoslim algorithm config
+num_samples = 2
+model = dict(
+    _scope_='mmrazor',
+    type='AutoSlim',
+    num_samples=num_samples,
+    architecture=supernet,
+    data_preprocessor=data_preprocessor,
+    distiller=dict(
+        type='ConfigurableDistiller',
+        teacher_recorders=dict(
+            fc=dict(type='ModuleOutputs', source='head.fc')),
+        student_recorders=dict(
+            fc=dict(type='ModuleOutputs', source='head.fc')),
+        distill_losses=dict(
+            loss_kl=dict(type='KLDivergence', tau=1, loss_weight=1)),
+        loss_forward_mappings=dict(
+            loss_kl=dict(
+                preds_S=dict(recorder='fc', from_student=True),
+                preds_T=dict(recorder='fc', from_student=False)))),
+    mutator=dict(
+        type='OneShotChannelMutator',
+        mutable_cfg=dict(
+            type='OneShotMutableChannel',
+            candidate_choices=list(i / 12 for i in range(2, 13)),
+            candidate_mode='ratio'),
+        tracer_cfg=dict(
+            type='BackwardTracer',
+            loss_calculator=dict(type='ImageClassifierPseudoLoss'))))
+
+model_wrapper_cfg = dict(
+    type='mmrazor.AutoSlimDDP',
+    broadcast_buffers=False,
+    find_unused_parameters=False)
+
+optim_wrapper = dict(accumulative_counts=num_samples + 2)
+
+# learning policy
+max_epochs = 50
+param_scheduler = dict(end=max_epochs)
+
+# train, val, test setting
+train_cfg = dict(max_epochs=max_epochs)
+val_cfg = dict(type='mmrazor.AutoSlimValLoop')

model-index.yml

@@ -1,8 +1,9 @@
 Import:
-  - configs/distill/cwd/metafile.yml
-  - configs/distill/wsld/metafile.yml
-  - configs/distill/rkd/metafile.yml
-  - configs/nas/darts/metafile.yml
-  - configs/nas/detnas/metafile.yml
-  - configs/nas/spos/metafile.yml
-  - configs/pruning/autoslim/metafile.yml
+  - configs/distill/mmseg/cwd/metafile.yml
+  - configs/distill/mmdet/cwd/metafile.yml
+  - configs/distill/mmcls/wsld/metafile.yml
+  - configs/distill/mmcls/rkd/metafile.yml
+  # - configs/nas/darts/metafile.yml
+  - configs/nas/mmdet/detnas/metafile.yml
+  - configs/nas/mmcls/spos/metafile.yml
+  # - configs/pruning/autoslim/metafile.yml

tests/test_metafiles.py

@@ -0,0 +1,53 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import os
+from pathlib import Path
+
+import requests
+import yaml
+
+MMRAZOR_ROOT = Path(__file__).absolute().parents[1]
+
+
+class TestMetafiles:
+
+    def get_metafiles(self, code_path):
+        """
+        Function: get the metafile of all configs from model-index.yml
+        """
+        metafile = os.path.join(code_path, 'model-index.yml')
+        with open(metafile, 'r') as f:
+            meta = yaml.safe_load(f)
+        return meta['Import']
+
+    def test_metafiles(self):
+        metafiles = self.get_metafiles(MMRAZOR_ROOT)
+        for mf in metafiles:
+            metafile = os.path.abspath(os.path.join(MMRAZOR_ROOT, mf))
+            with open(metafile, 'r') as f:
+                meta = yaml.safe_load(f)
+            for model in meta['Models']:
+                # 1. weights url check
+                r = requests.head(model['Weights'], timeout=4)
+                assert r.status_code != 404, \
+                    f"can't connect url {model['Weights']} in " \
+                    f'metafile {metafile}'
+
+                # 2. config check
+                dir_path = os.path.abspath(os.path.join(metafile, '../'))
+                # list all files which are in the same directory of
+                # current metafile
+                config_files = os.listdir(dir_path)
+
+                if isinstance(model['Config'], list):
+                    # TODO: 3. log error
+                    continue
+
+                assert (model['Config'].split('/')[-1] in config_files), \
+                    f"config error in {metafile} model {model['Name']}"
+
+                # 4. name check
+                # erase '.py'
+                correct_name = model['Config'].split('/')[-1][:-3]
+                assert model['Name'] == correct_name, \
+                    f'name error in {metafile}, correct name should ' \
+                    f'be {correct_name}'

tools/model_converters/convert_kd_ckpt.py

@@ -0,0 +1,47 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import argparse
+from pathlib import Path
+
+import torch
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(
+        description='Process a checkpoint to be published')
+    parser.add_argument('checkpoint', help='input checkpoint filename')
+    parser.add_argument(
+        '--inplace', action='store_true', help='replace origin ckpt')
+    args = parser.parse_args()
+    return args
+
+
+def main():
+    args = parse_args()
+    checkpoint = torch.load(args.checkpoint, map_location='cpu')
+    new_state_dict = dict()
+
+    for key, value in checkpoint['state_dict'].items():
+        if key.startswith('architecture.model.distiller.teacher'):
+            new_key = key.replace('architecture.model.distiller.teacher',
+                                  'architecture.teacher')
+        elif key.startswith('architecture.model'):
+            new_key = key.replace('architecture.model', 'architecture')
+        else:
+            new_key = key
+
+        new_state_dict[new_key] = value
+
+    checkpoint['state_dict'] = new_state_dict
+
+    if args.inplace:
+        torch.save(checkpoint, args.checkpoint)
+    else:
+        ckpt_path = Path(args.checkpoint)
+        ckpt_name = ckpt_path.stem
+        ckpt_dir = ckpt_path.parent
+        new_ckpt_path = ckpt_dir / f'{ckpt_name}_latest.pth'
+        torch.save(checkpoint, new_ckpt_path)
+
+
+if __name__ == '__main__':
+    main()

tools/train.py

@@ -45,7 +45,8 @@ def main():
     cfg.launcher = args.launcher
     if args.cfg_options is not None:
         cfg.merge_from_dict(args.cfg_options)
-
+    import pdb
+    pdb.set_trace()
     # work_dir is determined in this priority: CLI > segment in file > filename
     if args.work_dir is not None:
         # update configs according to CLI args if args.work_dir is not None