Ross Wightman 3b9032ea48 Use Tensor.unfold().unfold() for HaloAttn, fast like as_strided but more clarity

2021-08-27 12:45:53 -07:00

..

__init__.py

Update attention / self-attn based models from a series of experiments:

2021-08-20 16:13:11 -07:00

activations_jit.py

…

activations_me.py

…

activations.py

…

adaptive_avgmax_pool.py

Refactoring, cleanup, improved test coverage.

2021-06-12 16:40:02 -07:00

blur_pool.py

…

bottleneck_attn.py

Improved (hopefully) init for SA/SA-like layers used in ByoaNets

2021-05-04 21:40:39 -07:00

cbam.py

Add Gather-Excite and Global Context attn modules. Refactor existing SE-like attn for consistency and refactor byob/byoanet for less redundancy.

2021-05-27 18:03:29 -07:00

classifier.py

Refactoring, cleanup, improved test coverage.

2021-06-12 16:40:02 -07:00

cond_conv2d.py

…

config.py

…

conv2d_same.py

…

conv_bn_act.py

…

create_act.py

Fix #661, move hardswish out of default args for LeViT. Enable native torch support for hardswish, hardsigmoid, mish if present.

2021-05-26 15:28:42 -07:00

create_attn.py

Update attention / self-attn based models from a series of experiments:

2021-08-20 16:13:11 -07:00

create_conv2d.py

…

create_norm_act.py

…

drop.py

…

eca.py

Add non-local and BAT attention. Merge attn and self-attn factories into one. Add attention references to README. Add mlp 'mode' to ECA.

2021-05-31 13:18:11 -07:00

evo_norm.py

…

gather_excite.py

Add Gather-Excite and Global Context attn modules. Refactor existing SE-like attn for consistency and refactor byob/byoanet for less redundancy.

2021-05-27 18:03:29 -07:00

global_context.py

Add gc_efficientnetv2_rw_t weights (global context instead of SE attn). Add TF XL weights even though the fine-tuned ones don't validate that well. Change default arg for GlobalContext to use scal (mul) mode.

2021-08-07 16:45:29 -07:00

halo_attn.py

Use Tensor.unfold().unfold() for HaloAttn, fast like as_strided but more clarity

2021-08-27 12:45:53 -07:00

helpers.py

Throw in some FBNetV3 code I had lying around, some refactoring of SE reduction channel calcs for all EffNet archs.

2021-05-28 20:47:24 -07:00

inplace_abn.py

…

lambda_layer.py

Improved (hopefully) init for SA/SA-like layers used in ByoaNets

2021-05-04 21:40:39 -07:00

linear.py

…

median_pool.py

…

mixed_conv2d.py

…

mlp.py

Refactoring, cleanup, improved test coverage.

2021-06-12 16:40:02 -07:00

non_local_attn.py

Fix torchscript issue in bat

2021-06-08 23:19:51 -07:00

norm_act.py

…

norm.py

Realized LayerNorm2d won't work in all cases as is, fixed.

2021-07-05 18:21:34 -07:00

padding.py

…

patch_embed.py

Add levit, levit_c, and visformer model defs. Largely untested and not finished cleanup.

2021-05-14 17:16:34 -07:00

pool2d_same.py

Support npz custom load for vision transformer hybrid models. Add posembed rescale for npz load.

2021-06-14 12:31:44 -07:00

selective_kernel.py

Remove min channels for SelectiveKernel, divisor should cover cases well enough.

2021-05-31 15:38:56 -07:00

separable_conv.py

…

space_to_depth.py

…

split_attn.py

Add non-local and BAT attention. Merge attn and self-attn factories into one. Add attention references to README. Add mlp 'mode' to ECA.

2021-05-31 13:18:11 -07:00

split_batchnorm.py

…

squeeze_excite.py

Add non-local and BAT attention. Merge attn and self-attn factories into one. Add attention references to README. Add mlp 'mode' to ECA.

2021-05-31 13:18:11 -07:00

std_conv.py

Add file docstring to std_conv.py

2021-06-13 12:31:06 -07:00

test_time_pool.py

…

weight_init.py

…