model-optimization

model-optimization copied to clipboard

Hi team `tensorflow-model-optimization`, I was applying clustering and pruning to a model based on Bert encoder from `tensorlfow-models-official`, and noticed that there is no API for registering custom layers, so...

aaarrti

Does Post-training full integer quantization support BERT?

1

Does Post-training full integer quantization in https://www.tensorflow.org/lite/performance/post_training_integer_quant#convert_using_float_fallback_quantization support BERT? I convert my pb model to tf lite: ``` dataset = create_dataset() def representative_dataset(): for data in dataset: yield { "token_type_ids":...

MrRace

I've read through the official guide and ran into problems understanding some concepts: 1. Is it possible to use Quantization Aware Training and not convert the model to a TF...

rthenamvar

**Describe the bug** Stripping the pruning layers seems to somehow disconnect the input layer from the graph. **System information** TensorFlow version (installed from source or binary): 2.11 (macos) TensorFlow Model...

christian-steinmeyer

CQAT fails to preserve clusters on ResNet-50

1

Prior to filing: check that this should be a bug instead of a feature request. Everything supported, including the compatible versions of TensorFlow, is listed in the overview page of...

funkyyyyyy

The input order of concatenate layer is changed, when use quantize_model interface.

1

use quantize_model interface:  original convert interface: 

fhahaha

How to get quantized weights from QAT model?

11

Hi all. I've recently trained a keras implementation of ssd-keras. I've managed to run QAT training on the model and got desired the accuracy. I wanted to get the quantised...

Hackerman28

guynich

sparsity.prune_low_magnitude fails with mixed precision policy mixed_float16

6

**Describe the bug** When using `tf.keras.mixed_precision.experimental.Policy("mixed_float16", loss_scale="dynamic")` the `sparsity.prune_low_magnitude` fails in tensor type conversion with the error `Tensor conversion requested dtype float32 for Tensor with dtype float16: `. Things work...

dsuthar-nvidia

Fix broken link in comprehensive_guide.ipynb

1

Remove extra parentheses in line 726 to fix the broken link.

synandi