tensorflow/model-optimization issues and pull requests

#1142 - Call `super().build(input_shape)` instead of `self.built = True` in all Keras layers.

Pull Request - State: closed - Opened by copybara-service[bot] 10 days ago
Labels: technique:pruning, technique:clustering, comp:weight compression

#1141 - ValueError: `prune_low_magnitude` can only prune an object of the following types: keras.models.Sequential, keras functional model, keras.layers.Layer, list of keras.layers.Layer. You passed an object of type: Sequential.

Issue - State: closed - Opened by FrancescoSorrentino about 2 months ago - 1 comment
Labels: bug

#1140 - Failed to apply the QAT function 'quantize_model' to the sequential model that is defined using tensorflow.keras

Issue - State: open - Opened by wwwind 2 months ago - 2 comments
Labels: bug

#1139 - RuntimeError: Layer tf.operators.getitem:<class 'tf_keras.src.layers.core.tf_op_layer.SlicingOpLambda'> is not supported.

Issue - State: open - Opened by reganh98 3 months ago
Labels: feature request

#1136 - Fix `dtype` and `assign*` in `AutocastVariable`.

Pull Request - State: closed - Opened by copybara-service[bot] 3 months ago
Labels: technique:pruning

#1135 - Unexpected Inference Time and Model Size for TensorFlow Lite and Pruned Models

Issue - State: open - Opened by experimentsym3 3 months ago
Labels: bug

#1134 - Is it possible to use this tool to optimize (quantize) a model trained with PyTorch?

Issue - State: open - Opened by zhuoran-guo 3 months ago
Labels: feature request

#1133 - Does structural pruning support pre-trained model?

Issue - State: open - Opened by aidevmin 4 months ago
Labels: feature request

#1132 - DepthwiseConv2D Layers cannot be clustered or sparsely pruned

Issue - State: open - Opened by nathanwbailey 4 months ago
Labels: bug

#1131 - Per-tensor QAT model Conv2d+BN+relu folding issue

Issue - State: open - Opened by sheh 5 months ago
Labels: bug

#1130 - Failed quantization of dilated convolution layers: tensorflow or tensorflow-model-optimization bug?

Issue - State: open - Opened by Ebanflo42 5 months ago
Labels: bug

#1129 - Permit tensorflow_model_optimization to use absl-py 2

Pull Request - State: closed - Opened by copybara-service[bot] 5 months ago

#1128 - Add `--test_output=errors` for kokoro tests of `tensorflow_model_optimization` for better debuggability.

Pull Request - State: closed - Opened by copybara-service[bot] 5 months ago

#1127 - Add required packages to setup.py to match requirements.txt

Pull Request - State: closed - Opened by copybara-service[bot] 6 months ago

#1126 - re-enable g3folder_to_gitfolder

Pull Request - State: closed - Opened by copybara-service[bot] 6 months ago
Labels: comp:weight compression

#1125 - Disable the reversible_check for g3folder_to_gitfolder workflow of model-optimization repo, since we are not receiving contributions from github.

Pull Request - State: closed - Opened by copybara-service[bot] 6 months ago
Labels: comp:weight compression

#1124 - Custom layer with Concat afterwards causes an error during QAT modeling

Issue - State: open - Opened by mhyeonsoo 6 months ago - 3 comments
Labels: bug

#1123 - Fix model loading from .keras files in Tensorflow 2.15

Pull Request - State: open - Opened by MaierOli2010 7 months ago - 1 comment
Labels: technique:qat

#1122 - Cannot use Quantize layer and use abstract class and methods

Issue - State: open - Opened by dhruven-god 7 months ago - 1 comment
Labels: bug

#1121 - Internal change.

Pull Request - State: closed - Opened by copybara-service[bot] 7 months ago
Labels: technique:qat

#1120 - Remove `[email protected]` from `SECURITY.md`

Pull Request - State: closed - Opened by copybara-service[bot] 7 months ago

#1119 - Any plans to support keras3?

Issue - State: open - Opened by Krovatkin 7 months ago - 4 comments
Labels: feature request

#1118 - Updates security files

Pull Request - State: open - Opened by copybara-service[bot] 8 months ago

#1117 - Can't use TFMOT version 0.8.0 due to missing dependency

Issue - State: open - Opened by vloncar 8 months ago - 1 comment
Labels: bug

#1116 - Fix keras importing failures at the colab.

Pull Request - State: closed - Opened by copybara-service[bot] 8 months ago - 1 comment
Labels: technique:pruning, technique:qat, technique:clustering

#1115 - Update release notes and version for TFMOT release 0.8.0.

Pull Request - State: closed - Opened by copybara-service[bot] 8 months ago

#1114 - Support for Recurrent layers for Quantization Aware Training.

Issue - State: open - Opened by parth-desai 8 months ago - 1 comment
Labels: feature request

#1113 - Force to use the keras v2 version to resolve breakages in OSS

Pull Request - State: closed - Opened by copybara-service[bot] 8 months ago
Labels: technique:pruning, technique:qat

#1112 - Update all `keras` references to use the keras v2 version instance

Pull Request - State: closed - Opened by copybara-service[bot] 8 months ago
Labels: technique:pruning, technique:qat, technique:clustering, comp:weight compression

#1111 - Require `tf-keras` for future keras compatibility

Pull Request - State: closed - Opened by copybara-service[bot] 8 months ago

#1110 - Update documentation in quantize.py

Pull Request - State: open - Opened by TheCrazyT 8 months ago - 1 comment
Labels: technique:qat

#1109 - strange behavior when quantizing a model.

Issue - State: open - Opened by IdrissARM 8 months ago - 2 comments
Labels: bug

#1108 - Error in MovingAverageQuantizer with per_axis=True due to missing parameters in _add_range_weights

Issue - State: open - Opened by Litschi123 8 months ago
Labels: bug

#1107 - MobileNetV3 QAT TFLite Conversion Issue

Issue - State: open - Opened by tarushbansal 9 months ago - 7 comments
Labels: bug

#1106 - Fix structural pruning sparsity notebook

Pull Request - State: closed - Opened by copybara-service[bot] 9 months ago - 1 comment
Labels: technique:pruning

#1105 - A error about quantization aware training

Issue - State: closed - Opened by panhu 9 months ago - 2 comments
Labels: bug

#1105 - A error about quantization aware training

Issue - State: open - Opened by panhu 9 months ago
Labels: bug

#1104 - Module Import Error

Issue - State: closed - Opened by surajpandey353 10 months ago - 2 comments

#1103 - Internal Code Change

Pull Request - State: closed - Opened by copybara-service[bot] 11 months ago
Labels: technique:pruning, technique:qat, technique:clustering

#1102 - Internal Code Change

Pull Request - State: closed - Opened by copybara-service[bot] 11 months ago
Labels: technique:pruning, technique:qat, technique:clustering, comp:weight compression

#1101 - [COLAB] No module named 'tensorflow_model_optimization'

Issue - State: closed - Opened by rjtp5670 11 months ago - 1 comment

#1100 - Quant aware training in tensorflow model optimization

Issue - State: closed - Opened by ardeal 12 months ago - 4 comments
Labels: help wanted

#1099 - Add batch norm to default_n_bit_quantize_registry and default_8_bit_quantize_registry

Issue - State: open - Opened by DerryFitz about 1 year ago - 2 comments
Labels: feature request

#1098 - Add a default PruningPolicy that filters out any layers not supported by the API

Issue - State: open - Opened by annietllnd about 1 year ago - 1 comment
Labels: feature request

#1097 - Rewire TensorFlow to rely on tf_keras target.

Pull Request - State: open - Opened by copybara-service[bot] about 1 year ago
Labels: technique:pruning

#1096 - Remove private Keras imports.

Pull Request - State: closed - Opened by copybara-service[bot] about 1 year ago
Labels: technique:pruning

#1095 - Remove invalid dependency on private py/keras package.

Pull Request - State: closed - Opened by copybara-service[bot] about 1 year ago
Labels: technique:pruning

#1094 - Remove private Keras imports.

Pull Request - State: closed - Opened by copybara-service[bot] about 1 year ago
Labels: technique:qat

#1093 - Remove private Keras imports.

Pull Request - State: closed - Opened by copybara-service[bot] about 1 year ago
Labels: technique:qat

#1093 - Remove private Keras imports.

Pull Request - State: open - Opened by copybara-service[bot] about 1 year ago
Labels: technique:qat

#1092 - float16 quantization runs out of memory for LSTM model

Issue - State: closed - Opened by Black3rror about 1 year ago - 1 comment
Labels: bug

#1092 - float16 quantization runs out of memory for LSTM model

Issue - State: closed - Opened by Black3rror about 1 year ago - 1 comment
Labels: bug

#1091 - float16 quantization runs out of memory for LSTM model

Issue - State: open - Opened by Black3rror about 1 year ago - 3 comments
Labels: bug

#1091 - float16 quantization runs out of memory for LSTM model

Issue - State: open - Opened by Black3rror about 1 year ago - 1 comment
Labels: bug

#1090 - 16x8 Quantization fails for RNN model - Max and min for dynamic tensors should be recorded during calibration

Issue - State: open - Opened by Black3rror about 1 year ago - 4 comments
Labels: bug

#1090 - 16x8 Quantization fails for RNN model - Max and min for dynamic tensors should be recorded during calibration

Issue - State: open - Opened by Black3rror about 1 year ago - 4 comments
Labels: bug

#1089 - batch norm layer quantization error

Issue - State: closed - Opened by mhyeonsoo about 1 year ago - 2 comments
Labels: bug

#1088 - remove early out in xnnpack pruning policy model traversal

Pull Request - State: open - Opened by christian-steinmeyer about 1 year ago
Labels: technique:pruning

#1087 - Determinism is not yet supported in GPU implementation of FakeQuantWithMinMaxVarsGradient

Issue - State: closed - Opened by puelon about 1 year ago - 2 comments

#1086 - QAT aware training for mobilenetV2 not working

Issue - State: open - Opened by christophezeinaty about 1 year ago - 1 comment

#1085 - AttributeError: 'CustomLayerMaxPooling1D' object has no attribute 'kernel'

Issue - State: open - Opened by konstantinatopali about 1 year ago - 1 comment
Labels: feature request

#1084 - Custom_Layer Quantization with custom training (QAT)

Issue - State: open - Opened by ManishwarG about 1 year ago - 2 comments
Labels: bug

#1083 - Internal Code Change

Pull Request - State: closed - Opened by copybara-service[bot] about 1 year ago
Labels: technique:pruning, technique:qat, technique:clustering

#1082 - [NumPy] Fix test failures under NumPy 1.25.

Pull Request - State: closed - Opened by copybara-service[bot] about 1 year ago

#1081 - Increase the tolerance to remove flakiness

Pull Request - State: closed - Opened by copybara-service[bot] about 1 year ago

#1080 - Fix broken TFMO tests after Sequential serialization fix.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: technique:qat

#1079 - Cannot `pip install` nightly

Issue - State: open - Opened by christian-steinmeyer over 1 year ago - 2 comments
Labels: bug

#1078 - add original weight name as prefix for mask and threshold weights

Pull Request - State: open - Opened by christian-steinmeyer over 1 year ago - 1 comment
Labels: technique:pruning

#1077 - Cannot save pruned model with MultiHeadAttention Layer

Issue - State: open - Opened by christian-steinmeyer over 1 year ago - 2 comments
Labels: bug

#1076 - Extract _clone_model_with_weights from quantize_apply

Pull Request - State: open - Opened by tbekolay over 1 year ago
Labels: technique:qat

#1075 - Internal Code Change

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: technique:qat, technique:clustering

#1074 - Block epr_test until deps issue is resolved.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: comp:weight compression

#1073 - Internal test

Pull Request - State: open - Opened by copybara-service[bot] over 1 year ago

#1072 - Update release notes and version for TFMOT release 0.7.5.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: comp:weight compression

#1071 - Fix compatibility issues for the TF/Keras 2.13.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: technique:pruning, technique:qat

#1070 - allow registering custom layers in prune and clustering registries.

Pull Request - State: open - Opened by aaarrti over 1 year ago
Labels: technique:pruning, technique:clustering, api-review

#1069 - allow registering custom layers

Pull Request - State: closed - Opened by aaarrti over 1 year ago - 1 comment
Labels: technique:pruning, technique:clustering, api-review

#1068 - Quantization introduces inconsistencies during distributed training

Issue - State: closed - Opened by jiannanWang over 1 year ago - 1 comment
Labels: bug

#1067 - trainable=False doesn't work for QuantizeWrapperV2

Issue - State: open - Opened by jiannanWang over 1 year ago - 1 comment
Labels: bug

#1066 - Does Post-training full integer quantization support BERT?

Issue - State: open - Opened by MrRace over 1 year ago - 1 comment
Labels: bug

#1065 - Silence some pytype errors.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: comp:weight compression

#1064 - some questions about quantization in TensorFlow

Issue - State: open - Opened by rthenamvar over 1 year ago

#1063 - Stripping disconnects input layer from graph

Issue - State: open - Opened by christian-steinmeyer over 1 year ago
Labels: bug

#1062 - `tf.split` or `tf.transpose` cause errors for quantize-aware training with `quantize_apply`

Issue - State: open - Opened by Janus-Shiau over 1 year ago - 4 comments
Labels: bug

#1061 - The input order of concatenate layer is changed, when use quantize_model interface.

Issue - State: open - Opened by fhahaha over 1 year ago - 1 comment
Labels: bug

#1060 - Handle the error type reflect to the keras update.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: technique:qat

#1059 - Some error checking logics are updated due to cl/524418506.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: technique:qat

#1058 - Change the term 'blacklist' -> 'denylist'.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago

#1057 - Create a local copy of `keras.utils.generic_utils.to_snake_case` at `pruning_wrapper.py`.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: technique:pruning

#1056 - CQAT fails to preserve clusters on ResNet-50

Issue - State: open - Opened by funkyyyyyy over 1 year ago - 1 comment
Labels: bug

#1055 - Split large target under python/training/ and redirect references to the small targets.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: technique:qat

#1054 - [notebook] Use stable tensorflow release rather than nightly version.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago - 1 comment
Labels: technique:qat

#1053 - tfl.concatenation op quantization parameters violate the same scale constraint

Issue - State: closed - Opened by akrapukhin over 1 year ago - 2 comments
Labels: bug

#1052 - Update release notes and version for TFMOT release 0.7.4.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago

#1051 - Cannot prune models with tensorflow operations

Issue - State: closed - Opened by useruser2023 over 1 year ago - 6 comments
Labels: bug

#1050 - Creates non-breaking changes where necessary in preparation for switching all of Keras to new serialization format.

Pull Request - State: closed - Opened by copybara-service[bot] over 1 year ago
Labels: technique:qat

GitHub / tensorflow/model-optimization issues and pull requests