TFLite `--int8` 'flatbuffers==1.12' fix (#6216)
* TFLite `--int8` 'flatbuffers==1.12' fix Temporary workaround for TFLite INT8 export. * Update export.py * Update export.pypull/6217/head
parent
5402753a53
commit
8125ec5d42
|
@ -277,8 +277,6 @@ def export_tflite(keras_model, im, file, int8, data, ncalib, prefix=colorstr('Te
|
|||
try:
|
||||
import tensorflow as tf
|
||||
|
||||
from models.tf import representative_dataset_gen
|
||||
|
||||
LOGGER.info(f'\n{prefix} starting export with tensorflow {tf.__version__}...')
|
||||
batch_size, ch, *imgsz = list(im.shape) # BCHW
|
||||
f = str(file).replace('.pt', '-fp16.tflite')
|
||||
|
@ -288,6 +286,8 @@ def export_tflite(keras_model, im, file, int8, data, ncalib, prefix=colorstr('Te
|
|||
converter.target_spec.supported_types = [tf.float16]
|
||||
converter.optimizations = [tf.lite.Optimize.DEFAULT]
|
||||
if int8:
|
||||
from models.tf import representative_dataset_gen
|
||||
check_requirements(('flatbuffers==1.12',)) # https://github.com/ultralytics/yolov5/issues/5707
|
||||
dataset = LoadImages(check_dataset(data)['train'], img_size=imgsz, auto=False) # representative data
|
||||
converter.representative_dataset = lambda: representative_dataset_gen(dataset, ncalib)
|
||||
converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]
|
||||
|
|
Loading…
Reference in New Issue