mct-nightly 2.4.0.20250707.643__py3-none-any.whl → 2.4.0.20250708.612__py3-none-any.whl

{mct_nightly-2.4.0.20250707.643.dist-info → mct_nightly-2.4.0.20250708.612.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mct-nightly
-Version: 2.4.0.20250707.643
+Version: 2.4.0.20250708.612
 Summary: A Model Compression Toolkit for neural networks
 Author-email: ssi-dnn-dev@sony.com
 Classifier: Programming Language :: Python :: 3

{mct_nightly-2.4.0.20250707.643.dist-info → mct_nightly-2.4.0.20250708.612.dist-info}/RECORD

@@ -1,5 +1,5 @@
-mct_nightly-2.4.0.20250707.643.dist-info/licenses/LICENSE.md,sha256=aYSSIb-5AFPeITTvXm1UAoe0uYBiMmSS8flvXaaFUks,10174
-model_compression_toolkit/__init__.py,sha256=n67zJIaNzqdkUrpJl9e_iXZ7xM4vgLz7U7d06AHkmTU,1557
+mct_nightly-2.4.0.20250708.612.dist-info/licenses/LICENSE.md,sha256=aYSSIb-5AFPeITTvXm1UAoe0uYBiMmSS8flvXaaFUks,10174
+model_compression_toolkit/__init__.py,sha256=TrwEn4n4YKNNqTg96Ud45uv2qh_Ob-Q1IzRMrsqjJG4,1557
 model_compression_toolkit/constants.py,sha256=KNgiNLpsMgSYyXMNEbHXd4bFNerQc1D6HH3vpbUq_Gs,4086
 model_compression_toolkit/defaultdict.py,sha256=LSc-sbZYXENMCw3U9F4GiXuv67IKpdn0Qm7Fr11jy-4,2277
 model_compression_toolkit/logger.py,sha256=L3q7tn3Uht0i_7phnlOWMR2Te2zvzrt2HOz9vYEInts,4529
@@ -16,7 +16,7 @@ model_compression_toolkit/core/common/framework_implementation.py,sha256=jrTupZb
 model_compression_toolkit/core/common/framework_info.py,sha256=vPGV28gm-kvNSkkWI6jY3YeKBUYmn6UQ98HVUnl_-tM,5449
 model_compression_toolkit/core/common/memory_computation.py,sha256=ixoSpV5ZYZGyzhre3kQcvR2sNA8KBsPZ3lgbkDnw9Cs,1205
 model_compression_toolkit/core/common/model_builder_mode.py,sha256=jll9-59OPaE3ug7Y9-lLyV99_FoNHxkGZMgcm0Vkpss,1324
-model_compression_toolkit/core/common/model_collector.py,sha256=fgSLqbi1YRyvISJIP9WyZv3IlvuMplywJ-ffUnDWN1Q,13655
+model_compression_toolkit/core/common/model_collector.py,sha256=I0O5SoUrwB045AGLleOGYVvZyR7HJx_t6vfbECahfoU,13914
 model_compression_toolkit/core/common/node_prior_info.py,sha256=WXX_PrGVG9M9I_REG5ZzFBohwmV4yf356sZnrja_FLo,2832
 model_compression_toolkit/core/common/similarity_analyzer.py,sha256=S3f6WgHyw62dGcxpX51FGKyfebe2zv9ABKbjtGyKRvY,9215
 model_compression_toolkit/core/common/user_info.py,sha256=dSRMnT-oewmdOziIpEuW-s9K7vTSeyUBxT4z9neXurI,1648
@@ -106,7 +106,7 @@ model_compression_toolkit/core/common/quantization/candidate_node_quantization_c
 model_compression_toolkit/core/common/quantization/core_config.py,sha256=yxCzWqldcHoe8GGxrH0tp99bhrc5jDT7SgZftnMUUBE,2374
 model_compression_toolkit/core/common/quantization/debug_config.py,sha256=uH45Uq3Tp9FIyMynex_WY2_y-Kv8LuPw2XXZydnpW5A,1649
 model_compression_toolkit/core/common/quantization/filter_nodes_candidates.py,sha256=FyYCYbfkAofEWO2mAvFIppPeq2I10f1ScPNiVa9F7x4,7687
-model_compression_toolkit/core/common/quantization/node_quantization_config.py,sha256=te18Lv-pwAbING6CnURANZZVeThFn3tDmNBBSXwBcuM,21306
+model_compression_toolkit/core/common/quantization/node_quantization_config.py,sha256=VRY_HxZl9D77eq6oJ61eBnL4QJTG5pVLzJtAcaZATRQ,21636
 model_compression_toolkit/core/common/quantization/quantization_config.py,sha256=EMhXeY1qkvwlMAY5fpKRDuLEAyRY5yKqe2fOmAD_rVI,4362
 model_compression_toolkit/core/common/quantization/quantization_fn_selection.py,sha256=VVq2cKjumlNWucUbaNw8s2J0IbI_vrQ-KR_eQPshGSg,3140
 model_compression_toolkit/core/common/quantization/quantize_graph_weights.py,sha256=N005MSvx8UypVpa7XrxNrB2G732n2wHj3RmLyjTgd3I,2728
@@ -117,8 +117,8 @@ model_compression_toolkit/core/common/quantization/quantization_params_generatio
 model_compression_toolkit/core/common/quantization/quantization_params_generation/lut_kmeans_params.py,sha256=RL-PklAjGyC-26anSt8fU07a6pB_LBQFQy9o4e9giN0,8739
 model_compression_toolkit/core/common/quantization/quantization_params_generation/outlier_filter.py,sha256=9gnfJV89jpGwAx8ImJ5E9NjCv3lDtbyulP4OtgWb62M,1772
 model_compression_toolkit/core/common/quantization/quantization_params_generation/power_of_two_selection.py,sha256=-cghHF5S11qbjTDRruHlc__uaDoofZHl7QTl8hCeKW0,11141
-model_compression_toolkit/core/common/quantization/quantization_params_generation/qparams_activations_computation.py,sha256=KoPYo6il5Pa5Y54KUa1GiB5G_hGzN8yMTgLOsC6QElY,7650
-model_compression_toolkit/core/common/quantization/quantization_params_generation/qparams_computation.py,sha256=bsVCs5KrVKtagHG4VLFzxN2f-YKs0-mhV9VmL57SE9E,7924
+model_compression_toolkit/core/common/quantization/quantization_params_generation/qparams_activations_computation.py,sha256=vOFquQ3_h0jyG8EEeHM8M57uIKMZQ7iobUgbvWbiXh4,7798
+model_compression_toolkit/core/common/quantization/quantization_params_generation/qparams_computation.py,sha256=TPSkKujdJ6jB7VOVp0kRgMnthmeuaBZgN1HQuJ2pqR0,7951
 model_compression_toolkit/core/common/quantization/quantization_params_generation/qparams_search.py,sha256=Nv_b3DECVjQnlrUet2kbuSvSKVnxcc-gf2zhFb2jSZk,43482
 model_compression_toolkit/core/common/quantization/quantization_params_generation/qparams_weights_computation.py,sha256=saOQMtj1qYqgQoAFjq31p7xtiRDxmanGGdm0DE81_cg,4820
 model_compression_toolkit/core/common/quantization/quantization_params_generation/symmetric_selection.py,sha256=6tRNgWvn-4r8hiSHqND7Qms1Nje1DUR4MR0JeWCNyvI,12531
@@ -530,7 +530,7 @@ model_compression_toolkit/xquant/pytorch/model_analyzer.py,sha256=b93o800yVB3Z-i
 model_compression_toolkit/xquant/pytorch/pytorch_report_utils.py,sha256=Y0oBl8qPFsdNrK49XczwmVacInJcOPHslVnFBs-iTCc,3742
 model_compression_toolkit/xquant/pytorch/similarity_functions.py,sha256=CERxq5K8rqaiE-DlwhZBTUd9x69dtYJlkHOPLB54vm8,2354
 model_compression_toolkit/xquant/pytorch/tensorboard_utils.py,sha256=n0HvWBzkBkUJZlS3WeynhpsRTps2qQkjlq7luliBHNU,9627
-mct_nightly-2.4.0.20250707.643.dist-info/METADATA,sha256=ZQC6URpJO-t-tH6PFUmNg-09N-hh6EnJg-P8ykkesoA,25555
-mct_nightly-2.4.0.20250707.643.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-mct_nightly-2.4.0.20250707.643.dist-info/top_level.txt,sha256=gsYA8juk0Z-ZmQRKULkb3JLGdOdz8jW_cMRjisn9ga4,26
-mct_nightly-2.4.0.20250707.643.dist-info/RECORD,,
+mct_nightly-2.4.0.20250708.612.dist-info/METADATA,sha256=DXcxuSC6OdHkGc28g74YdpfdWU7tRPQY-SPtmkfmJvw,25555
+mct_nightly-2.4.0.20250708.612.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+mct_nightly-2.4.0.20250708.612.dist-info/top_level.txt,sha256=gsYA8juk0Z-ZmQRKULkb3JLGdOdz8jW_cMRjisn9ga4,26
+mct_nightly-2.4.0.20250708.612.dist-info/RECORD,,

model_compression_toolkit/__init__.py

@@ -27,4 +27,4 @@ from model_compression_toolkit import data_generation
 from model_compression_toolkit import pruning
 from model_compression_toolkit.trainable_infrastructure.keras.load_model import keras_load_quantized_model
 
-__version__ = "2.4.0.20250707.000643"
+__version__ = "2.4.0.20250708.000612"

model_compression_toolkit/core/common/model_collector.py

@@ -57,19 +57,21 @@ def create_stats_collector_for_node(node: common.BaseNode,
 
 
 def create_tensor2node(graph: common.Graph,
-                       node: common.BaseNode):
+                       node: common.BaseNode,
+                       next_node_output_channel_axis: int):
     """
     Force statistic collector creation and assignment for a node.
     Args:
         graph: Graph of the node (for retrieving the current tensor).
         node: Node to create a tensor for.
+        next_node_output_channel_axis: channel output axis of next node.
 
     """
     current_sc = graph.get_out_stats_collector(node)
     is_list_nostat_collectors = isinstance(current_sc, list) and len(
         [sc for sc in current_sc if not isinstance(sc, common.NoStatsCollector)]) == 0
     if isinstance(current_sc, common.NoStatsCollector) or current_sc is None or is_list_nostat_collectors:
-        stats_collector = common.StatsCollector(node.out_channel_axis)
+        stats_collector = common.StatsCollector(next_node_output_channel_axis if node.out_channel_axis is None else node.out_channel_axis)
         graph.set_out_stats_collector_to_node(node, stats_collector)
 
 
@@ -175,7 +177,8 @@ class ModelCollector:
                 for ie in graph.incoming_edges(n):
                     input_node = ie.source_node
                     create_tensor2node(graph,
-                                       input_node)
+                                       input_node,
+                                       n.out_channel_axis)
             if sc is not None:
                 graph.set_out_stats_collector_to_node(n, sc)
 

model_compression_toolkit/core/common/quantization/node_quantization_config.py

@@ -98,6 +98,9 @@ class NodeActivationQuantizationConfig(BaseNodeQuantizationConfig):
         self.activation_quantization_params = {}
         # TODO: computed by compute_activation_bias_correction. Probably shouldnt be here.
         self.activation_bias_correction_term = None
+        # Z-threshold is a global param from QuantizationConfig, however it can be overridden per node by NetworkEditor.
+        # Since activation qparams are re-computed in several places, it's easier to keep it here and update it once.
+        self.z_threshold = None
 
     @property
     def enable_activation_quantization(self):
@@ -119,7 +122,7 @@ class NodeActivationQuantizationConfig(BaseNodeQuantizationConfig):
             activation_params: Dictionary that contains weight quantization params.
 
         """
-        assert self.quant_mode == ActivationQuantizationMode.QUANT
+        assert self.quant_mode == ActivationQuantizationMode.QUANT or self.quant_mode == ActivationQuantizationMode.FLN_QUANT
         for param_name, param_value in activation_params.items():
             self.activation_quantization_params[param_name] = param_value
 

model_compression_toolkit/core/common/quantization/quantization_params_generation/qparams_activations_computation.py

@@ -47,9 +47,12 @@ def compute_activation_qparams(quant_cfg: QuantizationConfig,
         node_activation_quant_cfg.activation_quantization_method, no_clipping=node_prior_info.is_output_bounded())
 
     # Extract and filter histogram data from the statistics container.
+    z_threshold = quant_cfg.z_threshold
+    if node_activation_quant_cfg.z_threshold is not None:
+        z_threshold = node_activation_quant_cfg.z_threshold
     bins_values, bins_counts = _get_histogram_data(out_stats_container,
                                                    activation_error_method=quant_cfg.activation_error_method,
-                                                   z_threshold=quant_cfg.z_threshold)
+                                                   z_threshold=z_threshold)
 
     # Retrieve the minimum and maximum values from the statistics container.
     min_value, max_value = out_stats_container.get_min_max_values()

model_compression_toolkit/core/common/quantization/quantization_params_generation/qparams_computation.py

@@ -105,7 +105,7 @@ def calculate_quantization_params(graph: Graph,
                     attr_cfg.weights_channels_axis = ChannelAxisMapping(output_channels_axis, attr_cfg.weights_channels_axis.input)
                     attr_cfg.set_weights_quantization_param(weights_params)
 
-            if n.is_activation_quantization_enabled():
+            if n.is_activation_quantization_enabled() or n.is_fln_quantization():
                 # If node's activations should be quantized as well, we compute its activation quantization parameters
                 activation_params = compute_activation_qparams(quant_cfg=quant_cfg,
                                                                node_activation_quant_cfg=candidate_qc.activation_quantization_cfg,