mct-nightly 1.8.0.4032023.post406__py3-none-any.whl → 1.8.0.4042023.post409__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (64) hide show
  1. {mct_nightly-1.8.0.4032023.post406.dist-info → mct_nightly-1.8.0.4042023.post409.dist-info}/METADATA +7 -7
  2. {mct_nightly-1.8.0.4032023.post406.dist-info → mct_nightly-1.8.0.4042023.post409.dist-info}/RECORD +63 -59
  3. {mct_nightly-1.8.0.4032023.post406.dist-info → mct_nightly-1.8.0.4042023.post409.dist-info}/WHEEL +1 -1
  4. model_compression_toolkit/__init__.py +9 -15
  5. model_compression_toolkit/core/common/logger.py +10 -2
  6. model_compression_toolkit/core/keras/back2framework/model_gradients.py +3 -2
  7. model_compression_toolkit/core/keras/quantization_facade.py +1 -1
  8. model_compression_toolkit/core/pytorch/back2framework/model_gradients.py +13 -6
  9. model_compression_toolkit/core/pytorch/constants.py +4 -0
  10. model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/reshape_with_static_shapes.py +16 -2
  11. model_compression_toolkit/exporter/__init__.py +5 -0
  12. model_compression_toolkit/exporter/model_exporter/__init__.py +0 -3
  13. model_compression_toolkit/exporter/model_exporter/tflite/fakely_quant_tflite_exporter.py +1 -1
  14. model_compression_toolkit/exporter/model_wrapper/__init__.py +4 -8
  15. model_compression_toolkit/exporter/model_wrapper/keras/builder/fully_quantized_model_builder.py +45 -39
  16. model_compression_toolkit/exporter/model_wrapper/keras/builder/node_to_quantizer.py +39 -24
  17. model_compression_toolkit/exporter/model_wrapper/keras/validate_layer.py +50 -42
  18. model_compression_toolkit/exporter/model_wrapper/pytorch/builder/fully_quantized_model_builder.py +43 -36
  19. model_compression_toolkit/exporter/model_wrapper/pytorch/builder/node_to_quantizer.py +24 -5
  20. model_compression_toolkit/exporter/model_wrapper/pytorch/validate_layer.py +25 -18
  21. model_compression_toolkit/gptq/__init__.py +6 -0
  22. model_compression_toolkit/gptq/common/gptq_config.py +57 -104
  23. model_compression_toolkit/gptq/common/gptq_constants.py +0 -7
  24. model_compression_toolkit/gptq/common/gptq_training.py +28 -38
  25. model_compression_toolkit/gptq/keras/gptq_training.py +10 -28
  26. model_compression_toolkit/gptq/keras/graph_info.py +8 -33
  27. model_compression_toolkit/gptq/keras/quantization_facade.py +6 -12
  28. model_compression_toolkit/gptq/keras/quantizer/base_keras_gptq_quantizer.py +0 -1
  29. model_compression_toolkit/gptq/keras/quantizer/quantization_builder.py +2 -2
  30. model_compression_toolkit/gptq/keras/quantizer/regularization_factory.py +45 -0
  31. model_compression_toolkit/gptq/keras/quantizer/soft_rounding/soft_quantizer_reg.py +112 -0
  32. model_compression_toolkit/gptq/keras/quantizer/soft_rounding/symmetric_soft_quantizer.py +38 -135
  33. model_compression_toolkit/gptq/keras/quantizer/ste_rounding/symmetric_ste.py +11 -41
  34. model_compression_toolkit/gptq/pytorch/gptq_training.py +9 -24
  35. model_compression_toolkit/gptq/pytorch/graph_info.py +7 -27
  36. model_compression_toolkit/gptq/pytorch/quantization_facade.py +9 -22
  37. model_compression_toolkit/gptq/pytorch/quantizer/__init__.py +1 -0
  38. model_compression_toolkit/gptq/pytorch/quantizer/base_pytorch_gptq_quantizer.py +0 -20
  39. model_compression_toolkit/gptq/pytorch/quantizer/quant_utils.py +10 -1
  40. model_compression_toolkit/gptq/pytorch/quantizer/quantization_builder.py +2 -2
  41. model_compression_toolkit/gptq/pytorch/quantizer/regularization_factory.py +45 -0
  42. model_compression_toolkit/gptq/pytorch/quantizer/soft_rounding/soft_quantizer_reg.py +115 -0
  43. model_compression_toolkit/gptq/pytorch/quantizer/soft_rounding/symmetric_soft_quantizer.py +30 -117
  44. model_compression_toolkit/gptq/pytorch/quantizer/soft_rounding/uniform_soft_quantizer.py +196 -0
  45. model_compression_toolkit/gptq/pytorch/quantizer/ste_rounding/symmetric_ste.py +9 -31
  46. model_compression_toolkit/qat/keras/quantizer/ste_rounding/symmetric_ste.py +30 -37
  47. model_compression_toolkit/qat/keras/quantizer/ste_rounding/uniform_ste.py +27 -36
  48. model_compression_toolkit/qat/pytorch/quantizer/ste_rounding/symmetric_ste.py +21 -21
  49. model_compression_toolkit/qat/pytorch/quantizer/ste_rounding/uniform_ste.py +25 -26
  50. model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/get_all_subclasses.py +1 -2
  51. model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/get_quantizers.py +1 -1
  52. model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/__init__.py +4 -0
  53. model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/constants.py +1 -0
  54. model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantize_wrapper.py +13 -3
  55. model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/__init__.py +6 -0
  56. model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/constants.py +3 -0
  57. model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/base_trainable_quantizer.py +53 -2
  58. model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/get_quantizers.py +2 -1
  59. model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/keras/base_keras_quantizer.py +22 -4
  60. model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/pytorch/base_pytorch_quantizer.py +24 -3
  61. model_compression_toolkit/gptq/common/gptq_quantizer_config.py +0 -93
  62. {mct_nightly-1.8.0.4032023.post406.dist-info → mct_nightly-1.8.0.4042023.post409.dist-info}/LICENSE.md +0 -0
  63. {mct_nightly-1.8.0.4032023.post406.dist-info → mct_nightly-1.8.0.4042023.post409.dist-info}/top_level.txt +0 -0
  64. /model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/{common → pytorch/quantizers/activation_inferable_quantizers}/activation_lut_pot_inferable_quantizer.py +0 -0
{mct_nightly-1.8.0.4032023.post406.dist-info → mct_nightly-1.8.0.4042023.post409.dist-info}/METADATA RENAMED
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.1
2
2
  Name: mct-nightly
3
- Version: 1.8.0.4032023.post406
3
+ Version: 1.8.0.4042023.post409
4
4
  Summary: A Model Compression Toolkit for neural networks
5
5
  Home-page: UNKNOWN
6
6
  License: UNKNOWN
@@ -49,8 +49,8 @@ MCT is developed by researchers and engineers working at Sony Semiconductor Isra
49
49
  ## Supported Features
50
50
 
51
51
  MCT supports different quantization methods:
52
- * Post training quantization (PTQ): [Keras API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/keras_post_training_quantization_experimental.html#ug-keras-post-training-quantization-experimental), [PyTorch API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/pytorch_post_training_quantization_experimental.html#ug-pytorch-post-training-quantization-experimental)
53
- * Gradient-based post training quantization (GPTQ): [Keras API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/keras_gradient_post_training_quantization_experimental.html#ug-keras-gradient-post-training-quantization-experimental), [PyTorch API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/pytorch_gradient_post_training_quantization_experimental.html#ug-pytorch-gradient-post-training-quantization-experimental)
52
+ * Post-training quantization (PTQ): [Keras API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/keras_post_training_quantization_experimental.html#ug-keras-post-training-quantization-experimental), [PyTorch API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/pytorch_post_training_quantization_experimental.html#ug-pytorch-post-training-quantization-experimental)
53
+ * Gradient-based post-training quantization (GPTQ): [Keras API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/keras_gradient_post_training_quantization_experimental.html#ug-keras-gradient-post-training-quantization-experimental), [PyTorch API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/pytorch_gradient_post_training_quantization_experimental.html#ug-pytorch-gradient-post-training-quantization-experimental)
54
54
  * Quantization aware training (QAT)[*](#experimental-features)
55
55
 
56
56
 
@@ -107,15 +107,15 @@ A nightly package is also available (unstable):
107
107
  pip install mct-nightly
108
108
  ```
109
109
 
110
- ### Requierments
110
+ ### Requirements
111
111
 
112
- To run MCT, one of the supported frameworks, Tenosflow/Pytorch, needs to be installed.
112
+ To run MCT, one of the supported frameworks, Tensorflow/Pytorch, needs to be installed.
113
113
 
114
- For using with Tensorflow please install the packages:
114
+ For use with Tensorflow please install the packages:
115
115
  [tensorflow](https://www.tensorflow.org/install),
116
116
  [tensorflow-model-optimization](https://www.tensorflow.org/model_optimization/guide/install)
117
117
 
118
- For using with PyTorch please install the packages:
118
+ For use with PyTorch please install the packages:
119
119
  [torch](https://pytorch.org/)
120
120
 
121
121
  Also, a [requirements](requirements.txt) file can be used to set up your environment.
{mct_nightly-1.8.0.4032023.post406.dist-info → mct_nightly-1.8.0.4042023.post409.dist-info}/RECORD RENAMED
@@ -1,4 +1,4 @@
1
- model_compression_toolkit/__init__.py,sha256=h7ysLzLClGy_hSX7fQRe3igfjaejc9k9-pMJ3vu_WcE,4254
1
+ model_compression_toolkit/__init__.py,sha256=oB0z1Vm7LRYpyc2-nZxeTLSddf3OSbIpXXTbL-bzVuk,3534
2
2
  model_compression_toolkit/core/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
3
3
  model_compression_toolkit/core/analyzer.py,sha256=etuO_VPRkHRlCKvfpQYGe56j8psu4QuXTumneLunj9g,2943
4
4
  model_compression_toolkit/core/exporter.py,sha256=U_-ea-zYHsnIt2ydameMLZ_gzDaCMI1dRa5IjA8RUuc,4233
@@ -11,7 +11,7 @@ model_compression_toolkit/core/common/defaultdict.py,sha256=n-F3dP-VTMnGy9KfCwp7
11
11
  model_compression_toolkit/core/common/framework_implementation.py,sha256=cW2P4G1Dq4SiQtATWWq3QjfIB9EbRFiWM43aV1ftZjw,22674
12
12
  model_compression_toolkit/core/common/framework_info.py,sha256=2mrdVpyTvDgrae8Wf_iAGI81vPlMCp9EwIbNw-Ywbfs,6430
13
13
  model_compression_toolkit/core/common/immutable.py,sha256=rSPd3Xpx4Rzt6-EwDA9tXvKygrrt4xvmq01JVCXY0hQ,1723
14
- model_compression_toolkit/core/common/logger.py,sha256=Pm5leGZT6CbcBadNvO9W2R2VTjUzVbr99KnAiSQwqgs,4673
14
+ model_compression_toolkit/core/common/logger.py,sha256=b9DVktZ-LymFcRxv2aL_sdiE6S2sSrFGWltx6dgEuUY,4863
15
15
  model_compression_toolkit/core/common/memory_computation.py,sha256=fYMU4fiCLlQRN4UIkoabuOvdwNHw4-lS1eNTuCVsBbQ,1217
16
16
  model_compression_toolkit/core/common/model_builder_mode.py,sha256=jll9-59OPaE3ug7Y9-lLyV99_FoNHxkGZMgcm0Vkpss,1324
17
17
  model_compression_toolkit/core/common/model_collector.py,sha256=9k1814EvteXo4WLaERIjU4KBRG_7WTjcQpAMPiCjAr4,5012
@@ -145,7 +145,7 @@ model_compression_toolkit/core/keras/keras_implementation.py,sha256=YJ5EAUbbcryj
145
145
  model_compression_toolkit/core/keras/keras_model_validation.py,sha256=IzlEriDsmTjZeWYIluIOeXNVlhTmaO-UGuFE7PBRG1o,1717
146
146
  model_compression_toolkit/core/keras/keras_node_prior_info.py,sha256=WGuyYfm3C2FScrI5WO-xPVeK_A9hJ-PyM4rb-CotsIs,3936
147
147
  model_compression_toolkit/core/keras/kpi_data_facade.py,sha256=xLYtLs-j0WiRtUFF_U7P3eH8xLR3606EabwtdklaGeg,8881
148
- model_compression_toolkit/core/keras/quantization_facade.py,sha256=D4ey6IAvDrrNs-90mOajJBzplzkDPqmNckOZ22Ukph4,18051
148
+ model_compression_toolkit/core/keras/quantization_facade.py,sha256=dluWNMmbUICeASvsSbTfPzmkqgVfuHzhwFyLIT_gLRQ,18032
149
149
  model_compression_toolkit/core/keras/tf_tensor_numpy.py,sha256=BauH-Ssoiuv5wu81fk1bm9HO_0yR0oTwKyLkXgE3qSE,2022
150
150
  model_compression_toolkit/core/keras/back2framework/__init__.py,sha256=rhIiXg_nBgUZ-baE3M6SzCuQbcnq4iebY1jtJBvKHOM,808
151
151
  model_compression_toolkit/core/keras/back2framework/factory_model_builder.py,sha256=MtfhDEou_7OWwFBgJIICxSUB0Flb18P1yM6nST5S9Xs,2231
@@ -153,7 +153,7 @@ model_compression_toolkit/core/keras/back2framework/float_model_builder.py,sha25
153
153
  model_compression_toolkit/core/keras/back2framework/instance_builder.py,sha256=V_7jxsA8jbBRNMVVazWeXNDhXeQFBkJORlM0TS5FN4Y,3837
154
154
  model_compression_toolkit/core/keras/back2framework/keras_model_builder.py,sha256=vvjdbKxAphIwdwAPcMitaDcfkp7BZSKXmZ9plmNO5Uo,15658
155
155
  model_compression_toolkit/core/keras/back2framework/mixed_precision_model_builder.py,sha256=3xeqOajVoL8V508BttX8ZOGyLLnhWLI0TOost-ae5fU,7162
156
- model_compression_toolkit/core/keras/back2framework/model_gradients.py,sha256=qeXebQh_ZHO8stxRqcqCkeJC1XP2uoK0wJn6IzFyJ1U,15305
156
+ model_compression_toolkit/core/keras/back2framework/model_gradients.py,sha256=a4qLD0pICSt6NyMtpC3-rsCxSVGDWdrUab0IfLtg9nw,15340
157
157
  model_compression_toolkit/core/keras/back2framework/quantized_model_builder.py,sha256=Qhhhc-JHx8BRupy5yZHU3XauJn9woDmmG8EmpcVZdAQ,2476
158
158
  model_compression_toolkit/core/keras/graph_substitutions/__init__.py,sha256=mjbqLD-KcG3eNeCYpu1GBS7VclGVOQ63x2p6mAAuba4,698
159
159
  model_compression_toolkit/core/keras/graph_substitutions/substitutions/__init__.py,sha256=mjbqLD-KcG3eNeCYpu1GBS7VclGVOQ63x2p6mAAuba4,698
@@ -199,7 +199,7 @@ model_compression_toolkit/core/keras/statistics_correction/__init__.py,sha256=9H
199
199
  model_compression_toolkit/core/keras/statistics_correction/apply_second_moment_correction.py,sha256=ZUUUxCkGUKRsdud0kHeosEhDg5MejDOa1NirRHbQYes,3055
200
200
  model_compression_toolkit/core/keras/visualization/__init__.py,sha256=mjbqLD-KcG3eNeCYpu1GBS7VclGVOQ63x2p6mAAuba4,698
201
201
  model_compression_toolkit/core/pytorch/__init__.py,sha256=Rf1RcYmelmdZmBV5qOKvKWF575ofc06JFQSq83Jz99A,696
202
- model_compression_toolkit/core/pytorch/constants.py,sha256=joqCvu_kmF1SrZNgxejNy3Dv1G1aeJalVywrjAy5PD0,2442
202
+ model_compression_toolkit/core/pytorch/constants.py,sha256=XC9uCV7zbkP47m0YoLla6VJCL_DBTmMqGe7iXLa-qes,2626
203
203
  model_compression_toolkit/core/pytorch/default_framework_info.py,sha256=lBNRSrY8LDJA2Oxk4qiVolUdeNIiDyUK1ek7-2ykc7Y,4219
204
204
  model_compression_toolkit/core/pytorch/kpi_data_facade.py,sha256=NtL-bRGF_0UWDUBlZt8mgdmvARYRwHSyWp3t3Y8Kq-U,8782
205
205
  model_compression_toolkit/core/pytorch/pytorch_implementation.py,sha256=3ENp_Hr5KN5OiLBqMzsag4AlAfZDywEDVOKCrKNkH-I,25965
@@ -211,7 +211,7 @@ model_compression_toolkit/core/pytorch/back2framework/factory_model_builder.py,s
211
211
  model_compression_toolkit/core/pytorch/back2framework/float_model_builder.py,sha256=GJWxQh8Lqshfi3vCqyAw0XmyAiDDe9q4YPQsQe9qXNQ,3414
212
212
  model_compression_toolkit/core/pytorch/back2framework/instance_builder.py,sha256=OkGH_MkrffaxADCbw75WKPtqVRQSi6AxOeNWoNgzOq0,1680
213
213
  model_compression_toolkit/core/pytorch/back2framework/mixed_precision_model_builder.py,sha256=Lx0w7vK4SmoxaVUeA1orLiQ7JbOknwDUs0kAHJJygrc,5152
214
- model_compression_toolkit/core/pytorch/back2framework/model_gradients.py,sha256=h3DgsIaNTRc02lu1WMDWBRULxWrfA8n7S0lUMXoHgDM,17808
214
+ model_compression_toolkit/core/pytorch/back2framework/model_gradients.py,sha256=sGEbv3TjzidMqxBRuQnCtAL2icV6CvNO2JHwTe5FAkg,18238
215
215
  model_compression_toolkit/core/pytorch/back2framework/pytorch_model_builder.py,sha256=DJ-NM2rhaRQ1qCX-ySq5CVSOaKDKbElo5DmjS_Se3Es,12685
216
216
  model_compression_toolkit/core/pytorch/back2framework/quantized_model_builder.py,sha256=Cbee-M10q5LaLo43zvndWwPcxvY4wfBiwpaVsI0Y0ng,3704
217
217
  model_compression_toolkit/core/pytorch/back2framework/quantization_wrapper/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
@@ -227,7 +227,7 @@ model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/linear_
227
227
  model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/multi_head_attention_decomposition.py,sha256=b6T6rA_6RZtaCtmcXn7tC04Bs7XvzkSE68Xq-YNZOP8,38365
228
228
  model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/permute_call_method.py,sha256=EMCviyFyJFLEKuAUz3rZHLfB9MAU1kywSBL2XQNzLlg,1953
229
229
  model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/relu_bound_to_power_of_2.py,sha256=Gxy_WuFyRhtmw6bSeuhoGwriiu7vcL4bvOTAMWT5SNs,5563
230
- model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/reshape_with_static_shapes.py,sha256=iMW_85CB-f6JAymCL3vhSEEyCCiq6sa0335Y0pCIn6c,3492
230
+ model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/reshape_with_static_shapes.py,sha256=u0bcymbcrO7XVRqcBwUqqNDq7PIjlndLp2OS8v0jxNo,4153
231
231
  model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/residual_collapsing.py,sha256=GCMXOWnXFQhgGoBUkdiZu08x33E4iJYq8mLlWwxt4vw,2911
232
232
  model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/scale_equalization.py,sha256=XFtU9yuBmoZlX0f0mS6otMPWMk-RcWs94XdvvTNhW8Y,3303
233
233
  model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/shift_negative_activation.py,sha256=EgBe7ij5LDD2i6yt320aeMl4AoJIAyOeKYg4MOsq7es,9833
@@ -300,8 +300,8 @@ model_compression_toolkit/core/tpc_models/tflite_tpc/v1/__init__.py,sha256=t4JKs
300
300
  model_compression_toolkit/core/tpc_models/tflite_tpc/v1/tp_model.py,sha256=3u8pxc6lgbOb_nT6BSSfoOML5LFdE0abXS0s83AZAHI,7770
301
301
  model_compression_toolkit/core/tpc_models/tflite_tpc/v1/tpc_keras.py,sha256=kEpZOfjUA5-qzdhuw4DSMNlMJdYjN05rWMQfo5rtM3I,6066
302
302
  model_compression_toolkit/core/tpc_models/tflite_tpc/v1/tpc_pytorch.py,sha256=TN51UfSK6lOQ_JRoXNzzglGpetrcYltLeg98Tz9mpC8,4937
303
- model_compression_toolkit/exporter/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
304
- model_compression_toolkit/exporter/model_exporter/__init__.py,sha256=zRTlsEE8qlPaSFbnvGWVrXtJw_r3eYCBvSEdLk4ySkA,1082
303
+ model_compression_toolkit/exporter/__init__.py,sha256=Ic52ZgHIPuAWsgWX7LuzA9TkRPo_flpQizJAc39ttrc,1083
304
+ model_compression_toolkit/exporter/model_exporter/__init__.py,sha256=9HIBmj8ROdCA-yvkpA8EcN6RHJe_2vEpLLW_gxOJtak,698
305
305
  model_compression_toolkit/exporter/model_exporter/fw_agonstic/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
306
306
  model_compression_toolkit/exporter/model_exporter/fw_agonstic/exporter.py,sha256=iwZZEs_1AwLvClepYG38P_oTrQrA2YXxFTQUNMVoyS4,2022
307
307
  model_compression_toolkit/exporter/model_exporter/keras/__init__.py,sha256=uZ2RigbY9O2PJ0Il8wPpS_s7frgg9WUGd_SHeKGyl1A,699
@@ -314,56 +314,60 @@ model_compression_toolkit/exporter/model_exporter/pytorch/fakely_quant_onnx_pyto
314
314
  model_compression_toolkit/exporter/model_exporter/pytorch/fakely_quant_torchscript_pytorch_exporter.py,sha256=T3K43YiDGa2g14SbZeDRqhr-3kFYVOyR6EZoSlIlLyc,2892
315
315
  model_compression_toolkit/exporter/model_exporter/pytorch/pytorch_export_facade.py,sha256=iB6o2arjkQIiFcXl8oVoTowmEu0WXBbPjrcZQLjVWbQ,3925
316
316
  model_compression_toolkit/exporter/model_exporter/tflite/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
317
- model_compression_toolkit/exporter/model_exporter/tflite/fakely_quant_tflite_exporter.py,sha256=-44sgg-NM89xy8ARjG9uvG-6-uZUQzLSR8A6XaJdT0g,3006
317
+ model_compression_toolkit/exporter/model_exporter/tflite/fakely_quant_tflite_exporter.py,sha256=59d__yulPJhD2_dIT2EiFWlw5qMfCLq0NwZarXOOHFU,3074
318
318
  model_compression_toolkit/exporter/model_exporter/tflite/int8_tflite_exporter.py,sha256=K_eHtLZ9jTUvEiLDL4uX-X_Q6HLwaIB--uUdNMvJFsE,8169
319
319
  model_compression_toolkit/exporter/model_exporter/tflite/tflite_export_facade.py,sha256=MFBd6DwO92Ub58uEXKcw6VQbtr40StokXjpRUA4XV-U,3338
320
- model_compression_toolkit/exporter/model_wrapper/__init__.py,sha256=GJg_o-jj48QDlDDX0eEVwWuD33iEx-CaBU4JFbcEYsQ,1314
320
+ model_compression_toolkit/exporter/model_wrapper/__init__.py,sha256=7CF2zvpTrIEm8qnbuHnLZyTZkwBBxV24V8QA0oxGbh0,1187
321
321
  model_compression_toolkit/exporter/model_wrapper/keras/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
322
- model_compression_toolkit/exporter/model_wrapper/keras/validate_layer.py,sha256=acSCl7AuPsg0FY4Y2dqOi8x5BEpfYpAEHDEqlaatELY,3335
322
+ model_compression_toolkit/exporter/model_wrapper/keras/validate_layer.py,sha256=QVSVD3ayI431id5e80BgBDELpUW5rYsHagR6tC2BcGo,3882
323
323
  model_compression_toolkit/exporter/model_wrapper/keras/builder/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
324
- model_compression_toolkit/exporter/model_wrapper/keras/builder/fully_quantized_model_builder.py,sha256=21ugFfg8kwO4Y9nbphJDqvh-pJyqUz0AG3emdGf38co,2494
325
- model_compression_toolkit/exporter/model_wrapper/keras/builder/node_to_quantizer.py,sha256=KZickC6F4HEyGxDcMG6tZX_KBfD7j8WLDMTyPBgIPc8,6853
324
+ model_compression_toolkit/exporter/model_wrapper/keras/builder/fully_quantized_model_builder.py,sha256=qhs7renKRKc5S-ad2ctiF-GeIhkpFIlDsL3Co0i9ViI,2988
325
+ model_compression_toolkit/exporter/model_wrapper/keras/builder/node_to_quantizer.py,sha256=84jpx_10f3DYFU4svDOHReykrYVj-MXcthxIQW4mwTk,8596
326
326
  model_compression_toolkit/exporter/model_wrapper/keras/builder/node_to_quantizers.py,sha256=n7VTA-a9TrLFpfdYAqrAKj6PGlAyLq8-xdwnMMpX71k,2077
327
327
  model_compression_toolkit/exporter/model_wrapper/pytorch/__init__.py,sha256=Rf1RcYmelmdZmBV5qOKvKWF575ofc06JFQSq83Jz99A,696
328
- model_compression_toolkit/exporter/model_wrapper/pytorch/validate_layer.py,sha256=qydJE_218J_8k2bzERbE4l305_MhEykrAZbEUtj6lKI,1571
328
+ model_compression_toolkit/exporter/model_wrapper/pytorch/validate_layer.py,sha256=-DbO4K3F7awXzCiIKi6vVWgiRqPBe3CaPiVGGfn4iQ4,2036
329
329
  model_compression_toolkit/exporter/model_wrapper/pytorch/builder/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
330
- model_compression_toolkit/exporter/model_wrapper/pytorch/builder/fully_quantized_model_builder.py,sha256=ijnRigugvu7APEsNxS0CMX_BAxv9Vn9m9M0sEecG9D4,2041
331
- model_compression_toolkit/exporter/model_wrapper/pytorch/builder/node_to_quantizer.py,sha256=DLD4ao29pBErnnp-ZT0x9JlUxb_X3IQ_B0cqvrhd6MA,6239
330
+ model_compression_toolkit/exporter/model_wrapper/pytorch/builder/fully_quantized_model_builder.py,sha256=XrAqmyUuC1EtoyR010Quy9qre-60RsvnTbqK7dhwU38,2496
331
+ model_compression_toolkit/exporter/model_wrapper/pytorch/builder/node_to_quantizer.py,sha256=BlbojVGO4QG0_f7kjdGkGGESQzsXX0AgmJAf6KHHbrU,7674
332
332
  model_compression_toolkit/exporter/model_wrapper/pytorch/builder/node_to_quantizers.py,sha256=hinP-wtyxZyoW860GdJAk6M3iPjmwwPXQTUxd56yhq8,2086
333
- model_compression_toolkit/gptq/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
333
+ model_compression_toolkit/gptq/__init__.py,sha256=tPxlcYl8JwK-EWVTy5IVgGOaUJsnG-6PnOKeYNeGJjQ,1250
334
334
  model_compression_toolkit/gptq/runner.py,sha256=rB11-U68ZcR_IosgMYNegwyO5mGRjMMgL9OdHEIZOGU,5484
335
335
  model_compression_toolkit/gptq/common/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
336
- model_compression_toolkit/gptq/common/gptq_config.py,sha256=GTHdtKrQ3-MHJ-Sr-IliOgsc-TSWe4HDfSmv2RNtwec,11798
337
- model_compression_toolkit/gptq/common/gptq_constants.py,sha256=dTSq70A7ySQaQr2OkDNNe4VidXPwS18CVrYkTEGDfh4,812
336
+ model_compression_toolkit/gptq/common/gptq_config.py,sha256=35Ywz9IKWl5iB_oe-aO5dyhNtXXxSm_8gZAlA8wUQng,9507
337
+ model_compression_toolkit/gptq/common/gptq_constants.py,sha256=QSm6laLkIV0LYmU0BLtmKp3Fi3SqDfbncFQWOGA1cGU,611
338
338
  model_compression_toolkit/gptq/common/gptq_graph.py,sha256=tXWLUtP52OLgC3WO9M9kaG2OYVDE9imY9L4ef16lAqY,2826
339
- model_compression_toolkit/gptq/common/gptq_quantizer_config.py,sha256=ACCgISnZ9ir5Qa3t7TdSWTxwbm27o_gt0K-HMLUQqkY,3389
340
- model_compression_toolkit/gptq/common/gptq_training.py,sha256=DCEVjHXoLPEVzmnGDvzQvPaTSDfog1O1Hq_0Ziuusc4,15250
339
+ model_compression_toolkit/gptq/common/gptq_training.py,sha256=5gTLlvP5NBT-zY7QxW5LmarMFUaHCTL01KZdbvl_-lA,15106
341
340
  model_compression_toolkit/gptq/keras/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
342
341
  model_compression_toolkit/gptq/keras/gptq_loss.py,sha256=rbRkF15MYd6nq4G49kcjb_dPTa-XNq9cTkrb93mXawo,6241
343
- model_compression_toolkit/gptq/keras/gptq_training.py,sha256=r61e6-fkfCOqZ5q8GeR20xiHjqzFg_x5VfZNjcGz1DM,16554
344
- model_compression_toolkit/gptq/keras/graph_info.py,sha256=FyHIMTMZCP0hqrSTvs63_q6QP7bLGZfRdP7OpNwvXCs,5174
345
- model_compression_toolkit/gptq/keras/quantization_facade.py,sha256=DSagaoLEljHZKAeKgmomuYifhjjDe5VJTVGAPtvffR0,14341
342
+ model_compression_toolkit/gptq/keras/gptq_training.py,sha256=d0M2RkiRvKROZlc3GCr4_SHRSWXNc3RU081K1GF4a0M,15891
343
+ model_compression_toolkit/gptq/keras/graph_info.py,sha256=nYKL3hrd6L3EiyxejrE1xJUeppxSmflHO4nt2fkE-aY,4399
344
+ model_compression_toolkit/gptq/keras/quantization_facade.py,sha256=tNL3_0NdsFUqtCQ9_Qjy4_hYfikXmZZGXdytl-lz8tQ,14112
346
345
  model_compression_toolkit/gptq/keras/quantizer/__init__.py,sha256=H76G9W-tYSpHBnqRRgIUoWOjhjKN7XN00njHgjBT_JA,872
347
- model_compression_toolkit/gptq/keras/quantizer/base_keras_gptq_quantizer.py,sha256=Ch1Mk8M_U9L5tBArFFeg_7g_3iTSr7n65LfQOSAZQ1E,4827
346
+ model_compression_toolkit/gptq/keras/quantizer/base_keras_gptq_quantizer.py,sha256=bEIgXk17VzyxDGHBqHVg_Ox9PV8I7zd_39Qkt90XzbQ,4782
348
347
  model_compression_toolkit/gptq/keras/quantizer/quant_utils.py,sha256=XT1idm84wl4GDGoqGI8L5XH-H9OjIlhTjGCjY-ylYQw,4604
349
- model_compression_toolkit/gptq/keras/quantizer/quantization_builder.py,sha256=8fbdLLJlMm8mf8MDjH0BV02G_1HA1gxuS-2qENVJ58U,4391
348
+ model_compression_toolkit/gptq/keras/quantizer/quantization_builder.py,sha256=Ngqk6rAfS5tWwF2DkpVE_u69Q1Kf15aaSuZ37bwOpBs,4392
349
+ model_compression_toolkit/gptq/keras/quantizer/regularization_factory.py,sha256=iKzHnxl2ZSEp09oatfJVoiDuu6Q_iN36mOxQzDr1cy8,2087
350
350
  model_compression_toolkit/gptq/keras/quantizer/soft_rounding/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
351
- model_compression_toolkit/gptq/keras/quantizer/soft_rounding/symmetric_soft_quantizer.py,sha256=rt9G-81c8s5HNknOhSe_r0zBbmHeOXNMOQuFlNmyJ-4,15326
351
+ model_compression_toolkit/gptq/keras/quantizer/soft_rounding/soft_quantizer_reg.py,sha256=HqvDvrQKRYvjvjEl4p4a_r0lrMSkO3gAUM1KV4EV5Js,3976
352
+ model_compression_toolkit/gptq/keras/quantizer/soft_rounding/symmetric_soft_quantizer.py,sha256=m33BS3VYGwOK-ovh_SBwOlg9lmsyKiB6U23WjWYLVY0,12060
352
353
  model_compression_toolkit/gptq/keras/quantizer/ste_rounding/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
353
- model_compression_toolkit/gptq/keras/quantizer/ste_rounding/symmetric_ste.py,sha256=XZQkZ8DDxEqt57NP_g096ijBOG--MhiR8S1snCFPA2o,9364
354
+ model_compression_toolkit/gptq/keras/quantizer/ste_rounding/symmetric_ste.py,sha256=2xjOHUFsy8A0kgmhw5AiXnofN8eSb_gzyLnrGT50Hm8,8539
354
355
  model_compression_toolkit/gptq/pytorch/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
355
356
  model_compression_toolkit/gptq/pytorch/gptq_loss.py,sha256=kDuWw-6zh17wZpYWh4Xa94rpoodf82DksgjQCnL7nBc,2719
356
- model_compression_toolkit/gptq/pytorch/gptq_training.py,sha256=kEScULboEMgc3OuxBPyJ6WggRLrFxLZQ8wuTuASlesc,14105
357
- model_compression_toolkit/gptq/pytorch/graph_info.py,sha256=YPMzFA0s7JjFKI7Mi6hYcszMVfKPJrSWaKMNG2HYwz0,4368
358
- model_compression_toolkit/gptq/pytorch/quantization_facade.py,sha256=lOEx4Gkdqk4EcpG98m20nEINwe56Kyr1EjfyY55Hfmw,13255
359
- model_compression_toolkit/gptq/pytorch/quantizer/__init__.py,sha256=z-KoLKfu53nL_sWm2ifM5yO5Ham1HE7rOHNtsk53QVU,876
360
- model_compression_toolkit/gptq/pytorch/quantizer/base_pytorch_gptq_quantizer.py,sha256=VgAyLxeOh24p9tn8H-VNlfZzIM6VMxnarYfIzNWjXsE,4884
361
- model_compression_toolkit/gptq/pytorch/quantizer/quant_utils.py,sha256=OxEudXw8liJc45ZnloQtu3P1UWw2C7d0S_VIx6s9UQw,3579
362
- model_compression_toolkit/gptq/pytorch/quantizer/quantization_builder.py,sha256=NsBKD8gvccZAvSUj98j7ISJc-IoXo9nIB4jJLav34d8,4220
357
+ model_compression_toolkit/gptq/pytorch/gptq_training.py,sha256=EaPX7MfaNlzFHogrxN14-G9zGPyt8Bpgya7O0WaUkgk,13516
358
+ model_compression_toolkit/gptq/pytorch/graph_info.py,sha256=Sphpr5wKADgwZ-sLxNqMAcsEiP_jaFEL7q2-zcrtUx8,3791
359
+ model_compression_toolkit/gptq/pytorch/quantization_facade.py,sha256=LPYUDDY393M9sk0iWiWXsIGjuGYPYiUPqiGnfCzmtrY,12468
360
+ model_compression_toolkit/gptq/pytorch/quantizer/__init__.py,sha256=ZHNHo1yzye44m9_ht4UUZfTpK01RiVR3Tr74-vtnOGI,968
361
+ model_compression_toolkit/gptq/pytorch/quantizer/base_pytorch_gptq_quantizer.py,sha256=IArGYcTb_c7aDnZOOlLGCuuZcV4A0DaxoYO3i-fbVNM,4291
362
+ model_compression_toolkit/gptq/pytorch/quantizer/quant_utils.py,sha256=rKRb6TgNko_NhZnwUA2xzwvXAM5qj_mWNjhy_h2SwI8,3888
363
+ model_compression_toolkit/gptq/pytorch/quantizer/quantization_builder.py,sha256=qprTfTkqqcAijNKsHwKsOlju75Ihu_PDEJxny_A5AD0,4221
364
+ model_compression_toolkit/gptq/pytorch/quantizer/regularization_factory.py,sha256=9owTzSu_xz29dsjONB-AYXuCZoPo_4nqxTk3yH18a0g,2089
363
365
  model_compression_toolkit/gptq/pytorch/quantizer/soft_rounding/__init__.py,sha256=lNJ29DYxaLUPDstRDA1PGI5r9Fulq_hvrZMlhst1Z5g,697
364
- model_compression_toolkit/gptq/pytorch/quantizer/soft_rounding/symmetric_soft_quantizer.py,sha256=VJBQVy7nhdBYQ_NmpltiiPOCPeRXqmmVQZBqPTEfcU8,15172
366
+ model_compression_toolkit/gptq/pytorch/quantizer/soft_rounding/soft_quantizer_reg.py,sha256=EGs3GfC73yl7E1xlDd_sHS7Vof1Td7RAq_Ny7hAbZZA,4178
367
+ model_compression_toolkit/gptq/pytorch/quantizer/soft_rounding/symmetric_soft_quantizer.py,sha256=dQ4tqk49V-nED6GZRx2iGqpnObKTatJq60FfcpfiJUg,12180
368
+ model_compression_toolkit/gptq/pytorch/quantizer/soft_rounding/uniform_soft_quantizer.py,sha256=5CmooZOZhIt9pTsgYUOSrJAoEz6YxOr0QtVXP3pZfuw,9191
365
369
  model_compression_toolkit/gptq/pytorch/quantizer/ste_rounding/__init__.py,sha256=Rf1RcYmelmdZmBV5qOKvKWF575ofc06JFQSq83Jz99A,696
366
- model_compression_toolkit/gptq/pytorch/quantizer/ste_rounding/symmetric_ste.py,sha256=vW8lRfYdxVg7Afki4wD2rFHk_MA28I6DcEWiFX9JHhw,9532
370
+ model_compression_toolkit/gptq/pytorch/quantizer/ste_rounding/symmetric_ste.py,sha256=T1z8KfcwwnWmq6Rh0Ncyx1JiaeTxKBQAkxVsI4yD4J4,8932
367
371
  model_compression_toolkit/ptq/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
368
372
  model_compression_toolkit/ptq/runner.py,sha256=_c1dSjlPPpsx59Vbg1buhG9bZq__OORz1VlPkwjJzoc,2552
369
373
  model_compression_toolkit/ptq/keras/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
@@ -381,8 +385,8 @@ model_compression_toolkit/qat/keras/quantizer/base_keras_qat_quantizer.py,sha256
381
385
  model_compression_toolkit/qat/keras/quantizer/quant_utils.py,sha256=rS2z_ozyjzQ07MMczaAFNZ7K6RKwAnBOKyRac4UvF44,2123
382
386
  model_compression_toolkit/qat/keras/quantizer/quantization_builder.py,sha256=ESYtJGA6SGT0103Q1r33VGTu60V05pux7fK8JOnRau0,4229
383
387
  model_compression_toolkit/qat/keras/quantizer/ste_rounding/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
384
- model_compression_toolkit/qat/keras/quantizer/ste_rounding/symmetric_ste.py,sha256=SP-EoQI5H2vnsmA8_T8R5kSv9tdNAgskXps_VqpS6A8,13820
385
- model_compression_toolkit/qat/keras/quantizer/ste_rounding/uniform_ste.py,sha256=pClxv-mjowimJ5jbV3Gtsv8pJx0arQ9nrtTLciQSqu4,11089
388
+ model_compression_toolkit/qat/keras/quantizer/ste_rounding/symmetric_ste.py,sha256=Cc1ohHnWOlWnjllAxEA2ZrLZqPiXt_oH6YfQWkEhPhI,13599
389
+ model_compression_toolkit/qat/keras/quantizer/ste_rounding/uniform_ste.py,sha256=XGvg0OPQdi1WvYZJ7GA9VUBKGTWIt_VKLzdPoXfw7xQ,10939
386
390
  model_compression_toolkit/qat/pytorch/__init__.py,sha256=cco4TmeIDIh32nj9ZZXVkws4dd9F2UDrmjKzTN8G0V0,697
387
391
  model_compression_toolkit/qat/pytorch/quantization_facade.py,sha256=8iH6ENPyVbPFzpJS2FaXPvjZ1VzkSgGWsb-EnHywjRA,11875
388
392
  model_compression_toolkit/qat/pytorch/quantizer/__init__.py,sha256=R4vwVcbg6QprCTNzibyF9PtbKKKBsfu9ffypKDNscJQ,859
@@ -390,25 +394,24 @@ model_compression_toolkit/qat/pytorch/quantizer/base_pytorch_qat_quantizer.py,sh
390
394
  model_compression_toolkit/qat/pytorch/quantizer/quantization_builder.py,sha256=V-oF596RdXW8W8nF8g57bEGsvB8ORvRIFoyrXBwyaWc,4086
391
395
  model_compression_toolkit/qat/pytorch/quantizer/quantizer_utils.py,sha256=5XswoF-5aaFangkHssWKAQTsk6lf_zzndzfCsBWBVMs,5004
392
396
  model_compression_toolkit/qat/pytorch/quantizer/ste_rounding/__init__.py,sha256=Rf1RcYmelmdZmBV5qOKvKWF575ofc06JFQSq83Jz99A,696
393
- model_compression_toolkit/qat/pytorch/quantizer/ste_rounding/symmetric_ste.py,sha256=ZEcs3IVngv_4TV7Zf6XVs5nD3q3y_FtDBkfBmtQxYu4,9735
394
- model_compression_toolkit/qat/pytorch/quantizer/ste_rounding/uniform_ste.py,sha256=yU1sijubpt6wXPqgWEB7yr24FYy-tg51Ly4vsN0ROIc,8673
397
+ model_compression_toolkit/qat/pytorch/quantizer/ste_rounding/symmetric_ste.py,sha256=qTVB979OYn7yPGfD319h_E4IgoP_o52EdWw3SSiiFis,9766
398
+ model_compression_toolkit/qat/pytorch/quantizer/ste_rounding/uniform_ste.py,sha256=_9gD5F-lAh4w7BkkiH4EO8VU_40_ySClXjLXf-7zPek,8810
395
399
  model_compression_toolkit/quantizers_infrastructure/__init__.py,sha256=wKRMA6H4aTPnQlvmx_gbe4d2KMZ-40QIGDJCZa-2VCU,1615
396
400
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
397
401
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
398
- model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/activation_lut_pot_inferable_quantizer.py,sha256=GH14y6rV3OKbfJTdv8NPee4tHwWPuIbuhwejkchMnn8,4599
399
402
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/base_inferable_quantizer.py,sha256=Zo2N4lAo1Lz91MSor0s-dng3ueFP6UpZlt9CcoCu5SM,3141
400
403
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/constants.py,sha256=tYmi5_PpsSJ-LyOcMslU0FEnG_c31VXyujc1_R7-EVk,1664
401
- model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/get_all_subclasses.py,sha256=6VyiXsp27SpERafPrWbTta0M9BQfH2MJgK44XV8Y6UY,1222
402
- model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/get_quantizers.py,sha256=xWd4gthIF_HXDbZvV1FWYyx02g0PGk0LU4-wOWcPSws,2968
404
+ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/get_all_subclasses.py,sha256=6KqemyFcM4a6KoCZ-6dm46iIZ_kusPnj5crH8RTAvuo,1213
405
+ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/get_quantizers.py,sha256=8GCjwp665nXcUZEv2RBeDT2MMZ1YvvORaxAkXooWJy0,2967
403
406
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/quant_utils.py,sha256=SDzQuh3q9ugSD80Z9IuaWOPskH5VsRRyuBOeIeWJDdQ,2153
404
407
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
405
408
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/load_model.py,sha256=H0-I_-4ChTLNk7gwcnft5YV-q9Cg1wue0OUuZ1W4fek,4279
406
409
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantize_wrapper.py,sha256=uE05T6VbLScwpyZoAXoJsCC_aYauv0LjyxSm89yZu84,14646
407
410
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizer_utils.py,sha256=Z0SoHkTl5dC0y3hrcj0bC5fSa-oU7IYuGN5sBTb4THA,3440
408
411
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/validation_functions.py,sha256=0heI5awgysKl9-XhIWxPiCbC_IBC7mANzBP95Tu1G9o,2942
409
- model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/__init__.py,sha256=Be0s33EOWPYCyyk1VsqU6FY6yeVPFdD5WH51CEPbODw,2112
412
+ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/__init__.py,sha256=Xg2t9VH61BIvqYtJg6GBKkAnDmFe1u4K0i0r2OBAK-I,2742
410
413
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/base_keras_inferable_quantizer.py,sha256=3ILK53o5RogFy1ZI2gyaJk6TjH2HV77j5uSqkSc_8W0,2228
411
- model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/constants.py,sha256=fxsgfxqcDHMloaPOaJU3FYjV3o_XS-yM7AeisIQpvPg,944
414
+ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/constants.py,sha256=ItUEs8c7LVxBPMopLD5BO2Ry9DIxFIrk_M7AdSEyBFg,979
412
415
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/activation_inferable_quantizers/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
413
416
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/activation_inferable_quantizers/activation_lut_pot_inferable_quantizer.py,sha256=aUiK_aOy8YImLyXT6_g3BIQ6Lt9FdPPPYjj1R93qsvc,7485
414
417
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/activation_inferable_quantizers/activation_pot_inferable_quantizer.py,sha256=z-t_WU9OGURVpG3e6uBQ8Mx4JPhOxqS0Tpg_ioa0MhA,3191
@@ -421,15 +424,16 @@ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/ker
421
424
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/weights_inferable_quantizers/weights_symmetric_inferable_quantizer.py,sha256=pfoZCZNKsovKq3XiSim5ErSBid3ihK9WORaZh7i8wIg,4376
422
425
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/weights_inferable_quantizers/weights_uniform_inferable_quantizer.py,sha256=PWl58Dpoxowz2HFg7IpT74A5o3_GjFKOsmonK2Nk0Uo,8664
423
426
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
424
- model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantize_wrapper.py,sha256=4xRw5ySkhVPjLvKqQ5zYQPnh2XYJCx4YsBieU_iycno,10937
427
+ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantize_wrapper.py,sha256=AXqeWmB570xucOdjUt5P1LbrRIkprdj26hW1q7YHjtE,11272
425
428
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizer_utils.py,sha256=2KN976TZTObiaEhoUL0-Rpceui-Nifw5LdKLdU7SRY0,5929
426
- model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/__init__.py,sha256=g6N2eOWr7UT56X4OoR5G4h0YF1hUqM-u9_xXLdlMQxc,2166
429
+ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/__init__.py,sha256=-hiXng1pF3wjI-YYYZqZ-NZ1TStGuec4bci3jxvYVY0,2820
427
430
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/base_lut_symmetric_inferable_quantizer.py,sha256=knIkFr8xnWFyBu8VT129eH8_Mj0Osb2jIZx32zQjgLY,4871
428
431
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/base_pytorch_inferable_quantizer.py,sha256=jZ8q-vzgsUCXv32OYpZySlCC5GeA07x6NVO5q5K4TsE,1919
429
432
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/base_symmetric_inferable_quantizer.py,sha256=gIKmFouy2KAJt7cbRZ1lmSitPtFNNkCgqgLJrnn9gRQ,3070
430
433
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/base_uniform_inferable_quantizer.py,sha256=G6ReERM4j9sGuRKuOPXXo-H1WMEMX7_OgQzxAZP0aaE,2501
431
- model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/constants.py,sha256=2K_ABBFM7BjXDLPwkhzvA3QxdsV_pTaBAowTW5wQTeo,920
434
+ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/constants.py,sha256=9bxo6snEJkRv5XWmhBGsV6g8LCe_1NgAE5ufIq2ewYU,1007
432
435
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/activation_inferable_quantizers/__init__.py,sha256=lNJ29DYxaLUPDstRDA1PGI5r9Fulq_hvrZMlhst1Z5g,697
436
+ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/activation_inferable_quantizers/activation_lut_pot_inferable_quantizer.py,sha256=GH14y6rV3OKbfJTdv8NPee4tHwWPuIbuhwejkchMnn8,4599
433
437
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/activation_inferable_quantizers/activation_pot_inferable_quantizer.py,sha256=mu17Qv9I4bzd3IAac7zFg-Goy6uQaNZlEplVRmZBArY,2928
434
438
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/activation_inferable_quantizers/activation_symmetric_inferable_quantizer.py,sha256=qV2F9wVUAAT0rn9STDN5gbQPbwh9EsAH8pQog-V_scg,3631
435
439
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/activation_inferable_quantizers/activation_uniform_inferable_quantizer.py,sha256=qSAo-pAPWnY_-DO-84-xJYUKrc9kih0AUo3KCyre0BY,4855
@@ -441,19 +445,19 @@ model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pyt
441
445
  model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/weights_inferable_quantizers/weights_uniform_inferable_quantizer.py,sha256=fbXuZBlVCGjNKp3hlIp3W9NM-dtzP75c19wkvjbNCMo,5394
442
446
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
443
447
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
444
- model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/base_trainable_quantizer.py,sha256=GMKp8NG_67lIHz1SyrteKuZK4wGMbNHQ5z8MrMVP01Y,6408
448
+ model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/base_trainable_quantizer.py,sha256=qnGMQfwttYetjeJJcdKsXJKpcb75Sy_HTS13Oorfgvo,7775
445
449
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/get_quantizer_config.py,sha256=uQPnn1tkD96EmTcg-zAnxKH5XzY5y9zYb6tJ9ZTm_oI,6333
446
- model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/get_quantizers.py,sha256=JMlDcOxuTgd9jVrtESq-dzBX24gbHs0PmYOzTkW88Sw,3789
450
+ model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/get_quantizers.py,sha256=RRtxOjiB1gFMiqYMlUC7hqZMdJGN5FFMBj7-sD2aWJ8,3831
447
451
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/quant_utils.py,sha256=zdiew1jwR7tUKm9XWlHnAPxIZsAdKqbzzC2vH02j5wA,1505
448
452
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/trainable_quantizer_config.py,sha256=WBTYdDQtXkscjO8b7leunBVSGG8JC__DhhpINx7lCEA,4774
449
453
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/keras/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
450
- model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/keras/base_keras_quantizer.py,sha256=2xAPiOCb2kXb6w8DML0fH17Yk425yAVmbvC6lTs9IUE,3411
454
+ model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/keras/base_keras_quantizer.py,sha256=VMwT-UDKKjZQtvEj7xEUUKAuGvzXr3ak05fPy64nnsw,4307
451
455
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/keras/config_serialization.py,sha256=Vj-59ImTj0YEGI7MgRgwqJWIeGcIlrdLzPDiedwHV_E,4062
452
456
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/keras/quantizer_utils.py,sha256=MVwXNymmFRB2NXIBx4e2mdJ1RfoHxRPYRgjb1MQP5kY,1797
453
457
  model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/pytorch/__init__.py,sha256=huHoBUcKNB6BnY6YaUCcFvdyBtBI172ZoUD8ZYeNc6o,696
454
- model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/pytorch/base_pytorch_quantizer.py,sha256=YmnoDMJUYkvkDV05aPVfr3Y73mR11CUz9YWSSAJ2mcI,2266
455
- mct_nightly-1.8.0.4032023.post406.dist-info/LICENSE.md,sha256=aYSSIb-5AFPeITTvXm1UAoe0uYBiMmSS8flvXaaFUks,10174
456
- mct_nightly-1.8.0.4032023.post406.dist-info/METADATA,sha256=irIopWoNXiuXKMSLM_Kj2ILgcEls5STvbh4_WUeE10g,10974
457
- mct_nightly-1.8.0.4032023.post406.dist-info/WHEEL,sha256=2wepM1nk4DS4eFpYrW1TTqPcoGNfHhhO_i5m4cOimbo,92
458
- mct_nightly-1.8.0.4032023.post406.dist-info/top_level.txt,sha256=gsYA8juk0Z-ZmQRKULkb3JLGdOdz8jW_cMRjisn9ga4,26
459
- mct_nightly-1.8.0.4032023.post406.dist-info/RECORD,,
458
+ model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/pytorch/base_pytorch_quantizer.py,sha256=V16tvEpBz5-Pfl0h8dkPs4F32DWXyYdEn6HnQHYKCfs,3161
459
+ mct_nightly-1.8.0.4042023.post409.dist-info/LICENSE.md,sha256=aYSSIb-5AFPeITTvXm1UAoe0uYBiMmSS8flvXaaFUks,10174
460
+ mct_nightly-1.8.0.4042023.post409.dist-info/METADATA,sha256=ykeWnMeaN2geVVupGfQh8pf-KlYPq3p6YvSiQpxdDiI,10971
461
+ mct_nightly-1.8.0.4042023.post409.dist-info/WHEEL,sha256=pkctZYzUS4AYVn6dJ-7367OJZivF2e8RA9b_ZBjif18,92
462
+ mct_nightly-1.8.0.4042023.post409.dist-info/top_level.txt,sha256=gsYA8juk0Z-ZmQRKULkb3JLGdOdz8jW_cMRjisn9ga4,26
463
+ mct_nightly-1.8.0.4042023.post409.dist-info/RECORD,,
{mct_nightly-1.8.0.4032023.post406.dist-info → mct_nightly-1.8.0.4042023.post409.dist-info}/WHEEL RENAMED
@@ -1,5 +1,5 @@
1
1
  Wheel-Version: 1.0
2
- Generator: bdist_wheel (0.38.4)
2
+ Generator: bdist_wheel (0.40.0)
3
3
  Root-Is-Purelib: true
4
4
  Tag: py3-none-any
5
5
 
model_compression_toolkit/__init__.py CHANGED
@@ -14,8 +14,6 @@
14
14
  # ==============================================================================
15
15
 
16
16
  from model_compression_toolkit.core.common.quantization.debug_config import DebugConfig
17
- from model_compression_toolkit.gptq.common.gptq_config import GradientPTQConfig, RoundingType, GradientPTQConfigV2
18
- from model_compression_toolkit.gptq.common.gptq_quantizer_config import GPTQQuantizerConfig, SoftQuantizerConfig
19
17
  from model_compression_toolkit.core.common.quantization import quantization_config
20
18
  from model_compression_toolkit.core.common.mixed_precision import mixed_precision_quantization_config
21
19
  from model_compression_toolkit.core.common.quantization.quantization_config import QuantizationConfig, \
@@ -36,25 +34,21 @@ from model_compression_toolkit.core.common import network_editors as network_edi
36
34
  from model_compression_toolkit.core.keras.quantization_facade import keras_post_training_quantization, \
37
35
  keras_post_training_quantization_mixed_precision
38
36
  from model_compression_toolkit.ptq.keras.quantization_facade import keras_post_training_quantization_experimental
39
- from model_compression_toolkit.gptq.keras.quantization_facade import \
40
- keras_gradient_post_training_quantization_experimental
41
- from model_compression_toolkit.gptq.keras.quantization_facade import get_keras_gptq_config
42
- from model_compression_toolkit.qat.keras.quantization_facade import keras_quantization_aware_training_init, \
43
- keras_quantization_aware_training_finalize
44
- from model_compression_toolkit.qat.pytorch.quantization_facade import pytorch_quantization_aware_training_init, \
45
- pytorch_quantization_aware_training_finalize
46
- from model_compression_toolkit.core.pytorch.quantization_facade import pytorch_post_training_quantization, \
47
- pytorch_post_training_quantization_mixed_precision
37
+ from model_compression_toolkit.qat.keras.quantization_facade import keras_quantization_aware_training_init, keras_quantization_aware_training_finalize
38
+ from model_compression_toolkit.qat.pytorch.quantization_facade import pytorch_quantization_aware_training_init, pytorch_quantization_aware_training_finalize
39
+ from model_compression_toolkit.core.pytorch.quantization_facade import pytorch_post_training_quantization, pytorch_post_training_quantization_mixed_precision
48
40
  from model_compression_toolkit.ptq.pytorch.quantization_facade import pytorch_post_training_quantization_experimental
49
- from model_compression_toolkit.gptq.pytorch.quantization_facade import \
50
- pytorch_gradient_post_training_quantization_experimental
51
- from model_compression_toolkit.gptq.pytorch.quantization_facade import get_pytorch_gptq_config
52
41
 
53
42
  from model_compression_toolkit.core.keras.kpi_data_facade import keras_kpi_data, keras_kpi_data_experimental
54
43
  from model_compression_toolkit.core.pytorch.kpi_data_facade import pytorch_kpi_data, pytorch_kpi_data_experimental
55
44
 
56
45
  from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.load_model import keras_load_quantized_model
57
46
 
58
- from model_compression_toolkit.exporter.model_exporter import tflite_export_model, TFLiteExportMode, keras_export_model, KerasExportMode, pytorch_export_model, PyTorchExportMode
47
+
48
+ from model_compression_toolkit import exporter
49
+
50
+ from model_compression_toolkit import gptq
51
+ from model_compression_toolkit.gptq import GradientPTQConfig
52
+
59
53
 
60
54
  __version__ = "1.8.0"
model_compression_toolkit/core/common/logger.py CHANGED
@@ -17,7 +17,6 @@
17
17
  import logging
18
18
  import os
19
19
  from datetime import datetime
20
- from os import path
21
20
  from pathlib import Path
22
21
 
23
22
  LOGGER_NAME = 'Constrained Model Optimization'
@@ -43,7 +42,7 @@ class Logger:
43
42
 
44
43
  """
45
44
 
46
- if not path.exists(log_path):
45
+ if not os.path.exists(log_path):
47
46
  Path(log_path).mkdir(parents=True, exist_ok=True)
48
47
 
49
48
  @staticmethod
@@ -93,6 +92,15 @@ class Logger:
93
92
 
94
93
  print(f'log file is in {log_name}')
95
94
 
95
+ @staticmethod
96
+ def shutdown():
97
+ """
98
+ An orderly command to shutdown by flushing and closing all logging handlers.
99
+
100
+ """
101
+ Logger.LOG_PATH = None
102
+ logging.shutdown()
103
+
96
104
  ########################################
97
105
  # Delegating methods to wrapped logger
98
106
  ########################################
model_compression_toolkit/core/keras/back2framework/model_gradients.py CHANGED
@@ -171,8 +171,9 @@ def keras_iterative_approx_jacobian_trace(graph_float: common.Graph,
171
171
 
172
172
  # If the change to the mean Jacobian approximation is insignificant we stop the calculation
173
173
  if j > MIN_JACOBIANS_ITER:
174
- delta = np.mean([jac_trace_approx, *trace_jv]) - np.mean(trace_jv)
175
- if np.abs(delta) / (np.abs(np.mean(trace_jv)) + 1e-6) < JACOBIANS_COMP_TOLERANCE:
174
+ new_mean = np.mean([jac_trace_approx, *trace_jv])
175
+ delta = new_mean - np.mean(trace_jv)
176
+ if np.abs(delta) / (np.abs(new_mean) + 1e-6) < JACOBIANS_COMP_TOLERANCE:
176
177
  trace_jv.append(jac_trace_approx)
177
178
  break
178
179
 
model_compression_toolkit/core/keras/quantization_facade.py CHANGED
@@ -19,7 +19,7 @@ from model_compression_toolkit.core import common
19
19
  from model_compression_toolkit.core.common import Logger
20
20
  from model_compression_toolkit.core.common.constants import TENSORFLOW
21
21
  from model_compression_toolkit.core.common.user_info import UserInformation
22
- from model_compression_toolkit.gptq.common.gptq_config import GradientPTQConfig, GradientPTQConfigV2
22
+ from model_compression_toolkit.gptq import GradientPTQConfig, GradientPTQConfigV2
23
23
  from model_compression_toolkit.core.common.mixed_precision.kpi_tools.kpi import KPI
24
24
  from model_compression_toolkit.core.common.framework_info import FrameworkInfo
25
25
  from model_compression_toolkit.core.common.network_editors.actions import EditRule
model_compression_toolkit/core/pytorch/back2framework/model_gradients.py CHANGED
@@ -26,8 +26,9 @@ from model_compression_toolkit.core.common.constants import EPS, MIN_JACOBIANS_I
26
26
  from model_compression_toolkit.core.common.graph.edge import EDGE_SINK_INDEX
27
27
  from model_compression_toolkit.core.common.graph.functional_node import FunctionalNode
28
28
  from model_compression_toolkit.core.pytorch.back2framework.instance_builder import node_builder
29
- from model_compression_toolkit.core.pytorch.reader.node_holders import DummyPlaceHolder
30
- from model_compression_toolkit.core.pytorch.utils import torch_tensor_to_numpy
29
+ from model_compression_toolkit.core.pytorch.constants import BUFFER
30
+ from model_compression_toolkit.core.pytorch.reader.node_holders import DummyPlaceHolder, BufferHolder
31
+ from model_compression_toolkit.core.pytorch.utils import torch_tensor_to_numpy, get_working_device
31
32
  from model_compression_toolkit.core.common.logger import Logger
32
33
 
33
34
 
@@ -133,7 +134,13 @@ class PytorchModelGradients(torch.nn.Module):
133
134
 
134
135
  for n in self.node_sort:
135
136
  if not isinstance(n, FunctionalNode):
136
- self.add_module(n.name, node_builder(n))
137
+ if n.type == BufferHolder:
138
+ self.add_module(n.name, node_builder(n))
139
+ self.get_submodule(n.name). \
140
+ register_buffer(n.name,
141
+ torch.Tensor(n.get_weights_by_keys(BUFFER)).to(get_working_device()))
142
+ else:
143
+ self.add_module(n.name, node_builder(n))
137
144
 
138
145
  def forward(self,
139
146
  *args: Any) -> Any:
@@ -289,9 +296,9 @@ def pytorch_iterative_approx_jacobian_trace(graph_float: common.Graph,
289
296
 
290
297
  # If the change to the mean Jacobian approximation is insignificant we stop the calculation
291
298
  if j > MIN_JACOBIANS_ITER:
292
- delta = torch.mean(torch.stack([jac_trace_approx, *trace_jv])) - torch.mean(
293
- torch.stack(trace_jv))
294
- if torch.abs(delta) / (torch.abs(torch.mean(torch.stack(trace_jv))) + 1e-6) < JACOBIANS_COMP_TOLERANCE:
299
+ new_mean = torch.mean(torch.stack([jac_trace_approx, *trace_jv]))
300
+ delta = new_mean - torch.mean(torch.stack(trace_jv))
301
+ if torch.abs(delta) / (torch.abs(new_mean) + 1e-6) < JACOBIANS_COMP_TOLERANCE:
295
302
  trace_jv.append(jac_trace_approx)
296
303
  break
297
304
 
model_compression_toolkit/core/pytorch/constants.py CHANGED
@@ -92,3 +92,7 @@ IN_PROJ_WEIGHT = 'in_proj_weight'
92
92
  IN_PROJ_BIAS = 'in_proj_bias'
93
93
  BIAS_K = 'bias_k'
94
94
  BIAS_V = 'bias_v'
95
+
96
+ # # Batch size value for 'reshape' and 'view' operators,
97
+ # # the value is -1 so the batch size is inferred from the length of the array and remaining dimensions.
98
+ BATCH_DIM_VALUE = -1
model_compression_toolkit/core/pytorch/graph_substitutions/substitutions/reshape_with_static_shapes.py CHANGED
@@ -14,10 +14,13 @@
14
14
  # ==============================================================================
15
15
  from torch import reshape
16
16
  import torch
17
+
18
+ from model_compression_toolkit.core.common import Logger
17
19
  from model_compression_toolkit.core.common.graph.graph_matchers import NodeOperationMatcher
18
20
  from model_compression_toolkit.core import common
19
21
  from model_compression_toolkit.core.common.graph.base_graph import Graph
20
22
  from model_compression_toolkit.core.common.graph.base_node import BaseNode
23
+ from model_compression_toolkit.core.pytorch.constants import BATCH_DIM_VALUE
21
24
 
22
25
 
23
26
  class ReshapeWithStaticShapes(common.BaseSubstitution):
@@ -47,14 +50,25 @@ class ReshapeWithStaticShapes(common.BaseSubstitution):
47
50
  Returns:
48
51
  Graph after applying the substitution.
49
52
  """
53
+ # we want the batch size value to infer from the length of the array and remaining dimensions
54
+ if len(node.output_shape) == 1:
55
+ node.output_shape[0][0] = BATCH_DIM_VALUE
56
+ else:
57
+ Logger.error('Reshape or view nodes should have a single output shape') # pragma: no cover
58
+
50
59
  # configure the new static output shape attribute
51
60
  node.op_call_args = node.output_shape
52
61
 
53
62
  # modify the node input info
54
63
  node.input_shape = [node.input_shape[0]]
64
+
65
+ # the first input is the tensor to be reshaped, we want his batch size value to infer
66
+ # from the length of the array and remaining dimensions
67
+ node.input_shape[0][0] = BATCH_DIM_VALUE
68
+
55
69
  nodes_to_check = []
56
70
  for in_edge in graph.incoming_edges(node):
57
- if in_edge.sink_index > 0: # the first input is the tensor to be reshaped
71
+ if in_edge.sink_index > 0: # the first input is the tensor to be reshaped
58
72
  nodes_to_check.append(in_edge.source_node)
59
73
  graph.remove_edge(in_edge.source_node, node)
60
74
  for n in nodes_to_check:
@@ -80,4 +94,4 @@ def clean_graph_from_nodes_without_out_edges(graph: Graph,
80
94
  graph.remove_edge(in_edge.source_node, node)
81
95
  graph.remove_node(node)
82
96
  for n in nodes_to_check:
83
- clean_graph_from_nodes_without_out_edges(graph, n)
97
+ clean_graph_from_nodes_without_out_edges(graph, n)
model_compression_toolkit/exporter/__init__.py CHANGED
@@ -12,3 +12,8 @@
12
12
  # See the License for the specific language governing permissions and
13
13
  # limitations under the License.
14
14
  # ==============================================================================
15
+
16
+ from model_compression_toolkit.exporter.model_exporter.keras.keras_export_facade import keras_export_model, KerasExportMode
17
+ from model_compression_toolkit.exporter.model_exporter.pytorch.pytorch_export_facade import PyTorchExportMode, pytorch_export_model
18
+ from model_compression_toolkit.exporter.model_exporter.tflite.tflite_export_facade import tflite_export_model, TFLiteExportMode
19
+
model_compression_toolkit/exporter/model_exporter/__init__.py CHANGED
@@ -13,6 +13,3 @@
13
13
  # limitations under the License.
14
14
  # ==============================================================================
15
15
 
16
- from model_compression_toolkit.exporter.model_exporter.keras.keras_export_facade import keras_export_model, KerasExportMode
17
- from model_compression_toolkit.exporter.model_exporter.pytorch.pytorch_export_facade import PyTorchExportMode, pytorch_export_model
18
- from model_compression_toolkit.exporter.model_exporter.tflite.tflite_export_facade import tflite_export_model, TFLiteExportMode
model_compression_toolkit/exporter/model_exporter/tflite/fakely_quant_tflite_exporter.py CHANGED
@@ -19,7 +19,7 @@ from typing import Callable
19
19
  import keras.models
20
20
  import tensorflow as tf
21
21
 
22
- from model_compression_toolkit import keras_load_quantized_model
22
+ from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.load_model import keras_load_quantized_model
23
23
  from model_compression_toolkit.core.common import Logger
24
24
  from model_compression_toolkit.exporter.model_exporter.keras.fakely_quant_keras_exporter import FakelyQuantKerasExporter
25
25
 
model_compression_toolkit/exporter/model_wrapper/__init__.py CHANGED
@@ -13,12 +13,8 @@
13
13
  # limitations under the License.
14
14
  # ==============================================================================
15
15
 
16
- from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH
16
+ from model_compression_toolkit.exporter.model_wrapper.keras.validate_layer import is_keras_layer_exportable
17
+ from model_compression_toolkit.exporter.model_wrapper.keras.builder.fully_quantized_model_builder import get_exportable_keras_model
17
18
 
18
- if FOUND_TF:
19
- from model_compression_toolkit.exporter.model_wrapper.keras.validate_layer import is_keras_layer_exportable
20
- from model_compression_toolkit.exporter.model_wrapper.keras.builder.fully_quantized_model_builder import get_exportable_keras_model
21
-
22
- if FOUND_TORCH:
23
- from model_compression_toolkit.exporter.model_wrapper.pytorch.validate_layer import is_pytorch_layer_exportable
24
- from model_compression_toolkit.exporter.model_wrapper.pytorch.builder.fully_quantized_model_builder import get_exportable_pytorch_model
19
+ from model_compression_toolkit.exporter.model_wrapper.pytorch.validate_layer import is_pytorch_layer_exportable
20
+ from model_compression_toolkit.exporter.model_wrapper.pytorch.builder.fully_quantized_model_builder import get_exportable_pytorch_model