Updated model
Browse files- README.md +37 -24
- all_results.json +10 -10
- compressed_graph.dot +0 -0
- eval_results.json +6 -6
- nncf_output.log +0 -0
- openvino_config.json +12 -9
- openvino_model.bin +2 -2
- openvino_model.xml +0 -0
- structured_sparsity.csv +68 -68
- train_results.json +5 -5
- trainer_state.json +0 -0
- training_args.bin +2 -2
README.md
CHANGED
|
@@ -8,58 +8,71 @@ datasets:
|
|
| 8 |
metrics:
|
| 9 |
- accuracy
|
| 10 |
model-index:
|
| 11 |
-
- name: wav2vec2-base-
|
| 12 |
results: []
|
| 13 |
---
|
| 14 |
|
| 15 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 16 |
should probably proofread and complete it, then remove this comment. -->
|
| 17 |
|
| 18 |
-
# wav2vec2-base-
|
| 19 |
|
| 20 |
-
This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the
|
|
|
|
| 21 |
|
| 22 |
It was compressed using [NNCF](https://github.com/openvinotoolkit/nncf) with [Optimum Intel](https://github.com/huggingface/optimum-intel#openvino) following the
|
| 23 |
JPQD image classification example.
|
| 24 |
|
|
|
|
| 25 |
It achieves the following results on the evaluation set:
|
| 26 |
-
- Loss: 0.
|
| 27 |
-
- Accuracy: 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
### Training hyperparameters
|
| 30 |
|
| 31 |
The following hyperparameters were used during training:
|
| 32 |
-
- learning_rate:
|
| 33 |
- train_batch_size: 32
|
| 34 |
- eval_batch_size: 64
|
| 35 |
-
- seed:
|
| 36 |
- gradient_accumulation_steps: 4
|
| 37 |
- total_train_batch_size: 128
|
| 38 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 39 |
- lr_scheduler_type: linear
|
| 40 |
-
- lr_scheduler_warmup_ratio: 0.
|
| 41 |
-
- num_epochs:
|
| 42 |
- mixed_precision_training: Native AMP
|
| 43 |
|
| 44 |
### Training results
|
| 45 |
|
| 46 |
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|
| 47 |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
|
| 48 |
-
| 2.
|
| 49 |
-
|
|
| 50 |
-
|
|
| 51 |
-
|
|
| 52 |
-
|
|
| 53 |
-
|
|
| 54 |
-
|
|
| 55 |
-
|
|
| 56 |
-
| 0.
|
| 57 |
-
| 0.
|
| 58 |
-
| 0.
|
| 59 |
-
| 0.
|
| 60 |
-
| 0.2492 | 13.0 | 5187 | 0.1028 | 0.9712 |
|
| 61 |
-
| 0.2658 | 14.0 | 5586 | 0.1026 | 0.9734 |
|
| 62 |
-
| 0.213 | 15.0 | 5985 | 0.0996 | 0.9734 |
|
| 63 |
|
| 64 |
|
| 65 |
### Framework versions
|
|
|
|
| 8 |
metrics:
|
| 9 |
- accuracy
|
| 10 |
model-index:
|
| 11 |
+
- name: jpqd-wav2vec2-base-ft-keyword-spotting
|
| 12 |
results: []
|
| 13 |
---
|
| 14 |
|
| 15 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
| 16 |
should probably proofread and complete it, then remove this comment. -->
|
| 17 |
|
| 18 |
+
# jpqd-wav2vec2-base-ft-keyword-spotting
|
| 19 |
|
| 20 |
+
This model is a fine-tuned version of [facebook/wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base) on the
|
| 21 |
+
superb dataset, using [superb/wav2vec2-base-superb-ks](https://huggingface.co/superb/wav2vec2-base-superb-ks) as a teacher model
|
| 22 |
|
| 23 |
It was compressed using [NNCF](https://github.com/openvinotoolkit/nncf) with [Optimum Intel](https://github.com/huggingface/optimum-intel#openvino) following the
|
| 24 |
JPQD image classification example.
|
| 25 |
|
| 26 |
+
|
| 27 |
It achieves the following results on the evaluation set:
|
| 28 |
+
- Loss: 0.5632
|
| 29 |
+
- Accuracy: 0.9756
|
| 30 |
+
|
| 31 |
+
## Model description
|
| 32 |
+
|
| 33 |
+
More information needed
|
| 34 |
+
|
| 35 |
+
## Intended uses & limitations
|
| 36 |
+
|
| 37 |
+
More information needed
|
| 38 |
+
|
| 39 |
+
## Training and evaluation data
|
| 40 |
+
|
| 41 |
+
More information needed
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
|
| 45 |
### Training hyperparameters
|
| 46 |
|
| 47 |
The following hyperparameters were used during training:
|
| 48 |
+
- learning_rate: 7e-05
|
| 49 |
- train_batch_size: 32
|
| 50 |
- eval_batch_size: 64
|
| 51 |
+
- seed: 42
|
| 52 |
- gradient_accumulation_steps: 4
|
| 53 |
- total_train_batch_size: 128
|
| 54 |
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 55 |
- lr_scheduler_type: linear
|
| 56 |
+
- lr_scheduler_warmup_ratio: 0.5
|
| 57 |
+
- num_epochs: 12.0
|
| 58 |
- mixed_precision_training: Native AMP
|
| 59 |
|
| 60 |
### Training results
|
| 61 |
|
| 62 |
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|
| 63 |
|:-------------:|:-----:|:----:|:---------------:|:--------:|
|
| 64 |
+
| 2.2245 | 1.0 | 399 | 2.2351 | 0.6209 |
|
| 65 |
+
| 6.9856 | 2.0 | 798 | 7.0597 | 0.7354 |
|
| 66 |
+
| 10.013 | 3.0 | 1197 | 9.8779 | 0.8069 |
|
| 67 |
+
| 11.3484 | 4.0 | 1596 | 11.1949 | 0.8719 |
|
| 68 |
+
| 11.6849 | 5.0 | 1995 | 11.5479 | 0.9014 |
|
| 69 |
+
| 11.5921 | 6.0 | 2394 | 11.4193 | 0.9495 |
|
| 70 |
+
| 0.8911 | 7.0 | 2793 | 0.7334 | 0.9500 |
|
| 71 |
+
| 0.8965 | 8.0 | 3192 | 0.6553 | 0.9685 |
|
| 72 |
+
| 0.7198 | 9.0 | 3591 | 0.6213 | 0.9669 |
|
| 73 |
+
| 0.7372 | 10.0 | 3990 | 0.5929 | 0.9675 |
|
| 74 |
+
| 0.7004 | 11.0 | 4389 | 0.5720 | 0.9721 |
|
| 75 |
+
| 0.6195 | 12.0 | 4788 | 0.5632 | 0.9756 |
|
|
|
|
|
|
|
|
|
|
| 76 |
|
| 77 |
|
| 78 |
### Framework versions
|
all_results.json
CHANGED
|
@@ -1,12 +1,12 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"eval_accuracy": 0.
|
| 4 |
-
"eval_loss": 0.
|
| 5 |
-
"eval_runtime":
|
| 6 |
-
"eval_samples_per_second":
|
| 7 |
-
"eval_steps_per_second": 3.
|
| 8 |
-
"train_loss":
|
| 9 |
-
"train_runtime":
|
| 10 |
-
"train_samples_per_second":
|
| 11 |
-
"train_steps_per_second": 0.
|
| 12 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 12.0,
|
| 3 |
+
"eval_accuracy": 0.9755810532509561,
|
| 4 |
+
"eval_loss": 0.5631720423698425,
|
| 5 |
+
"eval_runtime": 32.0594,
|
| 6 |
+
"eval_samples_per_second": 212.044,
|
| 7 |
+
"eval_steps_per_second": 3.338,
|
| 8 |
+
"train_loss": 4.577942104168304,
|
| 9 |
+
"train_runtime": 8786.4706,
|
| 10 |
+
"train_samples_per_second": 69.781,
|
| 11 |
+
"train_steps_per_second": 0.545
|
| 12 |
}
|
compressed_graph.dot
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
eval_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"eval_accuracy": 0.
|
| 4 |
-
"eval_loss": 0.
|
| 5 |
-
"eval_runtime":
|
| 6 |
-
"eval_samples_per_second":
|
| 7 |
-
"eval_steps_per_second": 3.
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 12.0,
|
| 3 |
+
"eval_accuracy": 0.9755810532509561,
|
| 4 |
+
"eval_loss": 0.5631720423698425,
|
| 5 |
+
"eval_runtime": 32.0594,
|
| 6 |
+
"eval_samples_per_second": 212.044,
|
| 7 |
+
"eval_steps_per_second": 3.338
|
| 8 |
}
|
nncf_output.log
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
openvino_config.json
CHANGED
|
@@ -11,9 +11,9 @@
|
|
| 11 |
],
|
| 12 |
"params": {
|
| 13 |
"enable_structured_masking": true,
|
| 14 |
-
"importance_regularization_factor": 0.
|
| 15 |
-
"warmup_end_epoch":
|
| 16 |
-
"warmup_start_epoch":
|
| 17 |
},
|
| 18 |
"sparse_structure_by_scopes": [
|
| 19 |
{
|
|
@@ -40,20 +40,23 @@
|
|
| 40 |
"algorithm": "quantization",
|
| 41 |
"export_to_onnx_standard_ops": false,
|
| 42 |
"ignored_scopes": [
|
| 43 |
-
"{re}.*feature_extractor.*",
|
| 44 |
"{re}.*__add___[0-1]",
|
| 45 |
"{re}.*layer_norm_0"
|
| 46 |
],
|
| 47 |
"initializer": {
|
| 48 |
"batchnorm_adaptation": {
|
| 49 |
-
"num_bn_adaptation_samples":
|
| 50 |
},
|
| 51 |
"range": {
|
| 52 |
-
"num_init_samples":
|
| 53 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
}
|
| 55 |
},
|
| 56 |
-
"overflow_fix": "
|
| 57 |
"preset": "mixed",
|
| 58 |
"quantize_inputs": false,
|
| 59 |
"scope_overrides": {
|
|
@@ -75,7 +78,7 @@
|
|
| 75 |
"type": "float"
|
| 76 |
}
|
| 77 |
],
|
| 78 |
-
"log_dir": "wav2vec2-base-
|
| 79 |
"optimum_version": "1.7.1",
|
| 80 |
"save_onnx_model": false,
|
| 81 |
"transformers_version": "4.26.1"
|
|
|
|
| 11 |
],
|
| 12 |
"params": {
|
| 13 |
"enable_structured_masking": true,
|
| 14 |
+
"importance_regularization_factor": 0.04,
|
| 15 |
+
"warmup_end_epoch": 6,
|
| 16 |
+
"warmup_start_epoch": 1
|
| 17 |
},
|
| 18 |
"sparse_structure_by_scopes": [
|
| 19 |
{
|
|
|
|
| 40 |
"algorithm": "quantization",
|
| 41 |
"export_to_onnx_standard_ops": false,
|
| 42 |
"ignored_scopes": [
|
|
|
|
| 43 |
"{re}.*__add___[0-1]",
|
| 44 |
"{re}.*layer_norm_0"
|
| 45 |
],
|
| 46 |
"initializer": {
|
| 47 |
"batchnorm_adaptation": {
|
| 48 |
+
"num_bn_adaptation_samples": 256
|
| 49 |
},
|
| 50 |
"range": {
|
| 51 |
+
"num_init_samples": 512,
|
| 52 |
+
"params": {
|
| 53 |
+
"max_percentile": 99.99,
|
| 54 |
+
"min_percentile": 0.01
|
| 55 |
+
},
|
| 56 |
+
"type": "percentile"
|
| 57 |
}
|
| 58 |
},
|
| 59 |
+
"overflow_fix": "disable",
|
| 60 |
"preset": "mixed",
|
| 61 |
"quantize_inputs": false,
|
| 62 |
"scope_overrides": {
|
|
|
|
| 78 |
"type": "float"
|
| 79 |
}
|
| 80 |
],
|
| 81 |
+
"log_dir": "jpqd-wav2vec2-base-ft-keyword-spotting",
|
| 82 |
"optimum_version": "1.7.1",
|
| 83 |
"save_onnx_model": false,
|
| 84 |
"transformers_version": "4.26.1"
|
openvino_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9822564c0c1a29d453de99e98f810fd4e84d24e3953c9bf6d4cc47eb836014d6
|
| 3 |
+
size 45682628
|
openvino_model.xml
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
structured_sparsity.csv
CHANGED
|
@@ -1,73 +1,73 @@
|
|
| 1 |
,group_id,type,torch_module,weight_shape,pruned_weight_shape,bias_shape,pruned_bias_shape,head_or_channel_id_to_keep,module_node_name
|
| 2 |
-
0,0,MHSA,nncf_module.wav2vec2.encoder.layers.0.attention.q_proj,"(768, 768)","(
|
| 3 |
-
1,0,MHSA,nncf_module.wav2vec2.encoder.layers.0.attention.k_proj,"(768, 768)","(
|
| 4 |
-
2,0,MHSA,nncf_module.wav2vec2.encoder.layers.0.attention.v_proj,"(768, 768)","(
|
| 5 |
-
3,0,MHSA,nncf_module.wav2vec2.encoder.layers.0.attention.out_proj,"(768, 768)","(768,
|
| 6 |
-
4,1,FF,nncf_module.wav2vec2.encoder.layers.0.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 7 |
-
5,1,FF,nncf_module.wav2vec2.encoder.layers.0.feed_forward.output_dense,"(768, 3072)","(768,
|
| 8 |
-
6,2,MHSA,nncf_module.wav2vec2.encoder.layers.1.attention.q_proj,"(768, 768)","(
|
| 9 |
-
7,2,MHSA,nncf_module.wav2vec2.encoder.layers.1.attention.k_proj,"(768, 768)","(
|
| 10 |
-
8,2,MHSA,nncf_module.wav2vec2.encoder.layers.1.attention.v_proj,"(768, 768)","(
|
| 11 |
-
9,2,MHSA,nncf_module.wav2vec2.encoder.layers.1.attention.out_proj,"(768, 768)","(768,
|
| 12 |
-
10,3,FF,nncf_module.wav2vec2.encoder.layers.1.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 13 |
-
11,3,FF,nncf_module.wav2vec2.encoder.layers.1.feed_forward.output_dense,"(768, 3072)","(768,
|
| 14 |
-
12,4,MHSA,nncf_module.wav2vec2.encoder.layers.2.attention.q_proj,"(768, 768)","(
|
| 15 |
-
13,4,MHSA,nncf_module.wav2vec2.encoder.layers.2.attention.k_proj,"(768, 768)","(
|
| 16 |
-
14,4,MHSA,nncf_module.wav2vec2.encoder.layers.2.attention.v_proj,"(768, 768)","(
|
| 17 |
-
15,4,MHSA,nncf_module.wav2vec2.encoder.layers.2.attention.out_proj,"(768, 768)","(768,
|
| 18 |
-
16,5,FF,nncf_module.wav2vec2.encoder.layers.2.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 19 |
-
17,5,FF,nncf_module.wav2vec2.encoder.layers.2.feed_forward.output_dense,"(768, 3072)","(768,
|
| 20 |
-
18,6,MHSA,nncf_module.wav2vec2.encoder.layers.3.attention.q_proj,"(768, 768)","(
|
| 21 |
-
19,6,MHSA,nncf_module.wav2vec2.encoder.layers.3.attention.k_proj,"(768, 768)","(
|
| 22 |
-
20,6,MHSA,nncf_module.wav2vec2.encoder.layers.3.attention.v_proj,"(768, 768)","(
|
| 23 |
-
21,6,MHSA,nncf_module.wav2vec2.encoder.layers.3.attention.out_proj,"(768, 768)","(768,
|
| 24 |
-
22,7,FF,nncf_module.wav2vec2.encoder.layers.3.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 25 |
-
23,7,FF,nncf_module.wav2vec2.encoder.layers.3.feed_forward.output_dense,"(768, 3072)","(768,
|
| 26 |
-
24,8,MHSA,nncf_module.wav2vec2.encoder.layers.4.attention.q_proj,"(768, 768)","(
|
| 27 |
-
25,8,MHSA,nncf_module.wav2vec2.encoder.layers.4.attention.k_proj,"(768, 768)","(
|
| 28 |
-
26,8,MHSA,nncf_module.wav2vec2.encoder.layers.4.attention.v_proj,"(768, 768)","(
|
| 29 |
-
27,8,MHSA,nncf_module.wav2vec2.encoder.layers.4.attention.out_proj,"(768, 768)","(768,
|
| 30 |
-
28,9,FF,nncf_module.wav2vec2.encoder.layers.4.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 31 |
-
29,9,FF,nncf_module.wav2vec2.encoder.layers.4.feed_forward.output_dense,"(768, 3072)","(768,
|
| 32 |
-
30,10,MHSA,nncf_module.wav2vec2.encoder.layers.5.attention.q_proj,"(768, 768)","(
|
| 33 |
-
31,10,MHSA,nncf_module.wav2vec2.encoder.layers.5.attention.k_proj,"(768, 768)","(
|
| 34 |
-
32,10,MHSA,nncf_module.wav2vec2.encoder.layers.5.attention.v_proj,"(768, 768)","(
|
| 35 |
-
33,10,MHSA,nncf_module.wav2vec2.encoder.layers.5.attention.out_proj,"(768, 768)","(768,
|
| 36 |
-
34,11,FF,nncf_module.wav2vec2.encoder.layers.5.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 37 |
-
35,11,FF,nncf_module.wav2vec2.encoder.layers.5.feed_forward.output_dense,"(768, 3072)","(768,
|
| 38 |
36,12,MHSA,nncf_module.wav2vec2.encoder.layers.6.attention.q_proj,"(768, 768)","(64, 768)","(768,)","(64,)",[9],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 39 |
37,12,MHSA,nncf_module.wav2vec2.encoder.layers.6.attention.k_proj,"(768, 768)","(64, 768)","(768,)","(64,)",[9],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 40 |
38,12,MHSA,nncf_module.wav2vec2.encoder.layers.6.attention.v_proj,"(768, 768)","(64, 768)","(768,)","(64,)",[9],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 41 |
39,12,MHSA,nncf_module.wav2vec2.encoder.layers.6.attention.out_proj,"(768, 768)","(768, 64)","(768,)","(768,)",[9],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 42 |
-
40,13,FF,nncf_module.wav2vec2.encoder.layers.6.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 43 |
-
41,13,FF,nncf_module.wav2vec2.encoder.layers.6.feed_forward.output_dense,"(768, 3072)","(768,
|
| 44 |
-
42,14,MHSA,nncf_module.wav2vec2.encoder.layers.7.attention.q_proj,"(768, 768)","(
|
| 45 |
-
43,14,MHSA,nncf_module.wav2vec2.encoder.layers.7.attention.k_proj,"(768, 768)","(
|
| 46 |
-
44,14,MHSA,nncf_module.wav2vec2.encoder.layers.7.attention.v_proj,"(768, 768)","(
|
| 47 |
-
45,14,MHSA,nncf_module.wav2vec2.encoder.layers.7.attention.out_proj,"(768, 768)","(768,
|
| 48 |
-
46,15,FF,nncf_module.wav2vec2.encoder.layers.7.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 49 |
-
47,15,FF,nncf_module.wav2vec2.encoder.layers.7.feed_forward.output_dense,"(768, 3072)","(768,
|
| 50 |
-
48,16,MHSA,nncf_module.wav2vec2.encoder.layers.8.attention.q_proj,"(768, 768)","(
|
| 51 |
-
49,16,MHSA,nncf_module.wav2vec2.encoder.layers.8.attention.k_proj,"(768, 768)","(
|
| 52 |
-
50,16,MHSA,nncf_module.wav2vec2.encoder.layers.8.attention.v_proj,"(768, 768)","(
|
| 53 |
-
51,16,MHSA,nncf_module.wav2vec2.encoder.layers.8.attention.out_proj,"(768, 768)","(768,
|
| 54 |
-
52,17,FF,nncf_module.wav2vec2.encoder.layers.8.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 55 |
-
53,17,FF,nncf_module.wav2vec2.encoder.layers.8.feed_forward.output_dense,"(768, 3072)","(768,
|
| 56 |
-
54,18,MHSA,nncf_module.wav2vec2.encoder.layers.9.attention.q_proj,"(768, 768)","(
|
| 57 |
-
55,18,MHSA,nncf_module.wav2vec2.encoder.layers.9.attention.k_proj,"(768, 768)","(
|
| 58 |
-
56,18,MHSA,nncf_module.wav2vec2.encoder.layers.9.attention.v_proj,"(768, 768)","(
|
| 59 |
-
57,18,MHSA,nncf_module.wav2vec2.encoder.layers.9.attention.out_proj,"(768, 768)","(768,
|
| 60 |
-
58,19,FF,nncf_module.wav2vec2.encoder.layers.9.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 61 |
-
59,19,FF,nncf_module.wav2vec2.encoder.layers.9.feed_forward.output_dense,"(768, 3072)","(768,
|
| 62 |
-
60,20,MHSA,nncf_module.wav2vec2.encoder.layers.10.attention.q_proj,"(768, 768)","(
|
| 63 |
-
61,20,MHSA,nncf_module.wav2vec2.encoder.layers.10.attention.k_proj,"(768, 768)","(
|
| 64 |
-
62,20,MHSA,nncf_module.wav2vec2.encoder.layers.10.attention.v_proj,"(768, 768)","(
|
| 65 |
-
63,20,MHSA,nncf_module.wav2vec2.encoder.layers.10.attention.out_proj,"(768, 768)","(768,
|
| 66 |
-
64,21,FF,nncf_module.wav2vec2.encoder.layers.10.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 67 |
-
65,21,FF,nncf_module.wav2vec2.encoder.layers.10.feed_forward.output_dense,"(768, 3072)","(768,
|
| 68 |
-
66,22,MHSA,nncf_module.wav2vec2.encoder.layers.11.attention.q_proj,"(768, 768)","(
|
| 69 |
-
67,22,MHSA,nncf_module.wav2vec2.encoder.layers.11.attention.k_proj,"(768, 768)","(
|
| 70 |
-
68,22,MHSA,nncf_module.wav2vec2.encoder.layers.11.attention.v_proj,"(768, 768)","(
|
| 71 |
-
69,22,MHSA,nncf_module.wav2vec2.encoder.layers.11.attention.out_proj,"(768, 768)","(768,
|
| 72 |
-
70,23,FF,nncf_module.wav2vec2.encoder.layers.11.feed_forward.intermediate_dense,"(3072, 768)","(
|
| 73 |
-
71,23,FF,nncf_module.wav2vec2.encoder.layers.11.feed_forward.output_dense,"(768, 3072)","(768,
|
|
|
|
| 1 |
,group_id,type,torch_module,weight_shape,pruned_weight_shape,bias_shape,pruned_bias_shape,head_or_channel_id_to_keep,module_node_name
|
| 2 |
+
0,0,MHSA,nncf_module.wav2vec2.encoder.layers.0.attention.q_proj,"(768, 768)","(448, 768)","(768,)","(448,)","[0, 2, 3, 4, 5, 6, 10]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[0]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 3 |
+
1,0,MHSA,nncf_module.wav2vec2.encoder.layers.0.attention.k_proj,"(768, 768)","(448, 768)","(768,)","(448,)","[0, 2, 3, 4, 5, 6, 10]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[0]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 4 |
+
2,0,MHSA,nncf_module.wav2vec2.encoder.layers.0.attention.v_proj,"(768, 768)","(448, 768)","(768,)","(448,)","[0, 2, 3, 4, 5, 6, 10]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[0]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 5 |
+
3,0,MHSA,nncf_module.wav2vec2.encoder.layers.0.attention.out_proj,"(768, 768)","(768, 448)","(768,)","(768,)","[0, 2, 3, 4, 5, 6, 10]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[0]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 6 |
+
4,1,FF,nncf_module.wav2vec2.encoder.layers.0.feed_forward.intermediate_dense,"(3072, 768)","(1846, 768)","(3072,)","(1846,)",[1846 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[0]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 7 |
+
5,1,FF,nncf_module.wav2vec2.encoder.layers.0.feed_forward.output_dense,"(768, 3072)","(768, 1846)","(768,)","(768,)",[1846 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[0]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 8 |
+
6,2,MHSA,nncf_module.wav2vec2.encoder.layers.1.attention.q_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[2, 3, 4, 5, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[1]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 9 |
+
7,2,MHSA,nncf_module.wav2vec2.encoder.layers.1.attention.k_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[2, 3, 4, 5, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[1]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 10 |
+
8,2,MHSA,nncf_module.wav2vec2.encoder.layers.1.attention.v_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[2, 3, 4, 5, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[1]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 11 |
+
9,2,MHSA,nncf_module.wav2vec2.encoder.layers.1.attention.out_proj,"(768, 768)","(768, 384)","(768,)","(768,)","[2, 3, 4, 5, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[1]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 12 |
+
10,3,FF,nncf_module.wav2vec2.encoder.layers.1.feed_forward.intermediate_dense,"(3072, 768)","(1527, 768)","(3072,)","(1527,)",[1527 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[1]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 13 |
+
11,3,FF,nncf_module.wav2vec2.encoder.layers.1.feed_forward.output_dense,"(768, 3072)","(768, 1527)","(768,)","(768,)",[1527 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[1]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 14 |
+
12,4,MHSA,nncf_module.wav2vec2.encoder.layers.2.attention.q_proj,"(768, 768)","(320, 768)","(768,)","(320,)","[4, 6, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[2]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 15 |
+
13,4,MHSA,nncf_module.wav2vec2.encoder.layers.2.attention.k_proj,"(768, 768)","(320, 768)","(768,)","(320,)","[4, 6, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[2]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 16 |
+
14,4,MHSA,nncf_module.wav2vec2.encoder.layers.2.attention.v_proj,"(768, 768)","(320, 768)","(768,)","(320,)","[4, 6, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[2]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 17 |
+
15,4,MHSA,nncf_module.wav2vec2.encoder.layers.2.attention.out_proj,"(768, 768)","(768, 320)","(768,)","(768,)","[4, 6, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[2]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 18 |
+
16,5,FF,nncf_module.wav2vec2.encoder.layers.2.feed_forward.intermediate_dense,"(3072, 768)","(1438, 768)","(3072,)","(1438,)",[1438 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[2]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 19 |
+
17,5,FF,nncf_module.wav2vec2.encoder.layers.2.feed_forward.output_dense,"(768, 3072)","(768, 1438)","(768,)","(768,)",[1438 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[2]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 20 |
+
18,6,MHSA,nncf_module.wav2vec2.encoder.layers.3.attention.q_proj,"(768, 768)","(192, 768)","(768,)","(192,)","[2, 4, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[3]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 21 |
+
19,6,MHSA,nncf_module.wav2vec2.encoder.layers.3.attention.k_proj,"(768, 768)","(192, 768)","(768,)","(192,)","[2, 4, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[3]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 22 |
+
20,6,MHSA,nncf_module.wav2vec2.encoder.layers.3.attention.v_proj,"(768, 768)","(192, 768)","(768,)","(192,)","[2, 4, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[3]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 23 |
+
21,6,MHSA,nncf_module.wav2vec2.encoder.layers.3.attention.out_proj,"(768, 768)","(768, 192)","(768,)","(768,)","[2, 4, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[3]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 24 |
+
22,7,FF,nncf_module.wav2vec2.encoder.layers.3.feed_forward.intermediate_dense,"(3072, 768)","(1466, 768)","(3072,)","(1466,)",[1466 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[3]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 25 |
+
23,7,FF,nncf_module.wav2vec2.encoder.layers.3.feed_forward.output_dense,"(768, 3072)","(768, 1466)","(768,)","(768,)",[1466 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[3]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 26 |
+
24,8,MHSA,nncf_module.wav2vec2.encoder.layers.4.attention.q_proj,"(768, 768)","(256, 768)","(768,)","(256,)","[1, 4, 7, 10]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[4]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 27 |
+
25,8,MHSA,nncf_module.wav2vec2.encoder.layers.4.attention.k_proj,"(768, 768)","(256, 768)","(768,)","(256,)","[1, 4, 7, 10]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[4]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 28 |
+
26,8,MHSA,nncf_module.wav2vec2.encoder.layers.4.attention.v_proj,"(768, 768)","(256, 768)","(768,)","(256,)","[1, 4, 7, 10]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[4]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 29 |
+
27,8,MHSA,nncf_module.wav2vec2.encoder.layers.4.attention.out_proj,"(768, 768)","(768, 256)","(768,)","(768,)","[1, 4, 7, 10]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[4]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 30 |
+
28,9,FF,nncf_module.wav2vec2.encoder.layers.4.feed_forward.intermediate_dense,"(3072, 768)","(985, 768)","(3072,)","(985,)",[985 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[4]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 31 |
+
29,9,FF,nncf_module.wav2vec2.encoder.layers.4.feed_forward.output_dense,"(768, 3072)","(768, 985)","(768,)","(768,)",[985 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[4]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 32 |
+
30,10,MHSA,nncf_module.wav2vec2.encoder.layers.5.attention.q_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[1, 2, 8, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[5]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 33 |
+
31,10,MHSA,nncf_module.wav2vec2.encoder.layers.5.attention.k_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[1, 2, 8, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[5]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 34 |
+
32,10,MHSA,nncf_module.wav2vec2.encoder.layers.5.attention.v_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[1, 2, 8, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[5]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 35 |
+
33,10,MHSA,nncf_module.wav2vec2.encoder.layers.5.attention.out_proj,"(768, 768)","(768, 384)","(768,)","(768,)","[1, 2, 8, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[5]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 36 |
+
34,11,FF,nncf_module.wav2vec2.encoder.layers.5.feed_forward.intermediate_dense,"(3072, 768)","(542, 768)","(3072,)","(542,)",[542 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[5]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 37 |
+
35,11,FF,nncf_module.wav2vec2.encoder.layers.5.feed_forward.output_dense,"(768, 3072)","(768, 542)","(768,)","(768,)",[542 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[5]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 38 |
36,12,MHSA,nncf_module.wav2vec2.encoder.layers.6.attention.q_proj,"(768, 768)","(64, 768)","(768,)","(64,)",[9],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 39 |
37,12,MHSA,nncf_module.wav2vec2.encoder.layers.6.attention.k_proj,"(768, 768)","(64, 768)","(768,)","(64,)",[9],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 40 |
38,12,MHSA,nncf_module.wav2vec2.encoder.layers.6.attention.v_proj,"(768, 768)","(64, 768)","(768,)","(64,)",[9],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 41 |
39,12,MHSA,nncf_module.wav2vec2.encoder.layers.6.attention.out_proj,"(768, 768)","(768, 64)","(768,)","(768,)",[9],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 42 |
+
40,13,FF,nncf_module.wav2vec2.encoder.layers.6.feed_forward.intermediate_dense,"(3072, 768)","(561, 768)","(3072,)","(561,)",[561 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 43 |
+
41,13,FF,nncf_module.wav2vec2.encoder.layers.6.feed_forward.output_dense,"(768, 3072)","(768, 561)","(768,)","(768,)",[561 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[6]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 44 |
+
42,14,MHSA,nncf_module.wav2vec2.encoder.layers.7.attention.q_proj,"(768, 768)","(256, 768)","(768,)","(256,)","[1, 4, 5, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[7]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 45 |
+
43,14,MHSA,nncf_module.wav2vec2.encoder.layers.7.attention.k_proj,"(768, 768)","(256, 768)","(768,)","(256,)","[1, 4, 5, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[7]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 46 |
+
44,14,MHSA,nncf_module.wav2vec2.encoder.layers.7.attention.v_proj,"(768, 768)","(256, 768)","(768,)","(256,)","[1, 4, 5, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[7]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 47 |
+
45,14,MHSA,nncf_module.wav2vec2.encoder.layers.7.attention.out_proj,"(768, 768)","(768, 256)","(768,)","(768,)","[1, 4, 5, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[7]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 48 |
+
46,15,FF,nncf_module.wav2vec2.encoder.layers.7.feed_forward.intermediate_dense,"(3072, 768)","(641, 768)","(3072,)","(641,)",[641 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[7]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 49 |
+
47,15,FF,nncf_module.wav2vec2.encoder.layers.7.feed_forward.output_dense,"(768, 3072)","(768, 641)","(768,)","(768,)",[641 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[7]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 50 |
+
48,16,MHSA,nncf_module.wav2vec2.encoder.layers.8.attention.q_proj,"(768, 768)","(640, 768)","(768,)","(640,)","[0, 1, 2, 3, 4, 5, 7, 8, 9, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[8]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 51 |
+
49,16,MHSA,nncf_module.wav2vec2.encoder.layers.8.attention.k_proj,"(768, 768)","(640, 768)","(768,)","(640,)","[0, 1, 2, 3, 4, 5, 7, 8, 9, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[8]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 52 |
+
50,16,MHSA,nncf_module.wav2vec2.encoder.layers.8.attention.v_proj,"(768, 768)","(640, 768)","(768,)","(640,)","[0, 1, 2, 3, 4, 5, 7, 8, 9, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[8]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 53 |
+
51,16,MHSA,nncf_module.wav2vec2.encoder.layers.8.attention.out_proj,"(768, 768)","(768, 640)","(768,)","(768,)","[0, 1, 2, 3, 4, 5, 7, 8, 9, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[8]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 54 |
+
52,17,FF,nncf_module.wav2vec2.encoder.layers.8.feed_forward.intermediate_dense,"(3072, 768)","(667, 768)","(3072,)","(667,)",[667 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[8]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 55 |
+
53,17,FF,nncf_module.wav2vec2.encoder.layers.8.feed_forward.output_dense,"(768, 3072)","(768, 667)","(768,)","(768,)",[667 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[8]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 56 |
+
54,18,MHSA,nncf_module.wav2vec2.encoder.layers.9.attention.q_proj,"(768, 768)","(576, 768)","(768,)","(576,)","[0, 2, 4, 6, 7, 8, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[9]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 57 |
+
55,18,MHSA,nncf_module.wav2vec2.encoder.layers.9.attention.k_proj,"(768, 768)","(576, 768)","(768,)","(576,)","[0, 2, 4, 6, 7, 8, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[9]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 58 |
+
56,18,MHSA,nncf_module.wav2vec2.encoder.layers.9.attention.v_proj,"(768, 768)","(576, 768)","(768,)","(576,)","[0, 2, 4, 6, 7, 8, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[9]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 59 |
+
57,18,MHSA,nncf_module.wav2vec2.encoder.layers.9.attention.out_proj,"(768, 768)","(768, 576)","(768,)","(768,)","[0, 2, 4, 6, 7, 8, 9, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[9]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 60 |
+
58,19,FF,nncf_module.wav2vec2.encoder.layers.9.feed_forward.intermediate_dense,"(3072, 768)","(1299, 768)","(3072,)","(1299,)",[1299 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[9]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 61 |
+
59,19,FF,nncf_module.wav2vec2.encoder.layers.9.feed_forward.output_dense,"(768, 3072)","(768, 1299)","(768,)","(768,)",[1299 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[9]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 62 |
+
60,20,MHSA,nncf_module.wav2vec2.encoder.layers.10.attention.q_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[2, 3, 5, 6, 8, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[10]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 63 |
+
61,20,MHSA,nncf_module.wav2vec2.encoder.layers.10.attention.k_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[2, 3, 5, 6, 8, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[10]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 64 |
+
62,20,MHSA,nncf_module.wav2vec2.encoder.layers.10.attention.v_proj,"(768, 768)","(384, 768)","(768,)","(384,)","[2, 3, 5, 6, 8, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[10]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 65 |
+
63,20,MHSA,nncf_module.wav2vec2.encoder.layers.10.attention.out_proj,"(768, 768)","(768, 384)","(768,)","(768,)","[2, 3, 5, 6, 8, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[10]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 66 |
+
64,21,FF,nncf_module.wav2vec2.encoder.layers.10.feed_forward.intermediate_dense,"(3072, 768)","(1524, 768)","(3072,)","(1524,)",[1524 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[10]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 67 |
+
65,21,FF,nncf_module.wav2vec2.encoder.layers.10.feed_forward.output_dense,"(768, 3072)","(768, 1524)","(768,)","(768,)",[1524 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[10]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
| 68 |
+
66,22,MHSA,nncf_module.wav2vec2.encoder.layers.11.attention.q_proj,"(768, 768)","(704, 768)","(768,)","(704,)","[0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[11]/Wav2Vec2Attention[attention]/NNCFLinear[q_proj]/linear_0
|
| 69 |
+
67,22,MHSA,nncf_module.wav2vec2.encoder.layers.11.attention.k_proj,"(768, 768)","(704, 768)","(768,)","(704,)","[0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[11]/Wav2Vec2Attention[attention]/NNCFLinear[k_proj]/linear_0
|
| 70 |
+
68,22,MHSA,nncf_module.wav2vec2.encoder.layers.11.attention.v_proj,"(768, 768)","(704, 768)","(768,)","(704,)","[0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[11]/Wav2Vec2Attention[attention]/NNCFLinear[v_proj]/linear_0
|
| 71 |
+
69,22,MHSA,nncf_module.wav2vec2.encoder.layers.11.attention.out_proj,"(768, 768)","(768, 704)","(768,)","(768,)","[0, 1, 2, 3, 4, 5, 6, 7, 8, 10, 11]",Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[11]/Wav2Vec2Attention[attention]/NNCFLinear[out_proj]/linear_0
|
| 72 |
+
70,23,FF,nncf_module.wav2vec2.encoder.layers.11.feed_forward.intermediate_dense,"(3072, 768)","(1471, 768)","(3072,)","(1471,)",[1471 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[11]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[intermediate_dense]/linear_0
|
| 73 |
+
71,23,FF,nncf_module.wav2vec2.encoder.layers.11.feed_forward.output_dense,"(768, 3072)","(768, 1471)","(768,)","(768,)",[1471 items],Wav2Vec2ForSequenceClassification/Wav2Vec2Model[wav2vec2]/Wav2Vec2Encoder[encoder]/ModuleList[layers]/Wav2Vec2EncoderLayer[11]/Wav2Vec2FeedForward[feed_forward]/NNCFLinear[output_dense]/linear_0
|
train_results.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_loss":
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples_per_second":
|
| 6 |
-
"train_steps_per_second": 0.
|
| 7 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 12.0,
|
| 3 |
+
"train_loss": 4.577942104168304,
|
| 4 |
+
"train_runtime": 8786.4706,
|
| 5 |
+
"train_samples_per_second": 69.781,
|
| 6 |
+
"train_steps_per_second": 0.545
|
| 7 |
}
|
trainer_state.json
CHANGED
|
The diff for this file is too large to render.
See raw diff
|
|
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f44269e45cc9a533c9d8c72aeca0705d418c3732a68bccac7f7e9159248c3dfd
|
| 3 |
+
size 3643
|