INFO:nncf:Not adding activation input quantizer for operation: 7 BertForSequenceClassification/BertModel[bert]/BertEmbeddings[embeddings]/NNCFEmbedding[position_embeddings]/embedding_0
INFO:nncf:Not adding activation input quantizer for operation: 4 BertForSequenceClassification/BertModel[bert]/BertEmbeddings[embeddings]/NNCFEmbedding[word_embeddings]/embedding_0
INFO:nncf:Not adding activation input quantizer for operation: 5 BertForSequenceClassification/BertModel[bert]/BertEmbeddings[embeddings]/NNCFEmbedding[token_type_embeddings]/embedding_0
INFO:nncf:Not adding activation input quantizer for operation: 6 BertForSequenceClassification/BertModel[bert]/BertEmbeddings[embeddings]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 8 BertForSequenceClassification/BertModel[bert]/BertEmbeddings[embeddings]/__iadd___0
INFO:nncf:Not adding activation input quantizer for operation: 9 BertForSequenceClassification/BertModel[bert]/BertEmbeddings[embeddings]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 10 BertForSequenceClassification/BertModel[bert]/BertEmbeddings[embeddings]/Dropout[dropout]/dropout_0
INFO:nncf:Not adding activation input quantizer for operation: 23 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 26 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertAttention[attention]/BertSelfAttention[self]/matmul_1
INFO:nncf:Not adding activation input quantizer for operation: 32 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertAttention[attention]/BertSelfOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 33 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertAttention[attention]/BertSelfOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 37 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 38 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[0]/BertOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 51 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 54 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertAttention[attention]/BertSelfAttention[self]/matmul_1
INFO:nncf:Not adding activation input quantizer for operation: 60 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertAttention[attention]/BertSelfOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 61 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertAttention[attention]/BertSelfOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 65 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 66 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[1]/BertOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 79 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 82 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertAttention[attention]/BertSelfAttention[self]/matmul_1
INFO:nncf:Not adding activation input quantizer for operation: 88 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertAttention[attention]/BertSelfOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 89 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertAttention[attention]/BertSelfOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 93 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 94 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[2]/BertOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 107 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 110 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertAttention[attention]/BertSelfAttention[self]/matmul_1
INFO:nncf:Not adding activation input quantizer for operation: 116 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertAttention[attention]/BertSelfOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 117 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertAttention[attention]/BertSelfOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 121 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 122 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[3]/BertOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 135 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 138 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertAttention[attention]/BertSelfAttention[self]/matmul_1
INFO:nncf:Not adding activation input quantizer for operation: 144 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertAttention[attention]/BertSelfOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 145 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertAttention[attention]/BertSelfOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 149 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 150 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[4]/BertOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 163 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 166 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertAttention[attention]/BertSelfAttention[self]/matmul_1
INFO:nncf:Not adding activation input quantizer for operation: 172 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertAttention[attention]/BertSelfOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 173 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertAttention[attention]/BertSelfOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Not adding activation input quantizer for operation: 177 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertOutput[output]/__add___0
INFO:nncf:Not adding activation input quantizer for operation: 178 BertForSequenceClassification/BertModel[bert]/BertEncoder[encoder]/ModuleList[layer]/BertLayer[5]/BertOutput[output]/NNCFLayerNorm[LayerNorm]/layer_norm_0
INFO:nncf:Collecting tensor statistics |█████           | 1 / 3
INFO:nncf:Collecting tensor statistics |██████████      | 2 / 3
INFO:nncf:Collecting tensor statistics |████████████████| 3 / 3
INFO:nncf:Compiling and loading torch extension: quantized_functions_cuda...
INFO:nncf:Finished loading torch extension: quantized_functions_cuda
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
INFO:nncf:Statistics of the quantization algorithm:
Epoch 0 |+--------------------------------+-------+
Epoch 0 ||        Statistic's name        | Value |
Epoch 0 |+================================+=======+
Epoch 0 || Ratio of enabled quantizations | 100   |
Epoch 0 |+--------------------------------+-------+
Epoch 0 |
Epoch 0 |Statistics of the quantization share:
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 ||         Statistic's name         |       Value        |
Epoch 0 |+==================================+====================+
Epoch 0 || Symmetric WQs / All placed WQs   | 100.00 % (38 / 38) |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Asymmetric WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Signed WQs / All placed WQs      | 100.00 % (38 / 38) |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Unsigned WQs / All placed WQs    | 0.00 % (0 / 38)    |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Per-tensor WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Per-channel WQs / All placed WQs | 100.00 % (38 / 38) |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Placed WQs / Potential WQs       | 70.37 % (38 / 54)  |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Symmetric AQs / All placed AQs   | 27.27 % (12 / 44)  |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Asymmetric AQs / All placed AQs  | 72.73 % (32 / 44)  |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Signed AQs / All placed AQs      | 100.00 % (44 / 44) |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Unsigned AQs / All placed AQs    | 0.00 % (0 / 44)    |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Per-tensor AQs / All placed AQs  | 100.00 % (44 / 44) |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 || Per-channel AQs / All placed AQs | 0.00 % (0 / 44)    |
Epoch 0 |+----------------------------------+--------------------+
Epoch 0 |
Epoch 0 |Statistics of the bitwidth distribution:
Epoch 0 |+--------------+---------------------+--------------------+--------------------+
Epoch 0 || Num bits (N) | N-bits WQs / Placed |    N-bits AQs /    | N-bits Qs / Placed |
Epoch 0 ||              |         WQs         |     Placed AQs     |         Qs         |
Epoch 0 |+==============+=====================+====================+====================+
Epoch 0 || 8            | 100.00 % (38 / 38)  | 100.00 % (44 / 44) | 100.00 % (82 / 82) |
Epoch 0 |+--------------+---------------------+--------------------+--------------------+
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
INFO:nncf:Statistics of the quantization algorithm:
Epoch 1 |+--------------------------------+-------+
Epoch 1 ||        Statistic's name        | Value |
Epoch 1 |+================================+=======+
Epoch 1 || Ratio of enabled quantizations | 100   |
Epoch 1 |+--------------------------------+-------+
Epoch 1 |
Epoch 1 |Statistics of the quantization share:
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 ||         Statistic's name         |       Value        |
Epoch 1 |+==================================+====================+
Epoch 1 || Symmetric WQs / All placed WQs   | 100.00 % (38 / 38) |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Asymmetric WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Signed WQs / All placed WQs      | 100.00 % (38 / 38) |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Unsigned WQs / All placed WQs    | 0.00 % (0 / 38)    |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Per-tensor WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Per-channel WQs / All placed WQs | 100.00 % (38 / 38) |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Placed WQs / Potential WQs       | 70.37 % (38 / 54)  |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Symmetric AQs / All placed AQs   | 27.27 % (12 / 44)  |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Asymmetric AQs / All placed AQs  | 72.73 % (32 / 44)  |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Signed AQs / All placed AQs      | 100.00 % (44 / 44) |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Unsigned AQs / All placed AQs    | 0.00 % (0 / 44)    |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Per-tensor AQs / All placed AQs  | 100.00 % (44 / 44) |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 || Per-channel AQs / All placed AQs | 0.00 % (0 / 44)    |
Epoch 1 |+----------------------------------+--------------------+
Epoch 1 |
Epoch 1 |Statistics of the bitwidth distribution:
Epoch 1 |+--------------+---------------------+--------------------+--------------------+
Epoch 1 || Num bits (N) | N-bits WQs / Placed |    N-bits AQs /    | N-bits Qs / Placed |
Epoch 1 ||              |         WQs         |     Placed AQs     |         Qs         |
Epoch 1 |+==============+=====================+====================+====================+
Epoch 1 || 8            | 100.00 % (38 / 38)  | 100.00 % (44 / 44) | 100.00 % (82 / 82) |
Epoch 1 |+--------------+---------------------+--------------------+--------------------+
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
INFO:nncf:Statistics of the quantization algorithm:
Epoch 2 |+--------------------------------+-------+
Epoch 2 ||        Statistic's name        | Value |
Epoch 2 |+================================+=======+
Epoch 2 || Ratio of enabled quantizations | 100   |
Epoch 2 |+--------------------------------+-------+
Epoch 2 |
Epoch 2 |Statistics of the quantization share:
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 ||         Statistic's name         |       Value        |
Epoch 2 |+==================================+====================+
Epoch 2 || Symmetric WQs / All placed WQs   | 100.00 % (38 / 38) |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Asymmetric WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Signed WQs / All placed WQs      | 100.00 % (38 / 38) |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Unsigned WQs / All placed WQs    | 0.00 % (0 / 38)    |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Per-tensor WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Per-channel WQs / All placed WQs | 100.00 % (38 / 38) |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Placed WQs / Potential WQs       | 70.37 % (38 / 54)  |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Symmetric AQs / All placed AQs   | 27.27 % (12 / 44)  |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Asymmetric AQs / All placed AQs  | 72.73 % (32 / 44)  |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Signed AQs / All placed AQs      | 100.00 % (44 / 44) |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Unsigned AQs / All placed AQs    | 0.00 % (0 / 44)    |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Per-tensor AQs / All placed AQs  | 100.00 % (44 / 44) |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 || Per-channel AQs / All placed AQs | 0.00 % (0 / 44)    |
Epoch 2 |+----------------------------------+--------------------+
Epoch 2 |
Epoch 2 |Statistics of the bitwidth distribution:
Epoch 2 |+--------------+---------------------+--------------------+--------------------+
Epoch 2 || Num bits (N) | N-bits WQs / Placed |    N-bits AQs /    | N-bits Qs / Placed |
Epoch 2 ||              |         WQs         |     Placed AQs     |         Qs         |
Epoch 2 |+==============+=====================+====================+====================+
Epoch 2 || 8            | 100.00 % (38 / 38)  | 100.00 % (44 / 44) | 100.00 % (82 / 82) |
Epoch 2 |+--------------+---------------------+--------------------+--------------------+
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
INFO:nncf:Statistics of the quantization algorithm:
Epoch 3 |+--------------------------------+-------+
Epoch 3 ||        Statistic's name        | Value |
Epoch 3 |+================================+=======+
Epoch 3 || Ratio of enabled quantizations | 100   |
Epoch 3 |+--------------------------------+-------+
Epoch 3 |
Epoch 3 |Statistics of the quantization share:
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 ||         Statistic's name         |       Value        |
Epoch 3 |+==================================+====================+
Epoch 3 || Symmetric WQs / All placed WQs   | 100.00 % (38 / 38) |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Asymmetric WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Signed WQs / All placed WQs      | 100.00 % (38 / 38) |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Unsigned WQs / All placed WQs    | 0.00 % (0 / 38)    |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Per-tensor WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Per-channel WQs / All placed WQs | 100.00 % (38 / 38) |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Placed WQs / Potential WQs       | 70.37 % (38 / 54)  |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Symmetric AQs / All placed AQs   | 27.27 % (12 / 44)  |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Asymmetric AQs / All placed AQs  | 72.73 % (32 / 44)  |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Signed AQs / All placed AQs      | 100.00 % (44 / 44) |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Unsigned AQs / All placed AQs    | 0.00 % (0 / 44)    |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Per-tensor AQs / All placed AQs  | 100.00 % (44 / 44) |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 || Per-channel AQs / All placed AQs | 0.00 % (0 / 44)    |
Epoch 3 |+----------------------------------+--------------------+
Epoch 3 |
Epoch 3 |Statistics of the bitwidth distribution:
Epoch 3 |+--------------+---------------------+--------------------+--------------------+
Epoch 3 || Num bits (N) | N-bits WQs / Placed |    N-bits AQs /    | N-bits Qs / Placed |
Epoch 3 ||              |         WQs         |     Placed AQs     |         Qs         |
Epoch 3 |+==============+=====================+====================+====================+
Epoch 3 || 8            | 100.00 % (38 / 38)  | 100.00 % (44 / 44) | 100.00 % (82 / 82) |
Epoch 3 |+--------------+---------------------+--------------------+--------------------+
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
INFO:nncf:Statistics of the quantization algorithm:
Epoch 4 |+--------------------------------+-------+
Epoch 4 ||        Statistic's name        | Value |
Epoch 4 |+================================+=======+
Epoch 4 || Ratio of enabled quantizations | 100   |
Epoch 4 |+--------------------------------+-------+
Epoch 4 |
Epoch 4 |Statistics of the quantization share:
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 ||         Statistic's name         |       Value        |
Epoch 4 |+==================================+====================+
Epoch 4 || Symmetric WQs / All placed WQs   | 100.00 % (38 / 38) |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Asymmetric WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Signed WQs / All placed WQs      | 100.00 % (38 / 38) |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Unsigned WQs / All placed WQs    | 0.00 % (0 / 38)    |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Per-tensor WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Per-channel WQs / All placed WQs | 100.00 % (38 / 38) |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Placed WQs / Potential WQs       | 70.37 % (38 / 54)  |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Symmetric AQs / All placed AQs   | 27.27 % (12 / 44)  |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Asymmetric AQs / All placed AQs  | 72.73 % (32 / 44)  |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Signed AQs / All placed AQs      | 100.00 % (44 / 44) |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Unsigned AQs / All placed AQs    | 0.00 % (0 / 44)    |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Per-tensor AQs / All placed AQs  | 100.00 % (44 / 44) |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 || Per-channel AQs / All placed AQs | 0.00 % (0 / 44)    |
Epoch 4 |+----------------------------------+--------------------+
Epoch 4 |
Epoch 4 |Statistics of the bitwidth distribution:
Epoch 4 |+--------------+---------------------+--------------------+--------------------+
Epoch 4 || Num bits (N) | N-bits WQs / Placed |    N-bits AQs /    | N-bits Qs / Placed |
Epoch 4 ||              |         WQs         |     Placed AQs     |         Qs         |
Epoch 4 |+==============+=====================+====================+====================+
Epoch 4 || 8            | 100.00 % (38 / 38)  | 100.00 % (44 / 44) | 100.00 % (82 / 82) |
Epoch 4 |+--------------+---------------------+--------------------+--------------------+
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
INFO:nncf:Statistics of the quantization algorithm:
Epoch 5 |+--------------------------------+-------+
Epoch 5 ||        Statistic's name        | Value |
Epoch 5 |+================================+=======+
Epoch 5 || Ratio of enabled quantizations | 100   |
Epoch 5 |+--------------------------------+-------+
Epoch 5 |
Epoch 5 |Statistics of the quantization share:
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 ||         Statistic's name         |       Value        |
Epoch 5 |+==================================+====================+
Epoch 5 || Symmetric WQs / All placed WQs   | 100.00 % (38 / 38) |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Asymmetric WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Signed WQs / All placed WQs      | 100.00 % (38 / 38) |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Unsigned WQs / All placed WQs    | 0.00 % (0 / 38)    |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Per-tensor WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Per-channel WQs / All placed WQs | 100.00 % (38 / 38) |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Placed WQs / Potential WQs       | 70.37 % (38 / 54)  |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Symmetric AQs / All placed AQs   | 27.27 % (12 / 44)  |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Asymmetric AQs / All placed AQs  | 72.73 % (32 / 44)  |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Signed AQs / All placed AQs      | 100.00 % (44 / 44) |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Unsigned AQs / All placed AQs    | 0.00 % (0 / 44)    |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Per-tensor AQs / All placed AQs  | 100.00 % (44 / 44) |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 || Per-channel AQs / All placed AQs | 0.00 % (0 / 44)    |
Epoch 5 |+----------------------------------+--------------------+
Epoch 5 |
Epoch 5 |Statistics of the bitwidth distribution:
Epoch 5 |+--------------+---------------------+--------------------+--------------------+
Epoch 5 || Num bits (N) | N-bits WQs / Placed |    N-bits AQs /    | N-bits Qs / Placed |
Epoch 5 ||              |         WQs         |     Placed AQs     |         Qs         |
Epoch 5 |+==============+=====================+====================+====================+
Epoch 5 || 8            | 100.00 % (38 / 38)  | 100.00 % (44 / 44) | 100.00 % (82 / 82) |
Epoch 5 |+--------------+---------------------+--------------------+--------------------+
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
INFO:nncf:Statistics of the quantization algorithm:
Epoch 6 |+--------------------------------+-------+
Epoch 6 ||        Statistic's name        | Value |
Epoch 6 |+================================+=======+
Epoch 6 || Ratio of enabled quantizations | 100   |
Epoch 6 |+--------------------------------+-------+
Epoch 6 |
Epoch 6 |Statistics of the quantization share:
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 ||         Statistic's name         |       Value        |
Epoch 6 |+==================================+====================+
Epoch 6 || Symmetric WQs / All placed WQs   | 100.00 % (38 / 38) |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Asymmetric WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Signed WQs / All placed WQs      | 100.00 % (38 / 38) |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Unsigned WQs / All placed WQs    | 0.00 % (0 / 38)    |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Per-tensor WQs / All placed WQs  | 0.00 % (0 / 38)    |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Per-channel WQs / All placed WQs | 100.00 % (38 / 38) |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Placed WQs / Potential WQs       | 70.37 % (38 / 54)  |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Symmetric AQs / All placed AQs   | 27.27 % (12 / 44)  |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Asymmetric AQs / All placed AQs  | 72.73 % (32 / 44)  |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Signed AQs / All placed AQs      | 100.00 % (44 / 44) |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Unsigned AQs / All placed AQs    | 0.00 % (0 / 44)    |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Per-tensor AQs / All placed AQs  | 100.00 % (44 / 44) |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 || Per-channel AQs / All placed AQs | 0.00 % (0 / 44)    |
Epoch 6 |+----------------------------------+--------------------+
Epoch 6 |
Epoch 6 |Statistics of the bitwidth distribution:
Epoch 6 |+--------------+---------------------+--------------------+--------------------+
Epoch 6 || Num bits (N) | N-bits WQs / Placed |    N-bits AQs /    | N-bits Qs / Placed |
Epoch 6 ||              |         WQs         |     Placed AQs     |         Qs         |
Epoch 6 |+==============+=====================+====================+====================+
Epoch 6 || 8            | 100.00 % (38 / 38)  | 100.00 % (44 / 44) | 100.00 % (82 / 82) |
Epoch 6 |+--------------+---------------------+--------------------+--------------------+
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.
WARNING:nncf:You are setting `forward` on an NNCF-processed model object.
NNCF relies on custom-wrapping the `forward` call in order to function properly.
Arbitrary adjustments to the forward function on an NNCFNetwork object have undefined behavior.
If you need to replace the underlying forward function of the original model so that NNCF should be using that instead of the original forward function that NNCF saved during the compressed model creation, you can do this by calling:
model.nncf.set_original_unbound_forward(fn)
if `fn` has an unbound 0-th `self` argument, or
with model.nncf.temporary_bound_original_forward(fn): ...
if `fn` already had 0-th `self` argument bound or never had it in the first place.