Commit
•
feaf824
1
Parent(s):
25de0cf
clean main branch
Browse files- checkpoint-4400/config.json +0 -115
- checkpoint-4400/optimizer.pt +0 -3
- checkpoint-4400/preprocessor_config.json +0 -9
- checkpoint-4400/pytorch_model.bin +0 -3
- checkpoint-4400/rng_state.pth +0 -3
- checkpoint-4400/scaler.pt +0 -3
- checkpoint-4400/scheduler.pt +0 -3
- checkpoint-4400/trainer_state.json +0 -1732
- checkpoint-4400/training_args.bin +0 -3
- checkpoint-4500/config.json +0 -115
- checkpoint-4500/optimizer.pt +0 -3
- checkpoint-4500/preprocessor_config.json +0 -9
- checkpoint-4500/pytorch_model.bin +0 -3
- checkpoint-4500/rng_state.pth +0 -3
- checkpoint-4500/scaler.pt +0 -3
- checkpoint-4500/scheduler.pt +0 -3
- checkpoint-4500/trainer_state.json +0 -1771
- checkpoint-4500/training_args.bin +0 -3
- old_run_asr_ctc.py → join_datasets_asr_ctc.py +0 -0
- old_run.sh → join_datasets_asr_ctc_run.sh +0 -0
checkpoint-4400/config.json
DELETED
|
@@ -1,115 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_name_or_path": "marinone94/xls-r-300m-sv-robust",
|
| 3 |
-
"activation_dropout": 0.1,
|
| 4 |
-
"adapter_kernel_size": 3,
|
| 5 |
-
"adapter_stride": 2,
|
| 6 |
-
"add_adapter": false,
|
| 7 |
-
"apply_spec_augment": true,
|
| 8 |
-
"architectures": [
|
| 9 |
-
"Wav2Vec2ForCTC"
|
| 10 |
-
],
|
| 11 |
-
"attention_dropout": 0.0,
|
| 12 |
-
"bos_token_id": 1,
|
| 13 |
-
"classifier_proj_size": 256,
|
| 14 |
-
"codevector_dim": 768,
|
| 15 |
-
"contrastive_logits_temperature": 0.1,
|
| 16 |
-
"conv_bias": true,
|
| 17 |
-
"conv_dim": [
|
| 18 |
-
512,
|
| 19 |
-
512,
|
| 20 |
-
512,
|
| 21 |
-
512,
|
| 22 |
-
512,
|
| 23 |
-
512,
|
| 24 |
-
512
|
| 25 |
-
],
|
| 26 |
-
"conv_kernel": [
|
| 27 |
-
10,
|
| 28 |
-
3,
|
| 29 |
-
3,
|
| 30 |
-
3,
|
| 31 |
-
3,
|
| 32 |
-
2,
|
| 33 |
-
2
|
| 34 |
-
],
|
| 35 |
-
"conv_stride": [
|
| 36 |
-
5,
|
| 37 |
-
2,
|
| 38 |
-
2,
|
| 39 |
-
2,
|
| 40 |
-
2,
|
| 41 |
-
2,
|
| 42 |
-
2
|
| 43 |
-
],
|
| 44 |
-
"ctc_loss_reduction": "mean",
|
| 45 |
-
"ctc_zero_infinity": false,
|
| 46 |
-
"diversity_loss_weight": 0.1,
|
| 47 |
-
"do_stable_layer_norm": true,
|
| 48 |
-
"eos_token_id": 2,
|
| 49 |
-
"feat_extract_activation": "gelu",
|
| 50 |
-
"feat_extract_dropout": 0.0,
|
| 51 |
-
"feat_extract_norm": "layer",
|
| 52 |
-
"feat_proj_dropout": 0.0,
|
| 53 |
-
"feat_quantizer_dropout": 0.0,
|
| 54 |
-
"final_dropout": 0.0,
|
| 55 |
-
"hidden_act": "gelu",
|
| 56 |
-
"hidden_dropout": 0.0,
|
| 57 |
-
"hidden_size": 1024,
|
| 58 |
-
"initializer_range": 0.02,
|
| 59 |
-
"intermediate_size": 4096,
|
| 60 |
-
"layer_norm_eps": 1e-05,
|
| 61 |
-
"layerdrop": 0.0,
|
| 62 |
-
"mask_channel_length": 10,
|
| 63 |
-
"mask_channel_min_space": 1,
|
| 64 |
-
"mask_channel_other": 0.0,
|
| 65 |
-
"mask_channel_prob": 0.0,
|
| 66 |
-
"mask_channel_selection": "static",
|
| 67 |
-
"mask_feature_length": 64,
|
| 68 |
-
"mask_feature_min_masks": 0,
|
| 69 |
-
"mask_feature_prob": 0.25,
|
| 70 |
-
"mask_time_length": 10,
|
| 71 |
-
"mask_time_min_masks": 2,
|
| 72 |
-
"mask_time_min_space": 1,
|
| 73 |
-
"mask_time_other": 0.0,
|
| 74 |
-
"mask_time_prob": 0.75,
|
| 75 |
-
"mask_time_selection": "static",
|
| 76 |
-
"model_type": "wav2vec2",
|
| 77 |
-
"num_adapter_layers": 3,
|
| 78 |
-
"num_attention_heads": 16,
|
| 79 |
-
"num_codevector_groups": 2,
|
| 80 |
-
"num_codevectors_per_group": 320,
|
| 81 |
-
"num_conv_pos_embedding_groups": 16,
|
| 82 |
-
"num_conv_pos_embeddings": 128,
|
| 83 |
-
"num_feat_extract_layers": 7,
|
| 84 |
-
"num_hidden_layers": 24,
|
| 85 |
-
"num_negatives": 100,
|
| 86 |
-
"output_hidden_size": 1024,
|
| 87 |
-
"pad_token_id": 31,
|
| 88 |
-
"proj_codevector_dim": 768,
|
| 89 |
-
"tdnn_dilation": [
|
| 90 |
-
1,
|
| 91 |
-
2,
|
| 92 |
-
3,
|
| 93 |
-
1,
|
| 94 |
-
1
|
| 95 |
-
],
|
| 96 |
-
"tdnn_dim": [
|
| 97 |
-
512,
|
| 98 |
-
512,
|
| 99 |
-
512,
|
| 100 |
-
512,
|
| 101 |
-
1500
|
| 102 |
-
],
|
| 103 |
-
"tdnn_kernel": [
|
| 104 |
-
5,
|
| 105 |
-
3,
|
| 106 |
-
3,
|
| 107 |
-
1,
|
| 108 |
-
1
|
| 109 |
-
],
|
| 110 |
-
"torch_dtype": "float32",
|
| 111 |
-
"transformers_version": "4.17.0.dev0",
|
| 112 |
-
"use_weighted_layer_sum": false,
|
| 113 |
-
"vocab_size": 34,
|
| 114 |
-
"xvector_output_dim": 512
|
| 115 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4400/optimizer.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:2ed693bd3b91cdbbe1430c3f7f43d32b0b9b099aa5180fb31ece0f93c73b26ee
|
| 3 |
-
size 2490337809
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4400/preprocessor_config.json
DELETED
|
@@ -1,9 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"do_normalize": true,
|
| 3 |
-
"feature_extractor_type": "Wav2Vec2FeatureExtractor",
|
| 4 |
-
"feature_size": 1,
|
| 5 |
-
"padding_side": "right",
|
| 6 |
-
"padding_value": 0,
|
| 7 |
-
"return_attention_mask": true,
|
| 8 |
-
"sampling_rate": 16000
|
| 9 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4400/pytorch_model.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:f3c90a3d0a519652bea358ec5bc82e2a0b669425ce58bf0947c048a13d83b74d
|
| 3 |
-
size 1262063089
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4400/rng_state.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:0bb367ac83047020ca661382c5402cfaed86ec3e8ea3ce8b448d44d9e43e2965
|
| 3 |
-
size 14567
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4400/scaler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:cb287cb35474e15fb80cdda16160c24d664008c453ecb31933c1042ff81dd3e9
|
| 3 |
-
size 559
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4400/scheduler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:379bb05b55352db422771e61095d894e4be45df2c25710a995432fffb6cdd028
|
| 3 |
-
size 623
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4400/trainer_state.json
DELETED
|
@@ -1,1732 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"best_metric": null,
|
| 3 |
-
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 48.349726775956285,
|
| 5 |
-
"global_step": 4400,
|
| 6 |
-
"is_hyper_param_search": false,
|
| 7 |
-
"is_local_process_zero": true,
|
| 8 |
-
"is_world_process_zero": true,
|
| 9 |
-
"log_history": [
|
| 10 |
-
{
|
| 11 |
-
"epoch": 0.22,
|
| 12 |
-
"learning_rate": 2.5e-06,
|
| 13 |
-
"loss": 3.5867,
|
| 14 |
-
"step": 20
|
| 15 |
-
},
|
| 16 |
-
{
|
| 17 |
-
"epoch": 0.44,
|
| 18 |
-
"learning_rate": 5e-06,
|
| 19 |
-
"loss": 3.5457,
|
| 20 |
-
"step": 40
|
| 21 |
-
},
|
| 22 |
-
{
|
| 23 |
-
"epoch": 0.66,
|
| 24 |
-
"learning_rate": 7.5e-06,
|
| 25 |
-
"loss": 3.4513,
|
| 26 |
-
"step": 60
|
| 27 |
-
},
|
| 28 |
-
{
|
| 29 |
-
"epoch": 0.87,
|
| 30 |
-
"learning_rate": 1e-05,
|
| 31 |
-
"loss": 3.3432,
|
| 32 |
-
"step": 80
|
| 33 |
-
},
|
| 34 |
-
{
|
| 35 |
-
"epoch": 1.1,
|
| 36 |
-
"learning_rate": 1.25e-05,
|
| 37 |
-
"loss": 3.3533,
|
| 38 |
-
"step": 100
|
| 39 |
-
},
|
| 40 |
-
{
|
| 41 |
-
"epoch": 1.1,
|
| 42 |
-
"eval_loss": 3.2806732654571533,
|
| 43 |
-
"eval_runtime": 190.4728,
|
| 44 |
-
"eval_samples_per_second": 25.426,
|
| 45 |
-
"eval_steps_per_second": 0.798,
|
| 46 |
-
"eval_wer": 1.0,
|
| 47 |
-
"step": 100
|
| 48 |
-
},
|
| 49 |
-
{
|
| 50 |
-
"epoch": 1.32,
|
| 51 |
-
"learning_rate": 1.5e-05,
|
| 52 |
-
"loss": 3.2217,
|
| 53 |
-
"step": 120
|
| 54 |
-
},
|
| 55 |
-
{
|
| 56 |
-
"epoch": 1.54,
|
| 57 |
-
"learning_rate": 1.7500000000000002e-05,
|
| 58 |
-
"loss": 3.1765,
|
| 59 |
-
"step": 140
|
| 60 |
-
},
|
| 61 |
-
{
|
| 62 |
-
"epoch": 1.75,
|
| 63 |
-
"learning_rate": 2e-05,
|
| 64 |
-
"loss": 3.1408,
|
| 65 |
-
"step": 160
|
| 66 |
-
},
|
| 67 |
-
{
|
| 68 |
-
"epoch": 1.97,
|
| 69 |
-
"learning_rate": 2.2499999999999998e-05,
|
| 70 |
-
"loss": 3.1165,
|
| 71 |
-
"step": 180
|
| 72 |
-
},
|
| 73 |
-
{
|
| 74 |
-
"epoch": 2.2,
|
| 75 |
-
"learning_rate": 2.5e-05,
|
| 76 |
-
"loss": 3.1709,
|
| 77 |
-
"step": 200
|
| 78 |
-
},
|
| 79 |
-
{
|
| 80 |
-
"epoch": 2.2,
|
| 81 |
-
"eval_loss": 3.1325438022613525,
|
| 82 |
-
"eval_runtime": 192.4978,
|
| 83 |
-
"eval_samples_per_second": 25.159,
|
| 84 |
-
"eval_steps_per_second": 0.79,
|
| 85 |
-
"eval_wer": 1.0,
|
| 86 |
-
"step": 200
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"epoch": 2.42,
|
| 90 |
-
"learning_rate": 2.75e-05,
|
| 91 |
-
"loss": 3.079,
|
| 92 |
-
"step": 220
|
| 93 |
-
},
|
| 94 |
-
{
|
| 95 |
-
"epoch": 2.63,
|
| 96 |
-
"learning_rate": 3e-05,
|
| 97 |
-
"loss": 3.0677,
|
| 98 |
-
"step": 240
|
| 99 |
-
},
|
| 100 |
-
{
|
| 101 |
-
"epoch": 2.85,
|
| 102 |
-
"learning_rate": 3.2500000000000004e-05,
|
| 103 |
-
"loss": 3.0656,
|
| 104 |
-
"step": 260
|
| 105 |
-
},
|
| 106 |
-
{
|
| 107 |
-
"epoch": 3.08,
|
| 108 |
-
"learning_rate": 3.5000000000000004e-05,
|
| 109 |
-
"loss": 3.1463,
|
| 110 |
-
"step": 280
|
| 111 |
-
},
|
| 112 |
-
{
|
| 113 |
-
"epoch": 3.3,
|
| 114 |
-
"learning_rate": 3.75e-05,
|
| 115 |
-
"loss": 3.0573,
|
| 116 |
-
"step": 300
|
| 117 |
-
},
|
| 118 |
-
{
|
| 119 |
-
"epoch": 3.3,
|
| 120 |
-
"eval_loss": 3.0614514350891113,
|
| 121 |
-
"eval_runtime": 194.36,
|
| 122 |
-
"eval_samples_per_second": 24.918,
|
| 123 |
-
"eval_steps_per_second": 0.782,
|
| 124 |
-
"eval_wer": 1.0,
|
| 125 |
-
"step": 300
|
| 126 |
-
},
|
| 127 |
-
{
|
| 128 |
-
"epoch": 3.51,
|
| 129 |
-
"learning_rate": 4e-05,
|
| 130 |
-
"loss": 3.0511,
|
| 131 |
-
"step": 320
|
| 132 |
-
},
|
| 133 |
-
{
|
| 134 |
-
"epoch": 3.73,
|
| 135 |
-
"learning_rate": 4.25e-05,
|
| 136 |
-
"loss": 3.0358,
|
| 137 |
-
"step": 340
|
| 138 |
-
},
|
| 139 |
-
{
|
| 140 |
-
"epoch": 3.95,
|
| 141 |
-
"learning_rate": 4.4999999999999996e-05,
|
| 142 |
-
"loss": 3.0416,
|
| 143 |
-
"step": 360
|
| 144 |
-
},
|
| 145 |
-
{
|
| 146 |
-
"epoch": 4.17,
|
| 147 |
-
"learning_rate": 4.75e-05,
|
| 148 |
-
"loss": 3.1053,
|
| 149 |
-
"step": 380
|
| 150 |
-
},
|
| 151 |
-
{
|
| 152 |
-
"epoch": 4.39,
|
| 153 |
-
"learning_rate": 5e-05,
|
| 154 |
-
"loss": 3.0314,
|
| 155 |
-
"step": 400
|
| 156 |
-
},
|
| 157 |
-
{
|
| 158 |
-
"epoch": 4.39,
|
| 159 |
-
"eval_loss": 3.0990231037139893,
|
| 160 |
-
"eval_runtime": 198.3688,
|
| 161 |
-
"eval_samples_per_second": 24.414,
|
| 162 |
-
"eval_steps_per_second": 0.766,
|
| 163 |
-
"eval_wer": 1.0,
|
| 164 |
-
"step": 400
|
| 165 |
-
},
|
| 166 |
-
{
|
| 167 |
-
"epoch": 4.61,
|
| 168 |
-
"learning_rate": 5.25e-05,
|
| 169 |
-
"loss": 3.0309,
|
| 170 |
-
"step": 420
|
| 171 |
-
},
|
| 172 |
-
{
|
| 173 |
-
"epoch": 4.83,
|
| 174 |
-
"learning_rate": 5.5e-05,
|
| 175 |
-
"loss": 3.0259,
|
| 176 |
-
"step": 440
|
| 177 |
-
},
|
| 178 |
-
{
|
| 179 |
-
"epoch": 5.05,
|
| 180 |
-
"learning_rate": 5.75e-05,
|
| 181 |
-
"loss": 3.0998,
|
| 182 |
-
"step": 460
|
| 183 |
-
},
|
| 184 |
-
{
|
| 185 |
-
"epoch": 5.27,
|
| 186 |
-
"learning_rate": 6e-05,
|
| 187 |
-
"loss": 3.0152,
|
| 188 |
-
"step": 480
|
| 189 |
-
},
|
| 190 |
-
{
|
| 191 |
-
"epoch": 5.49,
|
| 192 |
-
"learning_rate": 6.25e-05,
|
| 193 |
-
"loss": 3.0129,
|
| 194 |
-
"step": 500
|
| 195 |
-
},
|
| 196 |
-
{
|
| 197 |
-
"epoch": 5.49,
|
| 198 |
-
"eval_loss": 3.039973497390747,
|
| 199 |
-
"eval_runtime": 190.8567,
|
| 200 |
-
"eval_samples_per_second": 25.375,
|
| 201 |
-
"eval_steps_per_second": 0.796,
|
| 202 |
-
"eval_wer": 1.0,
|
| 203 |
-
"step": 500
|
| 204 |
-
},
|
| 205 |
-
{
|
| 206 |
-
"epoch": 5.71,
|
| 207 |
-
"learning_rate": 6.500000000000001e-05,
|
| 208 |
-
"loss": 3.0088,
|
| 209 |
-
"step": 520
|
| 210 |
-
},
|
| 211 |
-
{
|
| 212 |
-
"epoch": 5.93,
|
| 213 |
-
"learning_rate": 6.75e-05,
|
| 214 |
-
"loss": 3.0051,
|
| 215 |
-
"step": 540
|
| 216 |
-
},
|
| 217 |
-
{
|
| 218 |
-
"epoch": 6.15,
|
| 219 |
-
"learning_rate": 7.000000000000001e-05,
|
| 220 |
-
"loss": 3.073,
|
| 221 |
-
"step": 560
|
| 222 |
-
},
|
| 223 |
-
{
|
| 224 |
-
"epoch": 6.37,
|
| 225 |
-
"learning_rate": 7.25e-05,
|
| 226 |
-
"loss": 3.0031,
|
| 227 |
-
"step": 580
|
| 228 |
-
},
|
| 229 |
-
{
|
| 230 |
-
"epoch": 6.59,
|
| 231 |
-
"learning_rate": 7.5e-05,
|
| 232 |
-
"loss": 2.9964,
|
| 233 |
-
"step": 600
|
| 234 |
-
},
|
| 235 |
-
{
|
| 236 |
-
"epoch": 6.59,
|
| 237 |
-
"eval_loss": 2.998962640762329,
|
| 238 |
-
"eval_runtime": 193.5213,
|
| 239 |
-
"eval_samples_per_second": 25.026,
|
| 240 |
-
"eval_steps_per_second": 0.785,
|
| 241 |
-
"eval_wer": 1.0,
|
| 242 |
-
"step": 600
|
| 243 |
-
},
|
| 244 |
-
{
|
| 245 |
-
"epoch": 6.81,
|
| 246 |
-
"learning_rate": 7.75e-05,
|
| 247 |
-
"loss": 2.9921,
|
| 248 |
-
"step": 620
|
| 249 |
-
},
|
| 250 |
-
{
|
| 251 |
-
"epoch": 7.03,
|
| 252 |
-
"learning_rate": 8e-05,
|
| 253 |
-
"loss": 3.0665,
|
| 254 |
-
"step": 640
|
| 255 |
-
},
|
| 256 |
-
{
|
| 257 |
-
"epoch": 7.25,
|
| 258 |
-
"learning_rate": 8.25e-05,
|
| 259 |
-
"loss": 2.9826,
|
| 260 |
-
"step": 660
|
| 261 |
-
},
|
| 262 |
-
{
|
| 263 |
-
"epoch": 7.47,
|
| 264 |
-
"learning_rate": 8.5e-05,
|
| 265 |
-
"loss": 2.9689,
|
| 266 |
-
"step": 680
|
| 267 |
-
},
|
| 268 |
-
{
|
| 269 |
-
"epoch": 7.69,
|
| 270 |
-
"learning_rate": 8.75e-05,
|
| 271 |
-
"loss": 2.9602,
|
| 272 |
-
"step": 700
|
| 273 |
-
},
|
| 274 |
-
{
|
| 275 |
-
"epoch": 7.69,
|
| 276 |
-
"eval_loss": 2.9620397090911865,
|
| 277 |
-
"eval_runtime": 193.5851,
|
| 278 |
-
"eval_samples_per_second": 25.017,
|
| 279 |
-
"eval_steps_per_second": 0.785,
|
| 280 |
-
"eval_wer": 1.0,
|
| 281 |
-
"step": 700
|
| 282 |
-
},
|
| 283 |
-
{
|
| 284 |
-
"epoch": 7.91,
|
| 285 |
-
"learning_rate": 8.999999999999999e-05,
|
| 286 |
-
"loss": 2.9639,
|
| 287 |
-
"step": 720
|
| 288 |
-
},
|
| 289 |
-
{
|
| 290 |
-
"epoch": 8.13,
|
| 291 |
-
"learning_rate": 9.25e-05,
|
| 292 |
-
"loss": 3.0215,
|
| 293 |
-
"step": 740
|
| 294 |
-
},
|
| 295 |
-
{
|
| 296 |
-
"epoch": 8.35,
|
| 297 |
-
"learning_rate": 9.5e-05,
|
| 298 |
-
"loss": 2.9454,
|
| 299 |
-
"step": 760
|
| 300 |
-
},
|
| 301 |
-
{
|
| 302 |
-
"epoch": 8.57,
|
| 303 |
-
"learning_rate": 9.750000000000001e-05,
|
| 304 |
-
"loss": 2.9239,
|
| 305 |
-
"step": 780
|
| 306 |
-
},
|
| 307 |
-
{
|
| 308 |
-
"epoch": 8.79,
|
| 309 |
-
"learning_rate": 0.0001,
|
| 310 |
-
"loss": 2.8756,
|
| 311 |
-
"step": 800
|
| 312 |
-
},
|
| 313 |
-
{
|
| 314 |
-
"epoch": 8.79,
|
| 315 |
-
"eval_loss": 2.7302000522613525,
|
| 316 |
-
"eval_runtime": 191.8065,
|
| 317 |
-
"eval_samples_per_second": 25.249,
|
| 318 |
-
"eval_steps_per_second": 0.792,
|
| 319 |
-
"eval_wer": 1.0,
|
| 320 |
-
"step": 800
|
| 321 |
-
},
|
| 322 |
-
{
|
| 323 |
-
"epoch": 9.01,
|
| 324 |
-
"learning_rate": 0.0001025,
|
| 325 |
-
"loss": 2.8933,
|
| 326 |
-
"step": 820
|
| 327 |
-
},
|
| 328 |
-
{
|
| 329 |
-
"epoch": 9.23,
|
| 330 |
-
"learning_rate": 0.000105,
|
| 331 |
-
"loss": 2.7318,
|
| 332 |
-
"step": 840
|
| 333 |
-
},
|
| 334 |
-
{
|
| 335 |
-
"epoch": 9.45,
|
| 336 |
-
"learning_rate": 0.0001075,
|
| 337 |
-
"loss": 2.5941,
|
| 338 |
-
"step": 860
|
| 339 |
-
},
|
| 340 |
-
{
|
| 341 |
-
"epoch": 9.67,
|
| 342 |
-
"learning_rate": 0.00011,
|
| 343 |
-
"loss": 2.4441,
|
| 344 |
-
"step": 880
|
| 345 |
-
},
|
| 346 |
-
{
|
| 347 |
-
"epoch": 9.89,
|
| 348 |
-
"learning_rate": 0.00011250000000000001,
|
| 349 |
-
"loss": 2.2931,
|
| 350 |
-
"step": 900
|
| 351 |
-
},
|
| 352 |
-
{
|
| 353 |
-
"epoch": 9.89,
|
| 354 |
-
"eval_loss": 1.5057899951934814,
|
| 355 |
-
"eval_runtime": 196.4368,
|
| 356 |
-
"eval_samples_per_second": 24.654,
|
| 357 |
-
"eval_steps_per_second": 0.774,
|
| 358 |
-
"eval_wer": 0.9775759296054499,
|
| 359 |
-
"step": 900
|
| 360 |
-
},
|
| 361 |
-
{
|
| 362 |
-
"epoch": 10.11,
|
| 363 |
-
"learning_rate": 0.000115,
|
| 364 |
-
"loss": 2.1999,
|
| 365 |
-
"step": 920
|
| 366 |
-
},
|
| 367 |
-
{
|
| 368 |
-
"epoch": 10.33,
|
| 369 |
-
"learning_rate": 0.0001175,
|
| 370 |
-
"loss": 2.0574,
|
| 371 |
-
"step": 940
|
| 372 |
-
},
|
| 373 |
-
{
|
| 374 |
-
"epoch": 10.55,
|
| 375 |
-
"learning_rate": 0.00012,
|
| 376 |
-
"loss": 1.9672,
|
| 377 |
-
"step": 960
|
| 378 |
-
},
|
| 379 |
-
{
|
| 380 |
-
"epoch": 10.77,
|
| 381 |
-
"learning_rate": 0.0001225,
|
| 382 |
-
"loss": 1.9015,
|
| 383 |
-
"step": 980
|
| 384 |
-
},
|
| 385 |
-
{
|
| 386 |
-
"epoch": 10.98,
|
| 387 |
-
"learning_rate": 0.000125,
|
| 388 |
-
"loss": 1.8427,
|
| 389 |
-
"step": 1000
|
| 390 |
-
},
|
| 391 |
-
{
|
| 392 |
-
"epoch": 10.98,
|
| 393 |
-
"eval_loss": 0.9154536724090576,
|
| 394 |
-
"eval_runtime": 203.1168,
|
| 395 |
-
"eval_samples_per_second": 23.843,
|
| 396 |
-
"eval_steps_per_second": 0.748,
|
| 397 |
-
"eval_wer": 0.7832245245529378,
|
| 398 |
-
"step": 1000
|
| 399 |
-
},
|
| 400 |
-
{
|
| 401 |
-
"epoch": 11.21,
|
| 402 |
-
"learning_rate": 0.0001275,
|
| 403 |
-
"loss": 1.8155,
|
| 404 |
-
"step": 1020
|
| 405 |
-
},
|
| 406 |
-
{
|
| 407 |
-
"epoch": 11.43,
|
| 408 |
-
"learning_rate": 0.00013000000000000002,
|
| 409 |
-
"loss": 1.6767,
|
| 410 |
-
"step": 1040
|
| 411 |
-
},
|
| 412 |
-
{
|
| 413 |
-
"epoch": 11.64,
|
| 414 |
-
"learning_rate": 0.00013250000000000002,
|
| 415 |
-
"loss": 1.5184,
|
| 416 |
-
"step": 1060
|
| 417 |
-
},
|
| 418 |
-
{
|
| 419 |
-
"epoch": 11.86,
|
| 420 |
-
"learning_rate": 0.000135,
|
| 421 |
-
"loss": 1.4505,
|
| 422 |
-
"step": 1080
|
| 423 |
-
},
|
| 424 |
-
{
|
| 425 |
-
"epoch": 12.09,
|
| 426 |
-
"learning_rate": 0.0001375,
|
| 427 |
-
"loss": 1.4286,
|
| 428 |
-
"step": 1100
|
| 429 |
-
},
|
| 430 |
-
{
|
| 431 |
-
"epoch": 12.09,
|
| 432 |
-
"eval_loss": 0.4074769914150238,
|
| 433 |
-
"eval_runtime": 194.7798,
|
| 434 |
-
"eval_samples_per_second": 24.864,
|
| 435 |
-
"eval_steps_per_second": 0.78,
|
| 436 |
-
"eval_wer": 0.37964802724950325,
|
| 437 |
-
"step": 1100
|
| 438 |
-
},
|
| 439 |
-
{
|
| 440 |
-
"epoch": 12.31,
|
| 441 |
-
"learning_rate": 0.00014000000000000001,
|
| 442 |
-
"loss": 1.3478,
|
| 443 |
-
"step": 1120
|
| 444 |
-
},
|
| 445 |
-
{
|
| 446 |
-
"epoch": 12.52,
|
| 447 |
-
"learning_rate": 0.0001425,
|
| 448 |
-
"loss": 1.3164,
|
| 449 |
-
"step": 1140
|
| 450 |
-
},
|
| 451 |
-
{
|
| 452 |
-
"epoch": 12.74,
|
| 453 |
-
"learning_rate": 0.000145,
|
| 454 |
-
"loss": 1.269,
|
| 455 |
-
"step": 1160
|
| 456 |
-
},
|
| 457 |
-
{
|
| 458 |
-
"epoch": 12.96,
|
| 459 |
-
"learning_rate": 0.0001475,
|
| 460 |
-
"loss": 1.2292,
|
| 461 |
-
"step": 1180
|
| 462 |
-
},
|
| 463 |
-
{
|
| 464 |
-
"epoch": 13.19,
|
| 465 |
-
"learning_rate": 0.00015,
|
| 466 |
-
"loss": 1.2229,
|
| 467 |
-
"step": 1200
|
| 468 |
-
},
|
| 469 |
-
{
|
| 470 |
-
"epoch": 13.19,
|
| 471 |
-
"eval_loss": 0.28933778405189514,
|
| 472 |
-
"eval_runtime": 192.6398,
|
| 473 |
-
"eval_samples_per_second": 25.14,
|
| 474 |
-
"eval_steps_per_second": 0.789,
|
| 475 |
-
"eval_wer": 0.26520011353959694,
|
| 476 |
-
"step": 1200
|
| 477 |
-
},
|
| 478 |
-
{
|
| 479 |
-
"epoch": 13.4,
|
| 480 |
-
"learning_rate": 0.0001525,
|
| 481 |
-
"loss": 1.182,
|
| 482 |
-
"step": 1220
|
| 483 |
-
},
|
| 484 |
-
{
|
| 485 |
-
"epoch": 13.62,
|
| 486 |
-
"learning_rate": 0.000155,
|
| 487 |
-
"loss": 1.1768,
|
| 488 |
-
"step": 1240
|
| 489 |
-
},
|
| 490 |
-
{
|
| 491 |
-
"epoch": 13.84,
|
| 492 |
-
"learning_rate": 0.0001575,
|
| 493 |
-
"loss": 1.1534,
|
| 494 |
-
"step": 1260
|
| 495 |
-
},
|
| 496 |
-
{
|
| 497 |
-
"epoch": 14.07,
|
| 498 |
-
"learning_rate": 0.00016,
|
| 499 |
-
"loss": 1.1832,
|
| 500 |
-
"step": 1280
|
| 501 |
-
},
|
| 502 |
-
{
|
| 503 |
-
"epoch": 14.28,
|
| 504 |
-
"learning_rate": 0.00016250000000000002,
|
| 505 |
-
"loss": 1.1106,
|
| 506 |
-
"step": 1300
|
| 507 |
-
},
|
| 508 |
-
{
|
| 509 |
-
"epoch": 14.28,
|
| 510 |
-
"eval_loss": 0.24685777723789215,
|
| 511 |
-
"eval_runtime": 198.5067,
|
| 512 |
-
"eval_samples_per_second": 24.397,
|
| 513 |
-
"eval_steps_per_second": 0.766,
|
| 514 |
-
"eval_wer": 0.2253760999148453,
|
| 515 |
-
"step": 1300
|
| 516 |
-
},
|
| 517 |
-
{
|
| 518 |
-
"epoch": 14.5,
|
| 519 |
-
"learning_rate": 0.000165,
|
| 520 |
-
"loss": 1.1217,
|
| 521 |
-
"step": 1320
|
| 522 |
-
},
|
| 523 |
-
{
|
| 524 |
-
"epoch": 14.72,
|
| 525 |
-
"learning_rate": 0.0001675,
|
| 526 |
-
"loss": 1.1236,
|
| 527 |
-
"step": 1340
|
| 528 |
-
},
|
| 529 |
-
{
|
| 530 |
-
"epoch": 14.94,
|
| 531 |
-
"learning_rate": 0.00017,
|
| 532 |
-
"loss": 1.0969,
|
| 533 |
-
"step": 1360
|
| 534 |
-
},
|
| 535 |
-
{
|
| 536 |
-
"epoch": 15.16,
|
| 537 |
-
"learning_rate": 0.0001725,
|
| 538 |
-
"loss": 1.0971,
|
| 539 |
-
"step": 1380
|
| 540 |
-
},
|
| 541 |
-
{
|
| 542 |
-
"epoch": 15.38,
|
| 543 |
-
"learning_rate": 0.000175,
|
| 544 |
-
"loss": 1.0663,
|
| 545 |
-
"step": 1400
|
| 546 |
-
},
|
| 547 |
-
{
|
| 548 |
-
"epoch": 15.38,
|
| 549 |
-
"eval_loss": 0.22192780673503876,
|
| 550 |
-
"eval_runtime": 189.4566,
|
| 551 |
-
"eval_samples_per_second": 25.563,
|
| 552 |
-
"eval_steps_per_second": 0.802,
|
| 553 |
-
"eval_wer": 0.19727504967357365,
|
| 554 |
-
"step": 1400
|
| 555 |
-
},
|
| 556 |
-
{
|
| 557 |
-
"epoch": 15.6,
|
| 558 |
-
"learning_rate": 0.0001775,
|
| 559 |
-
"loss": 1.0733,
|
| 560 |
-
"step": 1420
|
| 561 |
-
},
|
| 562 |
-
{
|
| 563 |
-
"epoch": 15.82,
|
| 564 |
-
"learning_rate": 0.00017999999999999998,
|
| 565 |
-
"loss": 1.0635,
|
| 566 |
-
"step": 1440
|
| 567 |
-
},
|
| 568 |
-
{
|
| 569 |
-
"epoch": 16.04,
|
| 570 |
-
"learning_rate": 0.0001825,
|
| 571 |
-
"loss": 1.1079,
|
| 572 |
-
"step": 1460
|
| 573 |
-
},
|
| 574 |
-
{
|
| 575 |
-
"epoch": 16.26,
|
| 576 |
-
"learning_rate": 0.000185,
|
| 577 |
-
"loss": 1.0467,
|
| 578 |
-
"step": 1480
|
| 579 |
-
},
|
| 580 |
-
{
|
| 581 |
-
"epoch": 16.48,
|
| 582 |
-
"learning_rate": 0.0001875,
|
| 583 |
-
"loss": 1.0667,
|
| 584 |
-
"step": 1500
|
| 585 |
-
},
|
| 586 |
-
{
|
| 587 |
-
"epoch": 16.48,
|
| 588 |
-
"eval_loss": 0.21288961172103882,
|
| 589 |
-
"eval_runtime": 189.0078,
|
| 590 |
-
"eval_samples_per_second": 25.623,
|
| 591 |
-
"eval_steps_per_second": 0.804,
|
| 592 |
-
"eval_wer": 0.18944081748509792,
|
| 593 |
-
"step": 1500
|
| 594 |
-
},
|
| 595 |
-
{
|
| 596 |
-
"epoch": 16.7,
|
| 597 |
-
"learning_rate": 0.00019,
|
| 598 |
-
"loss": 1.0204,
|
| 599 |
-
"step": 1520
|
| 600 |
-
},
|
| 601 |
-
{
|
| 602 |
-
"epoch": 16.92,
|
| 603 |
-
"learning_rate": 0.00019250000000000002,
|
| 604 |
-
"loss": 1.0356,
|
| 605 |
-
"step": 1540
|
| 606 |
-
},
|
| 607 |
-
{
|
| 608 |
-
"epoch": 17.14,
|
| 609 |
-
"learning_rate": 0.00019500000000000002,
|
| 610 |
-
"loss": 1.0538,
|
| 611 |
-
"step": 1560
|
| 612 |
-
},
|
| 613 |
-
{
|
| 614 |
-
"epoch": 17.36,
|
| 615 |
-
"learning_rate": 0.0001975,
|
| 616 |
-
"loss": 1.0228,
|
| 617 |
-
"step": 1580
|
| 618 |
-
},
|
| 619 |
-
{
|
| 620 |
-
"epoch": 17.58,
|
| 621 |
-
"learning_rate": 0.0002,
|
| 622 |
-
"loss": 1.0193,
|
| 623 |
-
"step": 1600
|
| 624 |
-
},
|
| 625 |
-
{
|
| 626 |
-
"epoch": 17.58,
|
| 627 |
-
"eval_loss": 0.1991206556558609,
|
| 628 |
-
"eval_runtime": 191.1717,
|
| 629 |
-
"eval_samples_per_second": 25.333,
|
| 630 |
-
"eval_steps_per_second": 0.795,
|
| 631 |
-
"eval_wer": 0.17885325007096226,
|
| 632 |
-
"step": 1600
|
| 633 |
-
},
|
| 634 |
-
{
|
| 635 |
-
"epoch": 17.8,
|
| 636 |
-
"learning_rate": 0.00020250000000000002,
|
| 637 |
-
"loss": 1.0322,
|
| 638 |
-
"step": 1620
|
| 639 |
-
},
|
| 640 |
-
{
|
| 641 |
-
"epoch": 18.02,
|
| 642 |
-
"learning_rate": 0.000205,
|
| 643 |
-
"loss": 1.0176,
|
| 644 |
-
"step": 1640
|
| 645 |
-
},
|
| 646 |
-
{
|
| 647 |
-
"epoch": 18.24,
|
| 648 |
-
"learning_rate": 0.0002075,
|
| 649 |
-
"loss": 1.0272,
|
| 650 |
-
"step": 1660
|
| 651 |
-
},
|
| 652 |
-
{
|
| 653 |
-
"epoch": 18.46,
|
| 654 |
-
"learning_rate": 0.00021,
|
| 655 |
-
"loss": 0.9675,
|
| 656 |
-
"step": 1680
|
| 657 |
-
},
|
| 658 |
-
{
|
| 659 |
-
"epoch": 18.68,
|
| 660 |
-
"learning_rate": 0.0002125,
|
| 661 |
-
"loss": 0.9816,
|
| 662 |
-
"step": 1700
|
| 663 |
-
},
|
| 664 |
-
{
|
| 665 |
-
"epoch": 18.68,
|
| 666 |
-
"eval_loss": 0.19402356445789337,
|
| 667 |
-
"eval_runtime": 194.75,
|
| 668 |
-
"eval_samples_per_second": 24.868,
|
| 669 |
-
"eval_steps_per_second": 0.78,
|
| 670 |
-
"eval_wer": 0.18010218563724098,
|
| 671 |
-
"step": 1700
|
| 672 |
-
},
|
| 673 |
-
{
|
| 674 |
-
"epoch": 18.9,
|
| 675 |
-
"learning_rate": 0.000215,
|
| 676 |
-
"loss": 1.0023,
|
| 677 |
-
"step": 1720
|
| 678 |
-
},
|
| 679 |
-
{
|
| 680 |
-
"epoch": 19.12,
|
| 681 |
-
"learning_rate": 0.0002175,
|
| 682 |
-
"loss": 1.013,
|
| 683 |
-
"step": 1740
|
| 684 |
-
},
|
| 685 |
-
{
|
| 686 |
-
"epoch": 19.34,
|
| 687 |
-
"learning_rate": 0.00022,
|
| 688 |
-
"loss": 0.9664,
|
| 689 |
-
"step": 1760
|
| 690 |
-
},
|
| 691 |
-
{
|
| 692 |
-
"epoch": 19.56,
|
| 693 |
-
"learning_rate": 0.00022250000000000001,
|
| 694 |
-
"loss": 0.9736,
|
| 695 |
-
"step": 1780
|
| 696 |
-
},
|
| 697 |
-
{
|
| 698 |
-
"epoch": 19.78,
|
| 699 |
-
"learning_rate": 0.00022500000000000002,
|
| 700 |
-
"loss": 0.9814,
|
| 701 |
-
"step": 1800
|
| 702 |
-
},
|
| 703 |
-
{
|
| 704 |
-
"epoch": 19.78,
|
| 705 |
-
"eval_loss": 0.18596723675727844,
|
| 706 |
-
"eval_runtime": 188.4639,
|
| 707 |
-
"eval_samples_per_second": 25.697,
|
| 708 |
-
"eval_steps_per_second": 0.807,
|
| 709 |
-
"eval_wer": 0.16667612829974454,
|
| 710 |
-
"step": 1800
|
| 711 |
-
},
|
| 712 |
-
{
|
| 713 |
-
"epoch": 19.99,
|
| 714 |
-
"learning_rate": 0.0002275,
|
| 715 |
-
"loss": 1.0064,
|
| 716 |
-
"step": 1820
|
| 717 |
-
},
|
| 718 |
-
{
|
| 719 |
-
"epoch": 20.22,
|
| 720 |
-
"learning_rate": 0.00023,
|
| 721 |
-
"loss": 0.9583,
|
| 722 |
-
"step": 1840
|
| 723 |
-
},
|
| 724 |
-
{
|
| 725 |
-
"epoch": 20.44,
|
| 726 |
-
"learning_rate": 0.0002325,
|
| 727 |
-
"loss": 0.9646,
|
| 728 |
-
"step": 1860
|
| 729 |
-
},
|
| 730 |
-
{
|
| 731 |
-
"epoch": 20.66,
|
| 732 |
-
"learning_rate": 0.000235,
|
| 733 |
-
"loss": 0.9762,
|
| 734 |
-
"step": 1880
|
| 735 |
-
},
|
| 736 |
-
{
|
| 737 |
-
"epoch": 20.87,
|
| 738 |
-
"learning_rate": 0.0002375,
|
| 739 |
-
"loss": 0.9787,
|
| 740 |
-
"step": 1900
|
| 741 |
-
},
|
| 742 |
-
{
|
| 743 |
-
"epoch": 20.87,
|
| 744 |
-
"eval_loss": 0.18878202140331268,
|
| 745 |
-
"eval_runtime": 188.244,
|
| 746 |
-
"eval_samples_per_second": 25.727,
|
| 747 |
-
"eval_steps_per_second": 0.807,
|
| 748 |
-
"eval_wer": 0.16420664206642066,
|
| 749 |
-
"step": 1900
|
| 750 |
-
},
|
| 751 |
-
{
|
| 752 |
-
"epoch": 21.1,
|
| 753 |
-
"learning_rate": 0.00024,
|
| 754 |
-
"loss": 1.0218,
|
| 755 |
-
"step": 1920
|
| 756 |
-
},
|
| 757 |
-
{
|
| 758 |
-
"epoch": 21.32,
|
| 759 |
-
"learning_rate": 0.00024249999999999999,
|
| 760 |
-
"loss": 0.9505,
|
| 761 |
-
"step": 1940
|
| 762 |
-
},
|
| 763 |
-
{
|
| 764 |
-
"epoch": 21.54,
|
| 765 |
-
"learning_rate": 0.000245,
|
| 766 |
-
"loss": 0.9554,
|
| 767 |
-
"step": 1960
|
| 768 |
-
},
|
| 769 |
-
{
|
| 770 |
-
"epoch": 21.75,
|
| 771 |
-
"learning_rate": 0.0002475,
|
| 772 |
-
"loss": 0.9728,
|
| 773 |
-
"step": 1980
|
| 774 |
-
},
|
| 775 |
-
{
|
| 776 |
-
"epoch": 21.97,
|
| 777 |
-
"learning_rate": 0.00025,
|
| 778 |
-
"loss": 0.9699,
|
| 779 |
-
"step": 2000
|
| 780 |
-
},
|
| 781 |
-
{
|
| 782 |
-
"epoch": 21.97,
|
| 783 |
-
"eval_loss": 0.18748582899570465,
|
| 784 |
-
"eval_runtime": 190.7875,
|
| 785 |
-
"eval_samples_per_second": 25.384,
|
| 786 |
-
"eval_steps_per_second": 0.797,
|
| 787 |
-
"eval_wer": 0.17042293499858074,
|
| 788 |
-
"step": 2000
|
| 789 |
-
},
|
| 790 |
-
{
|
| 791 |
-
"epoch": 22.2,
|
| 792 |
-
"learning_rate": 0.00024803921568627453,
|
| 793 |
-
"loss": 0.9624,
|
| 794 |
-
"step": 2020
|
| 795 |
-
},
|
| 796 |
-
{
|
| 797 |
-
"epoch": 22.42,
|
| 798 |
-
"learning_rate": 0.000246078431372549,
|
| 799 |
-
"loss": 0.9419,
|
| 800 |
-
"step": 2040
|
| 801 |
-
},
|
| 802 |
-
{
|
| 803 |
-
"epoch": 22.63,
|
| 804 |
-
"learning_rate": 0.00024411764705882354,
|
| 805 |
-
"loss": 0.9563,
|
| 806 |
-
"step": 2060
|
| 807 |
-
},
|
| 808 |
-
{
|
| 809 |
-
"epoch": 22.85,
|
| 810 |
-
"learning_rate": 0.00024215686274509804,
|
| 811 |
-
"loss": 0.9643,
|
| 812 |
-
"step": 2080
|
| 813 |
-
},
|
| 814 |
-
{
|
| 815 |
-
"epoch": 23.08,
|
| 816 |
-
"learning_rate": 0.00024019607843137256,
|
| 817 |
-
"loss": 0.9616,
|
| 818 |
-
"step": 2100
|
| 819 |
-
},
|
| 820 |
-
{
|
| 821 |
-
"epoch": 23.08,
|
| 822 |
-
"eval_loss": 0.18017500638961792,
|
| 823 |
-
"eval_runtime": 191.3932,
|
| 824 |
-
"eval_samples_per_second": 25.304,
|
| 825 |
-
"eval_steps_per_second": 0.794,
|
| 826 |
-
"eval_wer": 0.16173715583309678,
|
| 827 |
-
"step": 2100
|
| 828 |
-
},
|
| 829 |
-
{
|
| 830 |
-
"epoch": 23.3,
|
| 831 |
-
"learning_rate": 0.00023823529411764704,
|
| 832 |
-
"loss": 0.917,
|
| 833 |
-
"step": 2120
|
| 834 |
-
},
|
| 835 |
-
{
|
| 836 |
-
"epoch": 23.51,
|
| 837 |
-
"learning_rate": 0.00023627450980392157,
|
| 838 |
-
"loss": 0.945,
|
| 839 |
-
"step": 2140
|
| 840 |
-
},
|
| 841 |
-
{
|
| 842 |
-
"epoch": 23.73,
|
| 843 |
-
"learning_rate": 0.0002343137254901961,
|
| 844 |
-
"loss": 0.9243,
|
| 845 |
-
"step": 2160
|
| 846 |
-
},
|
| 847 |
-
{
|
| 848 |
-
"epoch": 23.95,
|
| 849 |
-
"learning_rate": 0.0002323529411764706,
|
| 850 |
-
"loss": 0.9288,
|
| 851 |
-
"step": 2180
|
| 852 |
-
},
|
| 853 |
-
{
|
| 854 |
-
"epoch": 24.17,
|
| 855 |
-
"learning_rate": 0.0002303921568627451,
|
| 856 |
-
"loss": 0.9378,
|
| 857 |
-
"step": 2200
|
| 858 |
-
},
|
| 859 |
-
{
|
| 860 |
-
"epoch": 24.17,
|
| 861 |
-
"eval_loss": 0.17928896844387054,
|
| 862 |
-
"eval_runtime": 189.7619,
|
| 863 |
-
"eval_samples_per_second": 25.521,
|
| 864 |
-
"eval_steps_per_second": 0.801,
|
| 865 |
-
"eval_wer": 0.1577348850411581,
|
| 866 |
-
"step": 2200
|
| 867 |
-
},
|
| 868 |
-
{
|
| 869 |
-
"epoch": 24.39,
|
| 870 |
-
"learning_rate": 0.0002284313725490196,
|
| 871 |
-
"loss": 0.9071,
|
| 872 |
-
"step": 2220
|
| 873 |
-
},
|
| 874 |
-
{
|
| 875 |
-
"epoch": 24.61,
|
| 876 |
-
"learning_rate": 0.00022647058823529412,
|
| 877 |
-
"loss": 0.9054,
|
| 878 |
-
"step": 2240
|
| 879 |
-
},
|
| 880 |
-
{
|
| 881 |
-
"epoch": 24.83,
|
| 882 |
-
"learning_rate": 0.0002246078431372549,
|
| 883 |
-
"loss": 0.9303,
|
| 884 |
-
"step": 2260
|
| 885 |
-
},
|
| 886 |
-
{
|
| 887 |
-
"epoch": 25.05,
|
| 888 |
-
"learning_rate": 0.00022264705882352943,
|
| 889 |
-
"loss": 0.9376,
|
| 890 |
-
"step": 2280
|
| 891 |
-
},
|
| 892 |
-
{
|
| 893 |
-
"epoch": 25.27,
|
| 894 |
-
"learning_rate": 0.0002206862745098039,
|
| 895 |
-
"loss": 0.888,
|
| 896 |
-
"step": 2300
|
| 897 |
-
},
|
| 898 |
-
{
|
| 899 |
-
"epoch": 25.27,
|
| 900 |
-
"eval_loss": 0.17642559111118317,
|
| 901 |
-
"eval_runtime": 187.0437,
|
| 902 |
-
"eval_samples_per_second": 25.892,
|
| 903 |
-
"eval_steps_per_second": 0.813,
|
| 904 |
-
"eval_wer": 0.15452739142776042,
|
| 905 |
-
"step": 2300
|
| 906 |
-
},
|
| 907 |
-
{
|
| 908 |
-
"epoch": 25.49,
|
| 909 |
-
"learning_rate": 0.00021872549019607843,
|
| 910 |
-
"loss": 0.9135,
|
| 911 |
-
"step": 2320
|
| 912 |
-
},
|
| 913 |
-
{
|
| 914 |
-
"epoch": 25.71,
|
| 915 |
-
"learning_rate": 0.00021676470588235294,
|
| 916 |
-
"loss": 0.9094,
|
| 917 |
-
"step": 2340
|
| 918 |
-
},
|
| 919 |
-
{
|
| 920 |
-
"epoch": 25.93,
|
| 921 |
-
"learning_rate": 0.00021480392156862746,
|
| 922 |
-
"loss": 0.8879,
|
| 923 |
-
"step": 2360
|
| 924 |
-
},
|
| 925 |
-
{
|
| 926 |
-
"epoch": 26.15,
|
| 927 |
-
"learning_rate": 0.00021284313725490196,
|
| 928 |
-
"loss": 0.929,
|
| 929 |
-
"step": 2380
|
| 930 |
-
},
|
| 931 |
-
{
|
| 932 |
-
"epoch": 26.37,
|
| 933 |
-
"learning_rate": 0.00021088235294117647,
|
| 934 |
-
"loss": 0.8942,
|
| 935 |
-
"step": 2400
|
| 936 |
-
},
|
| 937 |
-
{
|
| 938 |
-
"epoch": 26.37,
|
| 939 |
-
"eval_loss": 0.16744859516620636,
|
| 940 |
-
"eval_runtime": 190.6796,
|
| 941 |
-
"eval_samples_per_second": 25.399,
|
| 942 |
-
"eval_steps_per_second": 0.797,
|
| 943 |
-
"eval_wer": 0.14916264547260857,
|
| 944 |
-
"step": 2400
|
| 945 |
-
},
|
| 946 |
-
{
|
| 947 |
-
"epoch": 26.59,
|
| 948 |
-
"learning_rate": 0.000208921568627451,
|
| 949 |
-
"loss": 0.8717,
|
| 950 |
-
"step": 2420
|
| 951 |
-
},
|
| 952 |
-
{
|
| 953 |
-
"epoch": 26.81,
|
| 954 |
-
"learning_rate": 0.0002069607843137255,
|
| 955 |
-
"loss": 0.8952,
|
| 956 |
-
"step": 2440
|
| 957 |
-
},
|
| 958 |
-
{
|
| 959 |
-
"epoch": 27.03,
|
| 960 |
-
"learning_rate": 0.000205,
|
| 961 |
-
"loss": 0.8849,
|
| 962 |
-
"step": 2460
|
| 963 |
-
},
|
| 964 |
-
{
|
| 965 |
-
"epoch": 27.25,
|
| 966 |
-
"learning_rate": 0.00020313725490196078,
|
| 967 |
-
"loss": 0.8781,
|
| 968 |
-
"step": 2480
|
| 969 |
-
},
|
| 970 |
-
{
|
| 971 |
-
"epoch": 27.47,
|
| 972 |
-
"learning_rate": 0.0002011764705882353,
|
| 973 |
-
"loss": 0.8701,
|
| 974 |
-
"step": 2500
|
| 975 |
-
},
|
| 976 |
-
{
|
| 977 |
-
"epoch": 27.47,
|
| 978 |
-
"eval_loss": 0.1738910973072052,
|
| 979 |
-
"eval_runtime": 187.7205,
|
| 980 |
-
"eval_samples_per_second": 25.799,
|
| 981 |
-
"eval_steps_per_second": 0.81,
|
| 982 |
-
"eval_wer": 0.1511779733181947,
|
| 983 |
-
"step": 2500
|
| 984 |
-
},
|
| 985 |
-
{
|
| 986 |
-
"epoch": 27.69,
|
| 987 |
-
"learning_rate": 0.0001992156862745098,
|
| 988 |
-
"loss": 0.8674,
|
| 989 |
-
"step": 2520
|
| 990 |
-
},
|
| 991 |
-
{
|
| 992 |
-
"epoch": 27.91,
|
| 993 |
-
"learning_rate": 0.00019725490196078433,
|
| 994 |
-
"loss": 0.8752,
|
| 995 |
-
"step": 2540
|
| 996 |
-
},
|
| 997 |
-
{
|
| 998 |
-
"epoch": 28.13,
|
| 999 |
-
"learning_rate": 0.0001952941176470588,
|
| 1000 |
-
"loss": 0.91,
|
| 1001 |
-
"step": 2560
|
| 1002 |
-
},
|
| 1003 |
-
{
|
| 1004 |
-
"epoch": 28.35,
|
| 1005 |
-
"learning_rate": 0.00019333333333333333,
|
| 1006 |
-
"loss": 0.8693,
|
| 1007 |
-
"step": 2580
|
| 1008 |
-
},
|
| 1009 |
-
{
|
| 1010 |
-
"epoch": 28.57,
|
| 1011 |
-
"learning_rate": 0.00019137254901960786,
|
| 1012 |
-
"loss": 0.8555,
|
| 1013 |
-
"step": 2600
|
| 1014 |
-
},
|
| 1015 |
-
{
|
| 1016 |
-
"epoch": 28.57,
|
| 1017 |
-
"eval_loss": 0.1689654141664505,
|
| 1018 |
-
"eval_runtime": 196.2665,
|
| 1019 |
-
"eval_samples_per_second": 24.676,
|
| 1020 |
-
"eval_steps_per_second": 0.774,
|
| 1021 |
-
"eval_wer": 0.14459267669599773,
|
| 1022 |
-
"step": 2600
|
| 1023 |
-
},
|
| 1024 |
-
{
|
| 1025 |
-
"epoch": 28.79,
|
| 1026 |
-
"learning_rate": 0.00018941176470588236,
|
| 1027 |
-
"loss": 0.8796,
|
| 1028 |
-
"step": 2620
|
| 1029 |
-
},
|
| 1030 |
-
{
|
| 1031 |
-
"epoch": 29.01,
|
| 1032 |
-
"learning_rate": 0.00018745098039215686,
|
| 1033 |
-
"loss": 0.9045,
|
| 1034 |
-
"step": 2640
|
| 1035 |
-
},
|
| 1036 |
-
{
|
| 1037 |
-
"epoch": 29.23,
|
| 1038 |
-
"learning_rate": 0.00018549019607843137,
|
| 1039 |
-
"loss": 0.8515,
|
| 1040 |
-
"step": 2660
|
| 1041 |
-
},
|
| 1042 |
-
{
|
| 1043 |
-
"epoch": 29.45,
|
| 1044 |
-
"learning_rate": 0.0001835294117647059,
|
| 1045 |
-
"loss": 0.861,
|
| 1046 |
-
"step": 2680
|
| 1047 |
-
},
|
| 1048 |
-
{
|
| 1049 |
-
"epoch": 29.67,
|
| 1050 |
-
"learning_rate": 0.0001815686274509804,
|
| 1051 |
-
"loss": 0.8513,
|
| 1052 |
-
"step": 2700
|
| 1053 |
-
},
|
| 1054 |
-
{
|
| 1055 |
-
"epoch": 29.67,
|
| 1056 |
-
"eval_loss": 0.16488835215568542,
|
| 1057 |
-
"eval_runtime": 189.1938,
|
| 1058 |
-
"eval_samples_per_second": 25.598,
|
| 1059 |
-
"eval_steps_per_second": 0.803,
|
| 1060 |
-
"eval_wer": 0.14774340051092819,
|
| 1061 |
-
"step": 2700
|
| 1062 |
-
},
|
| 1063 |
-
{
|
| 1064 |
-
"epoch": 29.89,
|
| 1065 |
-
"learning_rate": 0.0001796078431372549,
|
| 1066 |
-
"loss": 0.8573,
|
| 1067 |
-
"step": 2720
|
| 1068 |
-
},
|
| 1069 |
-
{
|
| 1070 |
-
"epoch": 30.11,
|
| 1071 |
-
"learning_rate": 0.00017764705882352942,
|
| 1072 |
-
"loss": 0.8686,
|
| 1073 |
-
"step": 2740
|
| 1074 |
-
},
|
| 1075 |
-
{
|
| 1076 |
-
"epoch": 30.33,
|
| 1077 |
-
"learning_rate": 0.00017568627450980392,
|
| 1078 |
-
"loss": 0.8462,
|
| 1079 |
-
"step": 2760
|
| 1080 |
-
},
|
| 1081 |
-
{
|
| 1082 |
-
"epoch": 30.55,
|
| 1083 |
-
"learning_rate": 0.00017372549019607845,
|
| 1084 |
-
"loss": 0.8419,
|
| 1085 |
-
"step": 2780
|
| 1086 |
-
},
|
| 1087 |
-
{
|
| 1088 |
-
"epoch": 30.77,
|
| 1089 |
-
"learning_rate": 0.00017176470588235293,
|
| 1090 |
-
"loss": 0.8659,
|
| 1091 |
-
"step": 2800
|
| 1092 |
-
},
|
| 1093 |
-
{
|
| 1094 |
-
"epoch": 30.77,
|
| 1095 |
-
"eval_loss": 0.1636602133512497,
|
| 1096 |
-
"eval_runtime": 190.245,
|
| 1097 |
-
"eval_samples_per_second": 25.457,
|
| 1098 |
-
"eval_steps_per_second": 0.799,
|
| 1099 |
-
"eval_wer": 0.1422367300596083,
|
| 1100 |
-
"step": 2800
|
| 1101 |
-
},
|
| 1102 |
-
{
|
| 1103 |
-
"epoch": 30.98,
|
| 1104 |
-
"learning_rate": 0.00016980392156862745,
|
| 1105 |
-
"loss": 0.8607,
|
| 1106 |
-
"step": 2820
|
| 1107 |
-
},
|
| 1108 |
-
{
|
| 1109 |
-
"epoch": 31.21,
|
| 1110 |
-
"learning_rate": 0.00016784313725490196,
|
| 1111 |
-
"loss": 0.8484,
|
| 1112 |
-
"step": 2840
|
| 1113 |
-
},
|
| 1114 |
-
{
|
| 1115 |
-
"epoch": 31.43,
|
| 1116 |
-
"learning_rate": 0.00016588235294117648,
|
| 1117 |
-
"loss": 0.8335,
|
| 1118 |
-
"step": 2860
|
| 1119 |
-
},
|
| 1120 |
-
{
|
| 1121 |
-
"epoch": 31.64,
|
| 1122 |
-
"learning_rate": 0.00016392156862745098,
|
| 1123 |
-
"loss": 0.8321,
|
| 1124 |
-
"step": 2880
|
| 1125 |
-
},
|
| 1126 |
-
{
|
| 1127 |
-
"epoch": 31.86,
|
| 1128 |
-
"learning_rate": 0.00016196078431372549,
|
| 1129 |
-
"loss": 0.8419,
|
| 1130 |
-
"step": 2900
|
| 1131 |
-
},
|
| 1132 |
-
{
|
| 1133 |
-
"epoch": 31.86,
|
| 1134 |
-
"eval_loss": 0.16137854754924774,
|
| 1135 |
-
"eval_runtime": 190.2053,
|
| 1136 |
-
"eval_samples_per_second": 25.462,
|
| 1137 |
-
"eval_steps_per_second": 0.799,
|
| 1138 |
-
"eval_wer": 0.1397104740278172,
|
| 1139 |
-
"step": 2900
|
| 1140 |
-
},
|
| 1141 |
-
{
|
| 1142 |
-
"epoch": 32.09,
|
| 1143 |
-
"learning_rate": 0.00016,
|
| 1144 |
-
"loss": 0.8634,
|
| 1145 |
-
"step": 2920
|
| 1146 |
-
},
|
| 1147 |
-
{
|
| 1148 |
-
"epoch": 32.31,
|
| 1149 |
-
"learning_rate": 0.00015803921568627451,
|
| 1150 |
-
"loss": 0.8351,
|
| 1151 |
-
"step": 2940
|
| 1152 |
-
},
|
| 1153 |
-
{
|
| 1154 |
-
"epoch": 32.52,
|
| 1155 |
-
"learning_rate": 0.00015607843137254901,
|
| 1156 |
-
"loss": 0.8427,
|
| 1157 |
-
"step": 2960
|
| 1158 |
-
},
|
| 1159 |
-
{
|
| 1160 |
-
"epoch": 32.74,
|
| 1161 |
-
"learning_rate": 0.00015411764705882352,
|
| 1162 |
-
"loss": 0.8324,
|
| 1163 |
-
"step": 2980
|
| 1164 |
-
},
|
| 1165 |
-
{
|
| 1166 |
-
"epoch": 32.96,
|
| 1167 |
-
"learning_rate": 0.00015215686274509804,
|
| 1168 |
-
"loss": 0.8491,
|
| 1169 |
-
"step": 3000
|
| 1170 |
-
},
|
| 1171 |
-
{
|
| 1172 |
-
"epoch": 32.96,
|
| 1173 |
-
"eval_loss": 0.1594749242067337,
|
| 1174 |
-
"eval_runtime": 191.4631,
|
| 1175 |
-
"eval_samples_per_second": 25.295,
|
| 1176 |
-
"eval_steps_per_second": 0.794,
|
| 1177 |
-
"eval_wer": 0.1401362475163213,
|
| 1178 |
-
"step": 3000
|
| 1179 |
-
},
|
| 1180 |
-
{
|
| 1181 |
-
"epoch": 33.19,
|
| 1182 |
-
"learning_rate": 0.00015019607843137257,
|
| 1183 |
-
"loss": 0.8263,
|
| 1184 |
-
"step": 3020
|
| 1185 |
-
},
|
| 1186 |
-
{
|
| 1187 |
-
"epoch": 33.4,
|
| 1188 |
-
"learning_rate": 0.00014823529411764705,
|
| 1189 |
-
"loss": 0.8203,
|
| 1190 |
-
"step": 3040
|
| 1191 |
-
},
|
| 1192 |
-
{
|
| 1193 |
-
"epoch": 33.62,
|
| 1194 |
-
"learning_rate": 0.00014627450980392157,
|
| 1195 |
-
"loss": 0.8277,
|
| 1196 |
-
"step": 3060
|
| 1197 |
-
},
|
| 1198 |
-
{
|
| 1199 |
-
"epoch": 33.84,
|
| 1200 |
-
"learning_rate": 0.00014431372549019607,
|
| 1201 |
-
"loss": 0.8183,
|
| 1202 |
-
"step": 3080
|
| 1203 |
-
},
|
| 1204 |
-
{
|
| 1205 |
-
"epoch": 34.07,
|
| 1206 |
-
"learning_rate": 0.0001423529411764706,
|
| 1207 |
-
"loss": 0.8395,
|
| 1208 |
-
"step": 3100
|
| 1209 |
-
},
|
| 1210 |
-
{
|
| 1211 |
-
"epoch": 34.07,
|
| 1212 |
-
"eval_loss": 0.1606692373752594,
|
| 1213 |
-
"eval_runtime": 190.8222,
|
| 1214 |
-
"eval_samples_per_second": 25.38,
|
| 1215 |
-
"eval_steps_per_second": 0.797,
|
| 1216 |
-
"eval_wer": 0.13755322168606302,
|
| 1217 |
-
"step": 3100
|
| 1218 |
-
},
|
| 1219 |
-
{
|
| 1220 |
-
"epoch": 34.28,
|
| 1221 |
-
"learning_rate": 0.0001403921568627451,
|
| 1222 |
-
"loss": 0.8207,
|
| 1223 |
-
"step": 3120
|
| 1224 |
-
},
|
| 1225 |
-
{
|
| 1226 |
-
"epoch": 34.5,
|
| 1227 |
-
"learning_rate": 0.0001384313725490196,
|
| 1228 |
-
"loss": 0.8106,
|
| 1229 |
-
"step": 3140
|
| 1230 |
-
},
|
| 1231 |
-
{
|
| 1232 |
-
"epoch": 34.72,
|
| 1233 |
-
"learning_rate": 0.00013647058823529413,
|
| 1234 |
-
"loss": 0.8329,
|
| 1235 |
-
"step": 3160
|
| 1236 |
-
},
|
| 1237 |
-
{
|
| 1238 |
-
"epoch": 34.94,
|
| 1239 |
-
"learning_rate": 0.00013450980392156863,
|
| 1240 |
-
"loss": 0.8048,
|
| 1241 |
-
"step": 3180
|
| 1242 |
-
},
|
| 1243 |
-
{
|
| 1244 |
-
"epoch": 35.16,
|
| 1245 |
-
"learning_rate": 0.00013254901960784313,
|
| 1246 |
-
"loss": 0.83,
|
| 1247 |
-
"step": 3200
|
| 1248 |
-
},
|
| 1249 |
-
{
|
| 1250 |
-
"epoch": 35.16,
|
| 1251 |
-
"eval_loss": 0.15384173393249512,
|
| 1252 |
-
"eval_runtime": 191.1912,
|
| 1253 |
-
"eval_samples_per_second": 25.331,
|
| 1254 |
-
"eval_steps_per_second": 0.795,
|
| 1255 |
-
"eval_wer": 0.1379222253760999,
|
| 1256 |
-
"step": 3200
|
| 1257 |
-
},
|
| 1258 |
-
{
|
| 1259 |
-
"epoch": 35.38,
|
| 1260 |
-
"learning_rate": 0.00013058823529411764,
|
| 1261 |
-
"loss": 0.7963,
|
| 1262 |
-
"step": 3220
|
| 1263 |
-
},
|
| 1264 |
-
{
|
| 1265 |
-
"epoch": 35.6,
|
| 1266 |
-
"learning_rate": 0.00012862745098039216,
|
| 1267 |
-
"loss": 0.7895,
|
| 1268 |
-
"step": 3240
|
| 1269 |
-
},
|
| 1270 |
-
{
|
| 1271 |
-
"epoch": 35.82,
|
| 1272 |
-
"learning_rate": 0.0001266666666666667,
|
| 1273 |
-
"loss": 0.7964,
|
| 1274 |
-
"step": 3260
|
| 1275 |
-
},
|
| 1276 |
-
{
|
| 1277 |
-
"epoch": 36.04,
|
| 1278 |
-
"learning_rate": 0.0001247058823529412,
|
| 1279 |
-
"loss": 0.7931,
|
| 1280 |
-
"step": 3280
|
| 1281 |
-
},
|
| 1282 |
-
{
|
| 1283 |
-
"epoch": 36.26,
|
| 1284 |
-
"learning_rate": 0.0001227450980392157,
|
| 1285 |
-
"loss": 0.7835,
|
| 1286 |
-
"step": 3300
|
| 1287 |
-
},
|
| 1288 |
-
{
|
| 1289 |
-
"epoch": 36.26,
|
| 1290 |
-
"eval_loss": 0.16020993888378143,
|
| 1291 |
-
"eval_runtime": 192.1405,
|
| 1292 |
-
"eval_samples_per_second": 25.206,
|
| 1293 |
-
"eval_steps_per_second": 0.791,
|
| 1294 |
-
"eval_wer": 0.1408458699971615,
|
| 1295 |
-
"step": 3300
|
| 1296 |
-
},
|
| 1297 |
-
{
|
| 1298 |
-
"epoch": 36.48,
|
| 1299 |
-
"learning_rate": 0.00012078431372549021,
|
| 1300 |
-
"loss": 0.772,
|
| 1301 |
-
"step": 3320
|
| 1302 |
-
},
|
| 1303 |
-
{
|
| 1304 |
-
"epoch": 36.7,
|
| 1305 |
-
"learning_rate": 0.00011882352941176471,
|
| 1306 |
-
"loss": 0.7818,
|
| 1307 |
-
"step": 3340
|
| 1308 |
-
},
|
| 1309 |
-
{
|
| 1310 |
-
"epoch": 36.92,
|
| 1311 |
-
"learning_rate": 0.00011696078431372549,
|
| 1312 |
-
"loss": 0.8016,
|
| 1313 |
-
"step": 3360
|
| 1314 |
-
},
|
| 1315 |
-
{
|
| 1316 |
-
"epoch": 37.14,
|
| 1317 |
-
"learning_rate": 0.000115,
|
| 1318 |
-
"loss": 0.8061,
|
| 1319 |
-
"step": 3380
|
| 1320 |
-
},
|
| 1321 |
-
{
|
| 1322 |
-
"epoch": 37.36,
|
| 1323 |
-
"learning_rate": 0.0001130392156862745,
|
| 1324 |
-
"loss": 0.7703,
|
| 1325 |
-
"step": 3400
|
| 1326 |
-
},
|
| 1327 |
-
{
|
| 1328 |
-
"epoch": 37.36,
|
| 1329 |
-
"eval_loss": 0.16011376678943634,
|
| 1330 |
-
"eval_runtime": 187.5367,
|
| 1331 |
-
"eval_samples_per_second": 25.824,
|
| 1332 |
-
"eval_steps_per_second": 0.811,
|
| 1333 |
-
"eval_wer": 0.13692875390292364,
|
| 1334 |
-
"step": 3400
|
| 1335 |
-
},
|
| 1336 |
-
{
|
| 1337 |
-
"epoch": 37.58,
|
| 1338 |
-
"learning_rate": 0.00011107843137254903,
|
| 1339 |
-
"loss": 0.7713,
|
| 1340 |
-
"step": 3420
|
| 1341 |
-
},
|
| 1342 |
-
{
|
| 1343 |
-
"epoch": 37.8,
|
| 1344 |
-
"learning_rate": 0.00010911764705882353,
|
| 1345 |
-
"loss": 0.7712,
|
| 1346 |
-
"step": 3440
|
| 1347 |
-
},
|
| 1348 |
-
{
|
| 1349 |
-
"epoch": 38.02,
|
| 1350 |
-
"learning_rate": 0.00010715686274509805,
|
| 1351 |
-
"loss": 0.8149,
|
| 1352 |
-
"step": 3460
|
| 1353 |
-
},
|
| 1354 |
-
{
|
| 1355 |
-
"epoch": 38.24,
|
| 1356 |
-
"learning_rate": 0.00010519607843137255,
|
| 1357 |
-
"loss": 0.7885,
|
| 1358 |
-
"step": 3480
|
| 1359 |
-
},
|
| 1360 |
-
{
|
| 1361 |
-
"epoch": 38.46,
|
| 1362 |
-
"learning_rate": 0.00010323529411764706,
|
| 1363 |
-
"loss": 0.7474,
|
| 1364 |
-
"step": 3500
|
| 1365 |
-
},
|
| 1366 |
-
{
|
| 1367 |
-
"epoch": 38.46,
|
| 1368 |
-
"eval_loss": 0.1514146625995636,
|
| 1369 |
-
"eval_runtime": 190.8228,
|
| 1370 |
-
"eval_samples_per_second": 25.38,
|
| 1371 |
-
"eval_steps_per_second": 0.797,
|
| 1372 |
-
"eval_wer": 0.1342321884757309,
|
| 1373 |
-
"step": 3500
|
| 1374 |
-
},
|
| 1375 |
-
{
|
| 1376 |
-
"epoch": 38.68,
|
| 1377 |
-
"learning_rate": 0.00010127450980392156,
|
| 1378 |
-
"loss": 0.785,
|
| 1379 |
-
"step": 3520
|
| 1380 |
-
},
|
| 1381 |
-
{
|
| 1382 |
-
"epoch": 38.9,
|
| 1383 |
-
"learning_rate": 9.931372549019609e-05,
|
| 1384 |
-
"loss": 0.7677,
|
| 1385 |
-
"step": 3540
|
| 1386 |
-
},
|
| 1387 |
-
{
|
| 1388 |
-
"epoch": 39.12,
|
| 1389 |
-
"learning_rate": 9.73529411764706e-05,
|
| 1390 |
-
"loss": 0.7849,
|
| 1391 |
-
"step": 3560
|
| 1392 |
-
},
|
| 1393 |
-
{
|
| 1394 |
-
"epoch": 39.34,
|
| 1395 |
-
"learning_rate": 9.539215686274511e-05,
|
| 1396 |
-
"loss": 0.7637,
|
| 1397 |
-
"step": 3580
|
| 1398 |
-
},
|
| 1399 |
-
{
|
| 1400 |
-
"epoch": 39.56,
|
| 1401 |
-
"learning_rate": 9.343137254901961e-05,
|
| 1402 |
-
"loss": 0.7719,
|
| 1403 |
-
"step": 3600
|
| 1404 |
-
},
|
| 1405 |
-
{
|
| 1406 |
-
"epoch": 39.56,
|
| 1407 |
-
"eval_loss": 0.15932896733283997,
|
| 1408 |
-
"eval_runtime": 189.6806,
|
| 1409 |
-
"eval_samples_per_second": 25.532,
|
| 1410 |
-
"eval_steps_per_second": 0.801,
|
| 1411 |
-
"eval_wer": 0.1352540448481408,
|
| 1412 |
-
"step": 3600
|
| 1413 |
-
},
|
| 1414 |
-
{
|
| 1415 |
-
"epoch": 39.78,
|
| 1416 |
-
"learning_rate": 9.147058823529412e-05,
|
| 1417 |
-
"loss": 0.7591,
|
| 1418 |
-
"step": 3620
|
| 1419 |
-
},
|
| 1420 |
-
{
|
| 1421 |
-
"epoch": 39.99,
|
| 1422 |
-
"learning_rate": 8.950980392156862e-05,
|
| 1423 |
-
"loss": 0.7706,
|
| 1424 |
-
"step": 3640
|
| 1425 |
-
},
|
| 1426 |
-
{
|
| 1427 |
-
"epoch": 40.22,
|
| 1428 |
-
"learning_rate": 8.754901960784314e-05,
|
| 1429 |
-
"loss": 0.7805,
|
| 1430 |
-
"step": 3660
|
| 1431 |
-
},
|
| 1432 |
-
{
|
| 1433 |
-
"epoch": 40.44,
|
| 1434 |
-
"learning_rate": 8.558823529411765e-05,
|
| 1435 |
-
"loss": 0.7753,
|
| 1436 |
-
"step": 3680
|
| 1437 |
-
},
|
| 1438 |
-
{
|
| 1439 |
-
"epoch": 40.66,
|
| 1440 |
-
"learning_rate": 8.362745098039217e-05,
|
| 1441 |
-
"loss": 0.7638,
|
| 1442 |
-
"step": 3700
|
| 1443 |
-
},
|
| 1444 |
-
{
|
| 1445 |
-
"epoch": 40.66,
|
| 1446 |
-
"eval_loss": 0.15362653136253357,
|
| 1447 |
-
"eval_runtime": 192.1118,
|
| 1448 |
-
"eval_samples_per_second": 25.209,
|
| 1449 |
-
"eval_steps_per_second": 0.791,
|
| 1450 |
-
"eval_wer": 0.13380641498722678,
|
| 1451 |
-
"step": 3700
|
| 1452 |
-
},
|
| 1453 |
-
{
|
| 1454 |
-
"epoch": 40.87,
|
| 1455 |
-
"learning_rate": 8.166666666666667e-05,
|
| 1456 |
-
"loss": 0.7791,
|
| 1457 |
-
"step": 3720
|
| 1458 |
-
},
|
| 1459 |
-
{
|
| 1460 |
-
"epoch": 41.1,
|
| 1461 |
-
"learning_rate": 7.970588235294118e-05,
|
| 1462 |
-
"loss": 0.7931,
|
| 1463 |
-
"step": 3740
|
| 1464 |
-
},
|
| 1465 |
-
{
|
| 1466 |
-
"epoch": 41.32,
|
| 1467 |
-
"learning_rate": 7.774509803921568e-05,
|
| 1468 |
-
"loss": 0.741,
|
| 1469 |
-
"step": 3760
|
| 1470 |
-
},
|
| 1471 |
-
{
|
| 1472 |
-
"epoch": 41.54,
|
| 1473 |
-
"learning_rate": 7.57843137254902e-05,
|
| 1474 |
-
"loss": 0.7434,
|
| 1475 |
-
"step": 3780
|
| 1476 |
-
},
|
| 1477 |
-
{
|
| 1478 |
-
"epoch": 41.75,
|
| 1479 |
-
"learning_rate": 7.38235294117647e-05,
|
| 1480 |
-
"loss": 0.771,
|
| 1481 |
-
"step": 3800
|
| 1482 |
-
},
|
| 1483 |
-
{
|
| 1484 |
-
"epoch": 41.75,
|
| 1485 |
-
"eval_loss": 0.1530592143535614,
|
| 1486 |
-
"eval_runtime": 186.414,
|
| 1487 |
-
"eval_samples_per_second": 25.98,
|
| 1488 |
-
"eval_steps_per_second": 0.815,
|
| 1489 |
-
"eval_wer": 0.13170593244393983,
|
| 1490 |
-
"step": 3800
|
| 1491 |
-
},
|
| 1492 |
-
{
|
| 1493 |
-
"epoch": 41.97,
|
| 1494 |
-
"learning_rate": 7.186274509803923e-05,
|
| 1495 |
-
"loss": 0.7765,
|
| 1496 |
-
"step": 3820
|
| 1497 |
-
},
|
| 1498 |
-
{
|
| 1499 |
-
"epoch": 42.2,
|
| 1500 |
-
"learning_rate": 6.990196078431373e-05,
|
| 1501 |
-
"loss": 0.7599,
|
| 1502 |
-
"step": 3840
|
| 1503 |
-
},
|
| 1504 |
-
{
|
| 1505 |
-
"epoch": 42.42,
|
| 1506 |
-
"learning_rate": 6.794117647058824e-05,
|
| 1507 |
-
"loss": 0.7782,
|
| 1508 |
-
"step": 3860
|
| 1509 |
-
},
|
| 1510 |
-
{
|
| 1511 |
-
"epoch": 42.63,
|
| 1512 |
-
"learning_rate": 6.598039215686274e-05,
|
| 1513 |
-
"loss": 0.7395,
|
| 1514 |
-
"step": 3880
|
| 1515 |
-
},
|
| 1516 |
-
{
|
| 1517 |
-
"epoch": 42.85,
|
| 1518 |
-
"learning_rate": 6.401960784313726e-05,
|
| 1519 |
-
"loss": 0.7594,
|
| 1520 |
-
"step": 3900
|
| 1521 |
-
},
|
| 1522 |
-
{
|
| 1523 |
-
"epoch": 42.85,
|
| 1524 |
-
"eval_loss": 0.14983513951301575,
|
| 1525 |
-
"eval_runtime": 190.4439,
|
| 1526 |
-
"eval_samples_per_second": 25.43,
|
| 1527 |
-
"eval_steps_per_second": 0.798,
|
| 1528 |
-
"eval_wer": 0.12883905762134545,
|
| 1529 |
-
"step": 3900
|
| 1530 |
-
},
|
| 1531 |
-
{
|
| 1532 |
-
"epoch": 43.08,
|
| 1533 |
-
"learning_rate": 6.205882352941176e-05,
|
| 1534 |
-
"loss": 0.7691,
|
| 1535 |
-
"step": 3920
|
| 1536 |
-
},
|
| 1537 |
-
{
|
| 1538 |
-
"epoch": 43.3,
|
| 1539 |
-
"learning_rate": 6.0098039215686274e-05,
|
| 1540 |
-
"loss": 0.728,
|
| 1541 |
-
"step": 3940
|
| 1542 |
-
},
|
| 1543 |
-
{
|
| 1544 |
-
"epoch": 43.51,
|
| 1545 |
-
"learning_rate": 5.813725490196078e-05,
|
| 1546 |
-
"loss": 0.7585,
|
| 1547 |
-
"step": 3960
|
| 1548 |
-
},
|
| 1549 |
-
{
|
| 1550 |
-
"epoch": 43.73,
|
| 1551 |
-
"learning_rate": 5.6176470588235296e-05,
|
| 1552 |
-
"loss": 0.7564,
|
| 1553 |
-
"step": 3980
|
| 1554 |
-
},
|
| 1555 |
-
{
|
| 1556 |
-
"epoch": 43.95,
|
| 1557 |
-
"learning_rate": 5.4215686274509804e-05,
|
| 1558 |
-
"loss": 0.7383,
|
| 1559 |
-
"step": 4000
|
| 1560 |
-
},
|
| 1561 |
-
{
|
| 1562 |
-
"epoch": 43.95,
|
| 1563 |
-
"eval_loss": 0.15269021689891815,
|
| 1564 |
-
"eval_runtime": 188.5816,
|
| 1565 |
-
"eval_samples_per_second": 25.681,
|
| 1566 |
-
"eval_steps_per_second": 0.806,
|
| 1567 |
-
"eval_wer": 0.13003122338915696,
|
| 1568 |
-
"step": 4000
|
| 1569 |
-
},
|
| 1570 |
-
{
|
| 1571 |
-
"epoch": 44.17,
|
| 1572 |
-
"learning_rate": 5.225490196078431e-05,
|
| 1573 |
-
"loss": 0.7693,
|
| 1574 |
-
"step": 4020
|
| 1575 |
-
},
|
| 1576 |
-
{
|
| 1577 |
-
"epoch": 44.39,
|
| 1578 |
-
"learning_rate": 5.0294117647058826e-05,
|
| 1579 |
-
"loss": 0.7347,
|
| 1580 |
-
"step": 4040
|
| 1581 |
-
},
|
| 1582 |
-
{
|
| 1583 |
-
"epoch": 44.61,
|
| 1584 |
-
"learning_rate": 4.8333333333333334e-05,
|
| 1585 |
-
"loss": 0.7185,
|
| 1586 |
-
"step": 4060
|
| 1587 |
-
},
|
| 1588 |
-
{
|
| 1589 |
-
"epoch": 44.83,
|
| 1590 |
-
"learning_rate": 4.637254901960784e-05,
|
| 1591 |
-
"loss": 0.7394,
|
| 1592 |
-
"step": 4080
|
| 1593 |
-
},
|
| 1594 |
-
{
|
| 1595 |
-
"epoch": 45.05,
|
| 1596 |
-
"learning_rate": 4.4411764705882356e-05,
|
| 1597 |
-
"loss": 0.7565,
|
| 1598 |
-
"step": 4100
|
| 1599 |
-
},
|
| 1600 |
-
{
|
| 1601 |
-
"epoch": 45.05,
|
| 1602 |
-
"eval_loss": 0.14823457598686218,
|
| 1603 |
-
"eval_runtime": 191.1254,
|
| 1604 |
-
"eval_samples_per_second": 25.339,
|
| 1605 |
-
"eval_steps_per_second": 0.795,
|
| 1606 |
-
"eval_wer": 0.12886744252057905,
|
| 1607 |
-
"step": 4100
|
| 1608 |
-
},
|
| 1609 |
-
{
|
| 1610 |
-
"epoch": 45.27,
|
| 1611 |
-
"learning_rate": 4.2450980392156864e-05,
|
| 1612 |
-
"loss": 0.7535,
|
| 1613 |
-
"step": 4120
|
| 1614 |
-
},
|
| 1615 |
-
{
|
| 1616 |
-
"epoch": 45.49,
|
| 1617 |
-
"learning_rate": 4.049019607843137e-05,
|
| 1618 |
-
"loss": 0.737,
|
| 1619 |
-
"step": 4140
|
| 1620 |
-
},
|
| 1621 |
-
{
|
| 1622 |
-
"epoch": 45.71,
|
| 1623 |
-
"learning_rate": 3.852941176470588e-05,
|
| 1624 |
-
"loss": 0.7501,
|
| 1625 |
-
"step": 4160
|
| 1626 |
-
},
|
| 1627 |
-
{
|
| 1628 |
-
"epoch": 45.93,
|
| 1629 |
-
"learning_rate": 3.6568627450980393e-05,
|
| 1630 |
-
"loss": 0.7285,
|
| 1631 |
-
"step": 4180
|
| 1632 |
-
},
|
| 1633 |
-
{
|
| 1634 |
-
"epoch": 46.15,
|
| 1635 |
-
"learning_rate": 3.46078431372549e-05,
|
| 1636 |
-
"loss": 0.7697,
|
| 1637 |
-
"step": 4200
|
| 1638 |
-
},
|
| 1639 |
-
{
|
| 1640 |
-
"epoch": 46.15,
|
| 1641 |
-
"eval_loss": 0.1494804471731186,
|
| 1642 |
-
"eval_runtime": 190.3868,
|
| 1643 |
-
"eval_samples_per_second": 25.438,
|
| 1644 |
-
"eval_steps_per_second": 0.798,
|
| 1645 |
-
"eval_wer": 0.1271927334657962,
|
| 1646 |
-
"step": 4200
|
| 1647 |
-
},
|
| 1648 |
-
{
|
| 1649 |
-
"epoch": 46.37,
|
| 1650 |
-
"learning_rate": 3.264705882352941e-05,
|
| 1651 |
-
"loss": 0.7283,
|
| 1652 |
-
"step": 4220
|
| 1653 |
-
},
|
| 1654 |
-
{
|
| 1655 |
-
"epoch": 46.59,
|
| 1656 |
-
"learning_rate": 3.0686274509803923e-05,
|
| 1657 |
-
"loss": 0.7218,
|
| 1658 |
-
"step": 4240
|
| 1659 |
-
},
|
| 1660 |
-
{
|
| 1661 |
-
"epoch": 46.81,
|
| 1662 |
-
"learning_rate": 2.872549019607843e-05,
|
| 1663 |
-
"loss": 0.7341,
|
| 1664 |
-
"step": 4260
|
| 1665 |
-
},
|
| 1666 |
-
{
|
| 1667 |
-
"epoch": 47.03,
|
| 1668 |
-
"learning_rate": 2.6764705882352942e-05,
|
| 1669 |
-
"loss": 0.7293,
|
| 1670 |
-
"step": 4280
|
| 1671 |
-
},
|
| 1672 |
-
{
|
| 1673 |
-
"epoch": 47.25,
|
| 1674 |
-
"learning_rate": 2.480392156862745e-05,
|
| 1675 |
-
"loss": 0.7194,
|
| 1676 |
-
"step": 4300
|
| 1677 |
-
},
|
| 1678 |
-
{
|
| 1679 |
-
"epoch": 47.25,
|
| 1680 |
-
"eval_loss": 0.14928147196769714,
|
| 1681 |
-
"eval_runtime": 189.625,
|
| 1682 |
-
"eval_samples_per_second": 25.54,
|
| 1683 |
-
"eval_steps_per_second": 0.802,
|
| 1684 |
-
"eval_wer": 0.12693726937269373,
|
| 1685 |
-
"step": 4300
|
| 1686 |
-
},
|
| 1687 |
-
{
|
| 1688 |
-
"epoch": 47.47,
|
| 1689 |
-
"learning_rate": 2.284313725490196e-05,
|
| 1690 |
-
"loss": 0.7326,
|
| 1691 |
-
"step": 4320
|
| 1692 |
-
},
|
| 1693 |
-
{
|
| 1694 |
-
"epoch": 47.69,
|
| 1695 |
-
"learning_rate": 2.088235294117647e-05,
|
| 1696 |
-
"loss": 0.724,
|
| 1697 |
-
"step": 4340
|
| 1698 |
-
},
|
| 1699 |
-
{
|
| 1700 |
-
"epoch": 47.91,
|
| 1701 |
-
"learning_rate": 1.892156862745098e-05,
|
| 1702 |
-
"loss": 0.7316,
|
| 1703 |
-
"step": 4360
|
| 1704 |
-
},
|
| 1705 |
-
{
|
| 1706 |
-
"epoch": 48.13,
|
| 1707 |
-
"learning_rate": 1.696078431372549e-05,
|
| 1708 |
-
"loss": 0.7566,
|
| 1709 |
-
"step": 4380
|
| 1710 |
-
},
|
| 1711 |
-
{
|
| 1712 |
-
"epoch": 48.35,
|
| 1713 |
-
"learning_rate": 1.5e-05,
|
| 1714 |
-
"loss": 0.7479,
|
| 1715 |
-
"step": 4400
|
| 1716 |
-
},
|
| 1717 |
-
{
|
| 1718 |
-
"epoch": 48.35,
|
| 1719 |
-
"eval_loss": 0.1490115076303482,
|
| 1720 |
-
"eval_runtime": 188.3681,
|
| 1721 |
-
"eval_samples_per_second": 25.71,
|
| 1722 |
-
"eval_steps_per_second": 0.807,
|
| 1723 |
-
"eval_wer": 0.12761850695430033,
|
| 1724 |
-
"step": 4400
|
| 1725 |
-
}
|
| 1726 |
-
],
|
| 1727 |
-
"max_steps": 4550,
|
| 1728 |
-
"num_train_epochs": 50,
|
| 1729 |
-
"total_flos": 6.805564586486029e+19,
|
| 1730 |
-
"trial_name": null,
|
| 1731 |
-
"trial_params": null
|
| 1732 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4400/training_args.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:61c53503eb760641941023d2dcab7bb32b620bf895e31db5c3910b2760135b07
|
| 3 |
-
size 3055
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/config.json
DELETED
|
@@ -1,115 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_name_or_path": "marinone94/xls-r-300m-sv-robust",
|
| 3 |
-
"activation_dropout": 0.1,
|
| 4 |
-
"adapter_kernel_size": 3,
|
| 5 |
-
"adapter_stride": 2,
|
| 6 |
-
"add_adapter": false,
|
| 7 |
-
"apply_spec_augment": true,
|
| 8 |
-
"architectures": [
|
| 9 |
-
"Wav2Vec2ForCTC"
|
| 10 |
-
],
|
| 11 |
-
"attention_dropout": 0.0,
|
| 12 |
-
"bos_token_id": 1,
|
| 13 |
-
"classifier_proj_size": 256,
|
| 14 |
-
"codevector_dim": 768,
|
| 15 |
-
"contrastive_logits_temperature": 0.1,
|
| 16 |
-
"conv_bias": true,
|
| 17 |
-
"conv_dim": [
|
| 18 |
-
512,
|
| 19 |
-
512,
|
| 20 |
-
512,
|
| 21 |
-
512,
|
| 22 |
-
512,
|
| 23 |
-
512,
|
| 24 |
-
512
|
| 25 |
-
],
|
| 26 |
-
"conv_kernel": [
|
| 27 |
-
10,
|
| 28 |
-
3,
|
| 29 |
-
3,
|
| 30 |
-
3,
|
| 31 |
-
3,
|
| 32 |
-
2,
|
| 33 |
-
2
|
| 34 |
-
],
|
| 35 |
-
"conv_stride": [
|
| 36 |
-
5,
|
| 37 |
-
2,
|
| 38 |
-
2,
|
| 39 |
-
2,
|
| 40 |
-
2,
|
| 41 |
-
2,
|
| 42 |
-
2
|
| 43 |
-
],
|
| 44 |
-
"ctc_loss_reduction": "mean",
|
| 45 |
-
"ctc_zero_infinity": false,
|
| 46 |
-
"diversity_loss_weight": 0.1,
|
| 47 |
-
"do_stable_layer_norm": true,
|
| 48 |
-
"eos_token_id": 2,
|
| 49 |
-
"feat_extract_activation": "gelu",
|
| 50 |
-
"feat_extract_dropout": 0.0,
|
| 51 |
-
"feat_extract_norm": "layer",
|
| 52 |
-
"feat_proj_dropout": 0.0,
|
| 53 |
-
"feat_quantizer_dropout": 0.0,
|
| 54 |
-
"final_dropout": 0.0,
|
| 55 |
-
"hidden_act": "gelu",
|
| 56 |
-
"hidden_dropout": 0.0,
|
| 57 |
-
"hidden_size": 1024,
|
| 58 |
-
"initializer_range": 0.02,
|
| 59 |
-
"intermediate_size": 4096,
|
| 60 |
-
"layer_norm_eps": 1e-05,
|
| 61 |
-
"layerdrop": 0.0,
|
| 62 |
-
"mask_channel_length": 10,
|
| 63 |
-
"mask_channel_min_space": 1,
|
| 64 |
-
"mask_channel_other": 0.0,
|
| 65 |
-
"mask_channel_prob": 0.0,
|
| 66 |
-
"mask_channel_selection": "static",
|
| 67 |
-
"mask_feature_length": 64,
|
| 68 |
-
"mask_feature_min_masks": 0,
|
| 69 |
-
"mask_feature_prob": 0.25,
|
| 70 |
-
"mask_time_length": 10,
|
| 71 |
-
"mask_time_min_masks": 2,
|
| 72 |
-
"mask_time_min_space": 1,
|
| 73 |
-
"mask_time_other": 0.0,
|
| 74 |
-
"mask_time_prob": 0.75,
|
| 75 |
-
"mask_time_selection": "static",
|
| 76 |
-
"model_type": "wav2vec2",
|
| 77 |
-
"num_adapter_layers": 3,
|
| 78 |
-
"num_attention_heads": 16,
|
| 79 |
-
"num_codevector_groups": 2,
|
| 80 |
-
"num_codevectors_per_group": 320,
|
| 81 |
-
"num_conv_pos_embedding_groups": 16,
|
| 82 |
-
"num_conv_pos_embeddings": 128,
|
| 83 |
-
"num_feat_extract_layers": 7,
|
| 84 |
-
"num_hidden_layers": 24,
|
| 85 |
-
"num_negatives": 100,
|
| 86 |
-
"output_hidden_size": 1024,
|
| 87 |
-
"pad_token_id": 31,
|
| 88 |
-
"proj_codevector_dim": 768,
|
| 89 |
-
"tdnn_dilation": [
|
| 90 |
-
1,
|
| 91 |
-
2,
|
| 92 |
-
3,
|
| 93 |
-
1,
|
| 94 |
-
1
|
| 95 |
-
],
|
| 96 |
-
"tdnn_dim": [
|
| 97 |
-
512,
|
| 98 |
-
512,
|
| 99 |
-
512,
|
| 100 |
-
512,
|
| 101 |
-
1500
|
| 102 |
-
],
|
| 103 |
-
"tdnn_kernel": [
|
| 104 |
-
5,
|
| 105 |
-
3,
|
| 106 |
-
3,
|
| 107 |
-
1,
|
| 108 |
-
1
|
| 109 |
-
],
|
| 110 |
-
"torch_dtype": "float32",
|
| 111 |
-
"transformers_version": "4.17.0.dev0",
|
| 112 |
-
"use_weighted_layer_sum": false,
|
| 113 |
-
"vocab_size": 34,
|
| 114 |
-
"xvector_output_dim": 512
|
| 115 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/optimizer.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:6ddcef79ae9c9cbc49da385ba75dde3728f6718eeb4a1ea9f6af1e9bb92ea574
|
| 3 |
-
size 2490337809
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/preprocessor_config.json
DELETED
|
@@ -1,9 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"do_normalize": true,
|
| 3 |
-
"feature_extractor_type": "Wav2Vec2FeatureExtractor",
|
| 4 |
-
"feature_size": 1,
|
| 5 |
-
"padding_side": "right",
|
| 6 |
-
"padding_value": 0,
|
| 7 |
-
"return_attention_mask": true,
|
| 8 |
-
"sampling_rate": 16000
|
| 9 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/pytorch_model.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:3eae58c0bb9c908100e64bb4cfc44af074bab8941dffb754f675e676ebd7972b
|
| 3 |
-
size 1262063089
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/rng_state.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:7ae13169f4754f47cf02dcd08d7a9d3852557179a00533dd7f00f4265d8a8fba
|
| 3 |
-
size 14567
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/scaler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:a4c027f185a945dda35351dc282299cb9a8d301ee07589b42eb812ad50ff4977
|
| 3 |
-
size 559
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/scheduler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:c583f6b4d18dff47ed1b5d6c7fe67a5bf74f8906e7b32b87ae71e701db2a021c
|
| 3 |
-
size 623
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/trainer_state.json
DELETED
|
@@ -1,1771 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"best_metric": null,
|
| 3 |
-
"best_model_checkpoint": null,
|
| 4 |
-
"epoch": 49.44808743169399,
|
| 5 |
-
"global_step": 4500,
|
| 6 |
-
"is_hyper_param_search": false,
|
| 7 |
-
"is_local_process_zero": true,
|
| 8 |
-
"is_world_process_zero": true,
|
| 9 |
-
"log_history": [
|
| 10 |
-
{
|
| 11 |
-
"epoch": 0.22,
|
| 12 |
-
"learning_rate": 2.5e-06,
|
| 13 |
-
"loss": 3.5867,
|
| 14 |
-
"step": 20
|
| 15 |
-
},
|
| 16 |
-
{
|
| 17 |
-
"epoch": 0.44,
|
| 18 |
-
"learning_rate": 5e-06,
|
| 19 |
-
"loss": 3.5457,
|
| 20 |
-
"step": 40
|
| 21 |
-
},
|
| 22 |
-
{
|
| 23 |
-
"epoch": 0.66,
|
| 24 |
-
"learning_rate": 7.5e-06,
|
| 25 |
-
"loss": 3.4513,
|
| 26 |
-
"step": 60
|
| 27 |
-
},
|
| 28 |
-
{
|
| 29 |
-
"epoch": 0.87,
|
| 30 |
-
"learning_rate": 1e-05,
|
| 31 |
-
"loss": 3.3432,
|
| 32 |
-
"step": 80
|
| 33 |
-
},
|
| 34 |
-
{
|
| 35 |
-
"epoch": 1.1,
|
| 36 |
-
"learning_rate": 1.25e-05,
|
| 37 |
-
"loss": 3.3533,
|
| 38 |
-
"step": 100
|
| 39 |
-
},
|
| 40 |
-
{
|
| 41 |
-
"epoch": 1.1,
|
| 42 |
-
"eval_loss": 3.2806732654571533,
|
| 43 |
-
"eval_runtime": 190.4728,
|
| 44 |
-
"eval_samples_per_second": 25.426,
|
| 45 |
-
"eval_steps_per_second": 0.798,
|
| 46 |
-
"eval_wer": 1.0,
|
| 47 |
-
"step": 100
|
| 48 |
-
},
|
| 49 |
-
{
|
| 50 |
-
"epoch": 1.32,
|
| 51 |
-
"learning_rate": 1.5e-05,
|
| 52 |
-
"loss": 3.2217,
|
| 53 |
-
"step": 120
|
| 54 |
-
},
|
| 55 |
-
{
|
| 56 |
-
"epoch": 1.54,
|
| 57 |
-
"learning_rate": 1.7500000000000002e-05,
|
| 58 |
-
"loss": 3.1765,
|
| 59 |
-
"step": 140
|
| 60 |
-
},
|
| 61 |
-
{
|
| 62 |
-
"epoch": 1.75,
|
| 63 |
-
"learning_rate": 2e-05,
|
| 64 |
-
"loss": 3.1408,
|
| 65 |
-
"step": 160
|
| 66 |
-
},
|
| 67 |
-
{
|
| 68 |
-
"epoch": 1.97,
|
| 69 |
-
"learning_rate": 2.2499999999999998e-05,
|
| 70 |
-
"loss": 3.1165,
|
| 71 |
-
"step": 180
|
| 72 |
-
},
|
| 73 |
-
{
|
| 74 |
-
"epoch": 2.2,
|
| 75 |
-
"learning_rate": 2.5e-05,
|
| 76 |
-
"loss": 3.1709,
|
| 77 |
-
"step": 200
|
| 78 |
-
},
|
| 79 |
-
{
|
| 80 |
-
"epoch": 2.2,
|
| 81 |
-
"eval_loss": 3.1325438022613525,
|
| 82 |
-
"eval_runtime": 192.4978,
|
| 83 |
-
"eval_samples_per_second": 25.159,
|
| 84 |
-
"eval_steps_per_second": 0.79,
|
| 85 |
-
"eval_wer": 1.0,
|
| 86 |
-
"step": 200
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"epoch": 2.42,
|
| 90 |
-
"learning_rate": 2.75e-05,
|
| 91 |
-
"loss": 3.079,
|
| 92 |
-
"step": 220
|
| 93 |
-
},
|
| 94 |
-
{
|
| 95 |
-
"epoch": 2.63,
|
| 96 |
-
"learning_rate": 3e-05,
|
| 97 |
-
"loss": 3.0677,
|
| 98 |
-
"step": 240
|
| 99 |
-
},
|
| 100 |
-
{
|
| 101 |
-
"epoch": 2.85,
|
| 102 |
-
"learning_rate": 3.2500000000000004e-05,
|
| 103 |
-
"loss": 3.0656,
|
| 104 |
-
"step": 260
|
| 105 |
-
},
|
| 106 |
-
{
|
| 107 |
-
"epoch": 3.08,
|
| 108 |
-
"learning_rate": 3.5000000000000004e-05,
|
| 109 |
-
"loss": 3.1463,
|
| 110 |
-
"step": 280
|
| 111 |
-
},
|
| 112 |
-
{
|
| 113 |
-
"epoch": 3.3,
|
| 114 |
-
"learning_rate": 3.75e-05,
|
| 115 |
-
"loss": 3.0573,
|
| 116 |
-
"step": 300
|
| 117 |
-
},
|
| 118 |
-
{
|
| 119 |
-
"epoch": 3.3,
|
| 120 |
-
"eval_loss": 3.0614514350891113,
|
| 121 |
-
"eval_runtime": 194.36,
|
| 122 |
-
"eval_samples_per_second": 24.918,
|
| 123 |
-
"eval_steps_per_second": 0.782,
|
| 124 |
-
"eval_wer": 1.0,
|
| 125 |
-
"step": 300
|
| 126 |
-
},
|
| 127 |
-
{
|
| 128 |
-
"epoch": 3.51,
|
| 129 |
-
"learning_rate": 4e-05,
|
| 130 |
-
"loss": 3.0511,
|
| 131 |
-
"step": 320
|
| 132 |
-
},
|
| 133 |
-
{
|
| 134 |
-
"epoch": 3.73,
|
| 135 |
-
"learning_rate": 4.25e-05,
|
| 136 |
-
"loss": 3.0358,
|
| 137 |
-
"step": 340
|
| 138 |
-
},
|
| 139 |
-
{
|
| 140 |
-
"epoch": 3.95,
|
| 141 |
-
"learning_rate": 4.4999999999999996e-05,
|
| 142 |
-
"loss": 3.0416,
|
| 143 |
-
"step": 360
|
| 144 |
-
},
|
| 145 |
-
{
|
| 146 |
-
"epoch": 4.17,
|
| 147 |
-
"learning_rate": 4.75e-05,
|
| 148 |
-
"loss": 3.1053,
|
| 149 |
-
"step": 380
|
| 150 |
-
},
|
| 151 |
-
{
|
| 152 |
-
"epoch": 4.39,
|
| 153 |
-
"learning_rate": 5e-05,
|
| 154 |
-
"loss": 3.0314,
|
| 155 |
-
"step": 400
|
| 156 |
-
},
|
| 157 |
-
{
|
| 158 |
-
"epoch": 4.39,
|
| 159 |
-
"eval_loss": 3.0990231037139893,
|
| 160 |
-
"eval_runtime": 198.3688,
|
| 161 |
-
"eval_samples_per_second": 24.414,
|
| 162 |
-
"eval_steps_per_second": 0.766,
|
| 163 |
-
"eval_wer": 1.0,
|
| 164 |
-
"step": 400
|
| 165 |
-
},
|
| 166 |
-
{
|
| 167 |
-
"epoch": 4.61,
|
| 168 |
-
"learning_rate": 5.25e-05,
|
| 169 |
-
"loss": 3.0309,
|
| 170 |
-
"step": 420
|
| 171 |
-
},
|
| 172 |
-
{
|
| 173 |
-
"epoch": 4.83,
|
| 174 |
-
"learning_rate": 5.5e-05,
|
| 175 |
-
"loss": 3.0259,
|
| 176 |
-
"step": 440
|
| 177 |
-
},
|
| 178 |
-
{
|
| 179 |
-
"epoch": 5.05,
|
| 180 |
-
"learning_rate": 5.75e-05,
|
| 181 |
-
"loss": 3.0998,
|
| 182 |
-
"step": 460
|
| 183 |
-
},
|
| 184 |
-
{
|
| 185 |
-
"epoch": 5.27,
|
| 186 |
-
"learning_rate": 6e-05,
|
| 187 |
-
"loss": 3.0152,
|
| 188 |
-
"step": 480
|
| 189 |
-
},
|
| 190 |
-
{
|
| 191 |
-
"epoch": 5.49,
|
| 192 |
-
"learning_rate": 6.25e-05,
|
| 193 |
-
"loss": 3.0129,
|
| 194 |
-
"step": 500
|
| 195 |
-
},
|
| 196 |
-
{
|
| 197 |
-
"epoch": 5.49,
|
| 198 |
-
"eval_loss": 3.039973497390747,
|
| 199 |
-
"eval_runtime": 190.8567,
|
| 200 |
-
"eval_samples_per_second": 25.375,
|
| 201 |
-
"eval_steps_per_second": 0.796,
|
| 202 |
-
"eval_wer": 1.0,
|
| 203 |
-
"step": 500
|
| 204 |
-
},
|
| 205 |
-
{
|
| 206 |
-
"epoch": 5.71,
|
| 207 |
-
"learning_rate": 6.500000000000001e-05,
|
| 208 |
-
"loss": 3.0088,
|
| 209 |
-
"step": 520
|
| 210 |
-
},
|
| 211 |
-
{
|
| 212 |
-
"epoch": 5.93,
|
| 213 |
-
"learning_rate": 6.75e-05,
|
| 214 |
-
"loss": 3.0051,
|
| 215 |
-
"step": 540
|
| 216 |
-
},
|
| 217 |
-
{
|
| 218 |
-
"epoch": 6.15,
|
| 219 |
-
"learning_rate": 7.000000000000001e-05,
|
| 220 |
-
"loss": 3.073,
|
| 221 |
-
"step": 560
|
| 222 |
-
},
|
| 223 |
-
{
|
| 224 |
-
"epoch": 6.37,
|
| 225 |
-
"learning_rate": 7.25e-05,
|
| 226 |
-
"loss": 3.0031,
|
| 227 |
-
"step": 580
|
| 228 |
-
},
|
| 229 |
-
{
|
| 230 |
-
"epoch": 6.59,
|
| 231 |
-
"learning_rate": 7.5e-05,
|
| 232 |
-
"loss": 2.9964,
|
| 233 |
-
"step": 600
|
| 234 |
-
},
|
| 235 |
-
{
|
| 236 |
-
"epoch": 6.59,
|
| 237 |
-
"eval_loss": 2.998962640762329,
|
| 238 |
-
"eval_runtime": 193.5213,
|
| 239 |
-
"eval_samples_per_second": 25.026,
|
| 240 |
-
"eval_steps_per_second": 0.785,
|
| 241 |
-
"eval_wer": 1.0,
|
| 242 |
-
"step": 600
|
| 243 |
-
},
|
| 244 |
-
{
|
| 245 |
-
"epoch": 6.81,
|
| 246 |
-
"learning_rate": 7.75e-05,
|
| 247 |
-
"loss": 2.9921,
|
| 248 |
-
"step": 620
|
| 249 |
-
},
|
| 250 |
-
{
|
| 251 |
-
"epoch": 7.03,
|
| 252 |
-
"learning_rate": 8e-05,
|
| 253 |
-
"loss": 3.0665,
|
| 254 |
-
"step": 640
|
| 255 |
-
},
|
| 256 |
-
{
|
| 257 |
-
"epoch": 7.25,
|
| 258 |
-
"learning_rate": 8.25e-05,
|
| 259 |
-
"loss": 2.9826,
|
| 260 |
-
"step": 660
|
| 261 |
-
},
|
| 262 |
-
{
|
| 263 |
-
"epoch": 7.47,
|
| 264 |
-
"learning_rate": 8.5e-05,
|
| 265 |
-
"loss": 2.9689,
|
| 266 |
-
"step": 680
|
| 267 |
-
},
|
| 268 |
-
{
|
| 269 |
-
"epoch": 7.69,
|
| 270 |
-
"learning_rate": 8.75e-05,
|
| 271 |
-
"loss": 2.9602,
|
| 272 |
-
"step": 700
|
| 273 |
-
},
|
| 274 |
-
{
|
| 275 |
-
"epoch": 7.69,
|
| 276 |
-
"eval_loss": 2.9620397090911865,
|
| 277 |
-
"eval_runtime": 193.5851,
|
| 278 |
-
"eval_samples_per_second": 25.017,
|
| 279 |
-
"eval_steps_per_second": 0.785,
|
| 280 |
-
"eval_wer": 1.0,
|
| 281 |
-
"step": 700
|
| 282 |
-
},
|
| 283 |
-
{
|
| 284 |
-
"epoch": 7.91,
|
| 285 |
-
"learning_rate": 8.999999999999999e-05,
|
| 286 |
-
"loss": 2.9639,
|
| 287 |
-
"step": 720
|
| 288 |
-
},
|
| 289 |
-
{
|
| 290 |
-
"epoch": 8.13,
|
| 291 |
-
"learning_rate": 9.25e-05,
|
| 292 |
-
"loss": 3.0215,
|
| 293 |
-
"step": 740
|
| 294 |
-
},
|
| 295 |
-
{
|
| 296 |
-
"epoch": 8.35,
|
| 297 |
-
"learning_rate": 9.5e-05,
|
| 298 |
-
"loss": 2.9454,
|
| 299 |
-
"step": 760
|
| 300 |
-
},
|
| 301 |
-
{
|
| 302 |
-
"epoch": 8.57,
|
| 303 |
-
"learning_rate": 9.750000000000001e-05,
|
| 304 |
-
"loss": 2.9239,
|
| 305 |
-
"step": 780
|
| 306 |
-
},
|
| 307 |
-
{
|
| 308 |
-
"epoch": 8.79,
|
| 309 |
-
"learning_rate": 0.0001,
|
| 310 |
-
"loss": 2.8756,
|
| 311 |
-
"step": 800
|
| 312 |
-
},
|
| 313 |
-
{
|
| 314 |
-
"epoch": 8.79,
|
| 315 |
-
"eval_loss": 2.7302000522613525,
|
| 316 |
-
"eval_runtime": 191.8065,
|
| 317 |
-
"eval_samples_per_second": 25.249,
|
| 318 |
-
"eval_steps_per_second": 0.792,
|
| 319 |
-
"eval_wer": 1.0,
|
| 320 |
-
"step": 800
|
| 321 |
-
},
|
| 322 |
-
{
|
| 323 |
-
"epoch": 9.01,
|
| 324 |
-
"learning_rate": 0.0001025,
|
| 325 |
-
"loss": 2.8933,
|
| 326 |
-
"step": 820
|
| 327 |
-
},
|
| 328 |
-
{
|
| 329 |
-
"epoch": 9.23,
|
| 330 |
-
"learning_rate": 0.000105,
|
| 331 |
-
"loss": 2.7318,
|
| 332 |
-
"step": 840
|
| 333 |
-
},
|
| 334 |
-
{
|
| 335 |
-
"epoch": 9.45,
|
| 336 |
-
"learning_rate": 0.0001075,
|
| 337 |
-
"loss": 2.5941,
|
| 338 |
-
"step": 860
|
| 339 |
-
},
|
| 340 |
-
{
|
| 341 |
-
"epoch": 9.67,
|
| 342 |
-
"learning_rate": 0.00011,
|
| 343 |
-
"loss": 2.4441,
|
| 344 |
-
"step": 880
|
| 345 |
-
},
|
| 346 |
-
{
|
| 347 |
-
"epoch": 9.89,
|
| 348 |
-
"learning_rate": 0.00011250000000000001,
|
| 349 |
-
"loss": 2.2931,
|
| 350 |
-
"step": 900
|
| 351 |
-
},
|
| 352 |
-
{
|
| 353 |
-
"epoch": 9.89,
|
| 354 |
-
"eval_loss": 1.5057899951934814,
|
| 355 |
-
"eval_runtime": 196.4368,
|
| 356 |
-
"eval_samples_per_second": 24.654,
|
| 357 |
-
"eval_steps_per_second": 0.774,
|
| 358 |
-
"eval_wer": 0.9775759296054499,
|
| 359 |
-
"step": 900
|
| 360 |
-
},
|
| 361 |
-
{
|
| 362 |
-
"epoch": 10.11,
|
| 363 |
-
"learning_rate": 0.000115,
|
| 364 |
-
"loss": 2.1999,
|
| 365 |
-
"step": 920
|
| 366 |
-
},
|
| 367 |
-
{
|
| 368 |
-
"epoch": 10.33,
|
| 369 |
-
"learning_rate": 0.0001175,
|
| 370 |
-
"loss": 2.0574,
|
| 371 |
-
"step": 940
|
| 372 |
-
},
|
| 373 |
-
{
|
| 374 |
-
"epoch": 10.55,
|
| 375 |
-
"learning_rate": 0.00012,
|
| 376 |
-
"loss": 1.9672,
|
| 377 |
-
"step": 960
|
| 378 |
-
},
|
| 379 |
-
{
|
| 380 |
-
"epoch": 10.77,
|
| 381 |
-
"learning_rate": 0.0001225,
|
| 382 |
-
"loss": 1.9015,
|
| 383 |
-
"step": 980
|
| 384 |
-
},
|
| 385 |
-
{
|
| 386 |
-
"epoch": 10.98,
|
| 387 |
-
"learning_rate": 0.000125,
|
| 388 |
-
"loss": 1.8427,
|
| 389 |
-
"step": 1000
|
| 390 |
-
},
|
| 391 |
-
{
|
| 392 |
-
"epoch": 10.98,
|
| 393 |
-
"eval_loss": 0.9154536724090576,
|
| 394 |
-
"eval_runtime": 203.1168,
|
| 395 |
-
"eval_samples_per_second": 23.843,
|
| 396 |
-
"eval_steps_per_second": 0.748,
|
| 397 |
-
"eval_wer": 0.7832245245529378,
|
| 398 |
-
"step": 1000
|
| 399 |
-
},
|
| 400 |
-
{
|
| 401 |
-
"epoch": 11.21,
|
| 402 |
-
"learning_rate": 0.0001275,
|
| 403 |
-
"loss": 1.8155,
|
| 404 |
-
"step": 1020
|
| 405 |
-
},
|
| 406 |
-
{
|
| 407 |
-
"epoch": 11.43,
|
| 408 |
-
"learning_rate": 0.00013000000000000002,
|
| 409 |
-
"loss": 1.6767,
|
| 410 |
-
"step": 1040
|
| 411 |
-
},
|
| 412 |
-
{
|
| 413 |
-
"epoch": 11.64,
|
| 414 |
-
"learning_rate": 0.00013250000000000002,
|
| 415 |
-
"loss": 1.5184,
|
| 416 |
-
"step": 1060
|
| 417 |
-
},
|
| 418 |
-
{
|
| 419 |
-
"epoch": 11.86,
|
| 420 |
-
"learning_rate": 0.000135,
|
| 421 |
-
"loss": 1.4505,
|
| 422 |
-
"step": 1080
|
| 423 |
-
},
|
| 424 |
-
{
|
| 425 |
-
"epoch": 12.09,
|
| 426 |
-
"learning_rate": 0.0001375,
|
| 427 |
-
"loss": 1.4286,
|
| 428 |
-
"step": 1100
|
| 429 |
-
},
|
| 430 |
-
{
|
| 431 |
-
"epoch": 12.09,
|
| 432 |
-
"eval_loss": 0.4074769914150238,
|
| 433 |
-
"eval_runtime": 194.7798,
|
| 434 |
-
"eval_samples_per_second": 24.864,
|
| 435 |
-
"eval_steps_per_second": 0.78,
|
| 436 |
-
"eval_wer": 0.37964802724950325,
|
| 437 |
-
"step": 1100
|
| 438 |
-
},
|
| 439 |
-
{
|
| 440 |
-
"epoch": 12.31,
|
| 441 |
-
"learning_rate": 0.00014000000000000001,
|
| 442 |
-
"loss": 1.3478,
|
| 443 |
-
"step": 1120
|
| 444 |
-
},
|
| 445 |
-
{
|
| 446 |
-
"epoch": 12.52,
|
| 447 |
-
"learning_rate": 0.0001425,
|
| 448 |
-
"loss": 1.3164,
|
| 449 |
-
"step": 1140
|
| 450 |
-
},
|
| 451 |
-
{
|
| 452 |
-
"epoch": 12.74,
|
| 453 |
-
"learning_rate": 0.000145,
|
| 454 |
-
"loss": 1.269,
|
| 455 |
-
"step": 1160
|
| 456 |
-
},
|
| 457 |
-
{
|
| 458 |
-
"epoch": 12.96,
|
| 459 |
-
"learning_rate": 0.0001475,
|
| 460 |
-
"loss": 1.2292,
|
| 461 |
-
"step": 1180
|
| 462 |
-
},
|
| 463 |
-
{
|
| 464 |
-
"epoch": 13.19,
|
| 465 |
-
"learning_rate": 0.00015,
|
| 466 |
-
"loss": 1.2229,
|
| 467 |
-
"step": 1200
|
| 468 |
-
},
|
| 469 |
-
{
|
| 470 |
-
"epoch": 13.19,
|
| 471 |
-
"eval_loss": 0.28933778405189514,
|
| 472 |
-
"eval_runtime": 192.6398,
|
| 473 |
-
"eval_samples_per_second": 25.14,
|
| 474 |
-
"eval_steps_per_second": 0.789,
|
| 475 |
-
"eval_wer": 0.26520011353959694,
|
| 476 |
-
"step": 1200
|
| 477 |
-
},
|
| 478 |
-
{
|
| 479 |
-
"epoch": 13.4,
|
| 480 |
-
"learning_rate": 0.0001525,
|
| 481 |
-
"loss": 1.182,
|
| 482 |
-
"step": 1220
|
| 483 |
-
},
|
| 484 |
-
{
|
| 485 |
-
"epoch": 13.62,
|
| 486 |
-
"learning_rate": 0.000155,
|
| 487 |
-
"loss": 1.1768,
|
| 488 |
-
"step": 1240
|
| 489 |
-
},
|
| 490 |
-
{
|
| 491 |
-
"epoch": 13.84,
|
| 492 |
-
"learning_rate": 0.0001575,
|
| 493 |
-
"loss": 1.1534,
|
| 494 |
-
"step": 1260
|
| 495 |
-
},
|
| 496 |
-
{
|
| 497 |
-
"epoch": 14.07,
|
| 498 |
-
"learning_rate": 0.00016,
|
| 499 |
-
"loss": 1.1832,
|
| 500 |
-
"step": 1280
|
| 501 |
-
},
|
| 502 |
-
{
|
| 503 |
-
"epoch": 14.28,
|
| 504 |
-
"learning_rate": 0.00016250000000000002,
|
| 505 |
-
"loss": 1.1106,
|
| 506 |
-
"step": 1300
|
| 507 |
-
},
|
| 508 |
-
{
|
| 509 |
-
"epoch": 14.28,
|
| 510 |
-
"eval_loss": 0.24685777723789215,
|
| 511 |
-
"eval_runtime": 198.5067,
|
| 512 |
-
"eval_samples_per_second": 24.397,
|
| 513 |
-
"eval_steps_per_second": 0.766,
|
| 514 |
-
"eval_wer": 0.2253760999148453,
|
| 515 |
-
"step": 1300
|
| 516 |
-
},
|
| 517 |
-
{
|
| 518 |
-
"epoch": 14.5,
|
| 519 |
-
"learning_rate": 0.000165,
|
| 520 |
-
"loss": 1.1217,
|
| 521 |
-
"step": 1320
|
| 522 |
-
},
|
| 523 |
-
{
|
| 524 |
-
"epoch": 14.72,
|
| 525 |
-
"learning_rate": 0.0001675,
|
| 526 |
-
"loss": 1.1236,
|
| 527 |
-
"step": 1340
|
| 528 |
-
},
|
| 529 |
-
{
|
| 530 |
-
"epoch": 14.94,
|
| 531 |
-
"learning_rate": 0.00017,
|
| 532 |
-
"loss": 1.0969,
|
| 533 |
-
"step": 1360
|
| 534 |
-
},
|
| 535 |
-
{
|
| 536 |
-
"epoch": 15.16,
|
| 537 |
-
"learning_rate": 0.0001725,
|
| 538 |
-
"loss": 1.0971,
|
| 539 |
-
"step": 1380
|
| 540 |
-
},
|
| 541 |
-
{
|
| 542 |
-
"epoch": 15.38,
|
| 543 |
-
"learning_rate": 0.000175,
|
| 544 |
-
"loss": 1.0663,
|
| 545 |
-
"step": 1400
|
| 546 |
-
},
|
| 547 |
-
{
|
| 548 |
-
"epoch": 15.38,
|
| 549 |
-
"eval_loss": 0.22192780673503876,
|
| 550 |
-
"eval_runtime": 189.4566,
|
| 551 |
-
"eval_samples_per_second": 25.563,
|
| 552 |
-
"eval_steps_per_second": 0.802,
|
| 553 |
-
"eval_wer": 0.19727504967357365,
|
| 554 |
-
"step": 1400
|
| 555 |
-
},
|
| 556 |
-
{
|
| 557 |
-
"epoch": 15.6,
|
| 558 |
-
"learning_rate": 0.0001775,
|
| 559 |
-
"loss": 1.0733,
|
| 560 |
-
"step": 1420
|
| 561 |
-
},
|
| 562 |
-
{
|
| 563 |
-
"epoch": 15.82,
|
| 564 |
-
"learning_rate": 0.00017999999999999998,
|
| 565 |
-
"loss": 1.0635,
|
| 566 |
-
"step": 1440
|
| 567 |
-
},
|
| 568 |
-
{
|
| 569 |
-
"epoch": 16.04,
|
| 570 |
-
"learning_rate": 0.0001825,
|
| 571 |
-
"loss": 1.1079,
|
| 572 |
-
"step": 1460
|
| 573 |
-
},
|
| 574 |
-
{
|
| 575 |
-
"epoch": 16.26,
|
| 576 |
-
"learning_rate": 0.000185,
|
| 577 |
-
"loss": 1.0467,
|
| 578 |
-
"step": 1480
|
| 579 |
-
},
|
| 580 |
-
{
|
| 581 |
-
"epoch": 16.48,
|
| 582 |
-
"learning_rate": 0.0001875,
|
| 583 |
-
"loss": 1.0667,
|
| 584 |
-
"step": 1500
|
| 585 |
-
},
|
| 586 |
-
{
|
| 587 |
-
"epoch": 16.48,
|
| 588 |
-
"eval_loss": 0.21288961172103882,
|
| 589 |
-
"eval_runtime": 189.0078,
|
| 590 |
-
"eval_samples_per_second": 25.623,
|
| 591 |
-
"eval_steps_per_second": 0.804,
|
| 592 |
-
"eval_wer": 0.18944081748509792,
|
| 593 |
-
"step": 1500
|
| 594 |
-
},
|
| 595 |
-
{
|
| 596 |
-
"epoch": 16.7,
|
| 597 |
-
"learning_rate": 0.00019,
|
| 598 |
-
"loss": 1.0204,
|
| 599 |
-
"step": 1520
|
| 600 |
-
},
|
| 601 |
-
{
|
| 602 |
-
"epoch": 16.92,
|
| 603 |
-
"learning_rate": 0.00019250000000000002,
|
| 604 |
-
"loss": 1.0356,
|
| 605 |
-
"step": 1540
|
| 606 |
-
},
|
| 607 |
-
{
|
| 608 |
-
"epoch": 17.14,
|
| 609 |
-
"learning_rate": 0.00019500000000000002,
|
| 610 |
-
"loss": 1.0538,
|
| 611 |
-
"step": 1560
|
| 612 |
-
},
|
| 613 |
-
{
|
| 614 |
-
"epoch": 17.36,
|
| 615 |
-
"learning_rate": 0.0001975,
|
| 616 |
-
"loss": 1.0228,
|
| 617 |
-
"step": 1580
|
| 618 |
-
},
|
| 619 |
-
{
|
| 620 |
-
"epoch": 17.58,
|
| 621 |
-
"learning_rate": 0.0002,
|
| 622 |
-
"loss": 1.0193,
|
| 623 |
-
"step": 1600
|
| 624 |
-
},
|
| 625 |
-
{
|
| 626 |
-
"epoch": 17.58,
|
| 627 |
-
"eval_loss": 0.1991206556558609,
|
| 628 |
-
"eval_runtime": 191.1717,
|
| 629 |
-
"eval_samples_per_second": 25.333,
|
| 630 |
-
"eval_steps_per_second": 0.795,
|
| 631 |
-
"eval_wer": 0.17885325007096226,
|
| 632 |
-
"step": 1600
|
| 633 |
-
},
|
| 634 |
-
{
|
| 635 |
-
"epoch": 17.8,
|
| 636 |
-
"learning_rate": 0.00020250000000000002,
|
| 637 |
-
"loss": 1.0322,
|
| 638 |
-
"step": 1620
|
| 639 |
-
},
|
| 640 |
-
{
|
| 641 |
-
"epoch": 18.02,
|
| 642 |
-
"learning_rate": 0.000205,
|
| 643 |
-
"loss": 1.0176,
|
| 644 |
-
"step": 1640
|
| 645 |
-
},
|
| 646 |
-
{
|
| 647 |
-
"epoch": 18.24,
|
| 648 |
-
"learning_rate": 0.0002075,
|
| 649 |
-
"loss": 1.0272,
|
| 650 |
-
"step": 1660
|
| 651 |
-
},
|
| 652 |
-
{
|
| 653 |
-
"epoch": 18.46,
|
| 654 |
-
"learning_rate": 0.00021,
|
| 655 |
-
"loss": 0.9675,
|
| 656 |
-
"step": 1680
|
| 657 |
-
},
|
| 658 |
-
{
|
| 659 |
-
"epoch": 18.68,
|
| 660 |
-
"learning_rate": 0.0002125,
|
| 661 |
-
"loss": 0.9816,
|
| 662 |
-
"step": 1700
|
| 663 |
-
},
|
| 664 |
-
{
|
| 665 |
-
"epoch": 18.68,
|
| 666 |
-
"eval_loss": 0.19402356445789337,
|
| 667 |
-
"eval_runtime": 194.75,
|
| 668 |
-
"eval_samples_per_second": 24.868,
|
| 669 |
-
"eval_steps_per_second": 0.78,
|
| 670 |
-
"eval_wer": 0.18010218563724098,
|
| 671 |
-
"step": 1700
|
| 672 |
-
},
|
| 673 |
-
{
|
| 674 |
-
"epoch": 18.9,
|
| 675 |
-
"learning_rate": 0.000215,
|
| 676 |
-
"loss": 1.0023,
|
| 677 |
-
"step": 1720
|
| 678 |
-
},
|
| 679 |
-
{
|
| 680 |
-
"epoch": 19.12,
|
| 681 |
-
"learning_rate": 0.0002175,
|
| 682 |
-
"loss": 1.013,
|
| 683 |
-
"step": 1740
|
| 684 |
-
},
|
| 685 |
-
{
|
| 686 |
-
"epoch": 19.34,
|
| 687 |
-
"learning_rate": 0.00022,
|
| 688 |
-
"loss": 0.9664,
|
| 689 |
-
"step": 1760
|
| 690 |
-
},
|
| 691 |
-
{
|
| 692 |
-
"epoch": 19.56,
|
| 693 |
-
"learning_rate": 0.00022250000000000001,
|
| 694 |
-
"loss": 0.9736,
|
| 695 |
-
"step": 1780
|
| 696 |
-
},
|
| 697 |
-
{
|
| 698 |
-
"epoch": 19.78,
|
| 699 |
-
"learning_rate": 0.00022500000000000002,
|
| 700 |
-
"loss": 0.9814,
|
| 701 |
-
"step": 1800
|
| 702 |
-
},
|
| 703 |
-
{
|
| 704 |
-
"epoch": 19.78,
|
| 705 |
-
"eval_loss": 0.18596723675727844,
|
| 706 |
-
"eval_runtime": 188.4639,
|
| 707 |
-
"eval_samples_per_second": 25.697,
|
| 708 |
-
"eval_steps_per_second": 0.807,
|
| 709 |
-
"eval_wer": 0.16667612829974454,
|
| 710 |
-
"step": 1800
|
| 711 |
-
},
|
| 712 |
-
{
|
| 713 |
-
"epoch": 19.99,
|
| 714 |
-
"learning_rate": 0.0002275,
|
| 715 |
-
"loss": 1.0064,
|
| 716 |
-
"step": 1820
|
| 717 |
-
},
|
| 718 |
-
{
|
| 719 |
-
"epoch": 20.22,
|
| 720 |
-
"learning_rate": 0.00023,
|
| 721 |
-
"loss": 0.9583,
|
| 722 |
-
"step": 1840
|
| 723 |
-
},
|
| 724 |
-
{
|
| 725 |
-
"epoch": 20.44,
|
| 726 |
-
"learning_rate": 0.0002325,
|
| 727 |
-
"loss": 0.9646,
|
| 728 |
-
"step": 1860
|
| 729 |
-
},
|
| 730 |
-
{
|
| 731 |
-
"epoch": 20.66,
|
| 732 |
-
"learning_rate": 0.000235,
|
| 733 |
-
"loss": 0.9762,
|
| 734 |
-
"step": 1880
|
| 735 |
-
},
|
| 736 |
-
{
|
| 737 |
-
"epoch": 20.87,
|
| 738 |
-
"learning_rate": 0.0002375,
|
| 739 |
-
"loss": 0.9787,
|
| 740 |
-
"step": 1900
|
| 741 |
-
},
|
| 742 |
-
{
|
| 743 |
-
"epoch": 20.87,
|
| 744 |
-
"eval_loss": 0.18878202140331268,
|
| 745 |
-
"eval_runtime": 188.244,
|
| 746 |
-
"eval_samples_per_second": 25.727,
|
| 747 |
-
"eval_steps_per_second": 0.807,
|
| 748 |
-
"eval_wer": 0.16420664206642066,
|
| 749 |
-
"step": 1900
|
| 750 |
-
},
|
| 751 |
-
{
|
| 752 |
-
"epoch": 21.1,
|
| 753 |
-
"learning_rate": 0.00024,
|
| 754 |
-
"loss": 1.0218,
|
| 755 |
-
"step": 1920
|
| 756 |
-
},
|
| 757 |
-
{
|
| 758 |
-
"epoch": 21.32,
|
| 759 |
-
"learning_rate": 0.00024249999999999999,
|
| 760 |
-
"loss": 0.9505,
|
| 761 |
-
"step": 1940
|
| 762 |
-
},
|
| 763 |
-
{
|
| 764 |
-
"epoch": 21.54,
|
| 765 |
-
"learning_rate": 0.000245,
|
| 766 |
-
"loss": 0.9554,
|
| 767 |
-
"step": 1960
|
| 768 |
-
},
|
| 769 |
-
{
|
| 770 |
-
"epoch": 21.75,
|
| 771 |
-
"learning_rate": 0.0002475,
|
| 772 |
-
"loss": 0.9728,
|
| 773 |
-
"step": 1980
|
| 774 |
-
},
|
| 775 |
-
{
|
| 776 |
-
"epoch": 21.97,
|
| 777 |
-
"learning_rate": 0.00025,
|
| 778 |
-
"loss": 0.9699,
|
| 779 |
-
"step": 2000
|
| 780 |
-
},
|
| 781 |
-
{
|
| 782 |
-
"epoch": 21.97,
|
| 783 |
-
"eval_loss": 0.18748582899570465,
|
| 784 |
-
"eval_runtime": 190.7875,
|
| 785 |
-
"eval_samples_per_second": 25.384,
|
| 786 |
-
"eval_steps_per_second": 0.797,
|
| 787 |
-
"eval_wer": 0.17042293499858074,
|
| 788 |
-
"step": 2000
|
| 789 |
-
},
|
| 790 |
-
{
|
| 791 |
-
"epoch": 22.2,
|
| 792 |
-
"learning_rate": 0.00024803921568627453,
|
| 793 |
-
"loss": 0.9624,
|
| 794 |
-
"step": 2020
|
| 795 |
-
},
|
| 796 |
-
{
|
| 797 |
-
"epoch": 22.42,
|
| 798 |
-
"learning_rate": 0.000246078431372549,
|
| 799 |
-
"loss": 0.9419,
|
| 800 |
-
"step": 2040
|
| 801 |
-
},
|
| 802 |
-
{
|
| 803 |
-
"epoch": 22.63,
|
| 804 |
-
"learning_rate": 0.00024411764705882354,
|
| 805 |
-
"loss": 0.9563,
|
| 806 |
-
"step": 2060
|
| 807 |
-
},
|
| 808 |
-
{
|
| 809 |
-
"epoch": 22.85,
|
| 810 |
-
"learning_rate": 0.00024215686274509804,
|
| 811 |
-
"loss": 0.9643,
|
| 812 |
-
"step": 2080
|
| 813 |
-
},
|
| 814 |
-
{
|
| 815 |
-
"epoch": 23.08,
|
| 816 |
-
"learning_rate": 0.00024019607843137256,
|
| 817 |
-
"loss": 0.9616,
|
| 818 |
-
"step": 2100
|
| 819 |
-
},
|
| 820 |
-
{
|
| 821 |
-
"epoch": 23.08,
|
| 822 |
-
"eval_loss": 0.18017500638961792,
|
| 823 |
-
"eval_runtime": 191.3932,
|
| 824 |
-
"eval_samples_per_second": 25.304,
|
| 825 |
-
"eval_steps_per_second": 0.794,
|
| 826 |
-
"eval_wer": 0.16173715583309678,
|
| 827 |
-
"step": 2100
|
| 828 |
-
},
|
| 829 |
-
{
|
| 830 |
-
"epoch": 23.3,
|
| 831 |
-
"learning_rate": 0.00023823529411764704,
|
| 832 |
-
"loss": 0.917,
|
| 833 |
-
"step": 2120
|
| 834 |
-
},
|
| 835 |
-
{
|
| 836 |
-
"epoch": 23.51,
|
| 837 |
-
"learning_rate": 0.00023627450980392157,
|
| 838 |
-
"loss": 0.945,
|
| 839 |
-
"step": 2140
|
| 840 |
-
},
|
| 841 |
-
{
|
| 842 |
-
"epoch": 23.73,
|
| 843 |
-
"learning_rate": 0.0002343137254901961,
|
| 844 |
-
"loss": 0.9243,
|
| 845 |
-
"step": 2160
|
| 846 |
-
},
|
| 847 |
-
{
|
| 848 |
-
"epoch": 23.95,
|
| 849 |
-
"learning_rate": 0.0002323529411764706,
|
| 850 |
-
"loss": 0.9288,
|
| 851 |
-
"step": 2180
|
| 852 |
-
},
|
| 853 |
-
{
|
| 854 |
-
"epoch": 24.17,
|
| 855 |
-
"learning_rate": 0.0002303921568627451,
|
| 856 |
-
"loss": 0.9378,
|
| 857 |
-
"step": 2200
|
| 858 |
-
},
|
| 859 |
-
{
|
| 860 |
-
"epoch": 24.17,
|
| 861 |
-
"eval_loss": 0.17928896844387054,
|
| 862 |
-
"eval_runtime": 189.7619,
|
| 863 |
-
"eval_samples_per_second": 25.521,
|
| 864 |
-
"eval_steps_per_second": 0.801,
|
| 865 |
-
"eval_wer": 0.1577348850411581,
|
| 866 |
-
"step": 2200
|
| 867 |
-
},
|
| 868 |
-
{
|
| 869 |
-
"epoch": 24.39,
|
| 870 |
-
"learning_rate": 0.0002284313725490196,
|
| 871 |
-
"loss": 0.9071,
|
| 872 |
-
"step": 2220
|
| 873 |
-
},
|
| 874 |
-
{
|
| 875 |
-
"epoch": 24.61,
|
| 876 |
-
"learning_rate": 0.00022647058823529412,
|
| 877 |
-
"loss": 0.9054,
|
| 878 |
-
"step": 2240
|
| 879 |
-
},
|
| 880 |
-
{
|
| 881 |
-
"epoch": 24.83,
|
| 882 |
-
"learning_rate": 0.0002246078431372549,
|
| 883 |
-
"loss": 0.9303,
|
| 884 |
-
"step": 2260
|
| 885 |
-
},
|
| 886 |
-
{
|
| 887 |
-
"epoch": 25.05,
|
| 888 |
-
"learning_rate": 0.00022264705882352943,
|
| 889 |
-
"loss": 0.9376,
|
| 890 |
-
"step": 2280
|
| 891 |
-
},
|
| 892 |
-
{
|
| 893 |
-
"epoch": 25.27,
|
| 894 |
-
"learning_rate": 0.0002206862745098039,
|
| 895 |
-
"loss": 0.888,
|
| 896 |
-
"step": 2300
|
| 897 |
-
},
|
| 898 |
-
{
|
| 899 |
-
"epoch": 25.27,
|
| 900 |
-
"eval_loss": 0.17642559111118317,
|
| 901 |
-
"eval_runtime": 187.0437,
|
| 902 |
-
"eval_samples_per_second": 25.892,
|
| 903 |
-
"eval_steps_per_second": 0.813,
|
| 904 |
-
"eval_wer": 0.15452739142776042,
|
| 905 |
-
"step": 2300
|
| 906 |
-
},
|
| 907 |
-
{
|
| 908 |
-
"epoch": 25.49,
|
| 909 |
-
"learning_rate": 0.00021872549019607843,
|
| 910 |
-
"loss": 0.9135,
|
| 911 |
-
"step": 2320
|
| 912 |
-
},
|
| 913 |
-
{
|
| 914 |
-
"epoch": 25.71,
|
| 915 |
-
"learning_rate": 0.00021676470588235294,
|
| 916 |
-
"loss": 0.9094,
|
| 917 |
-
"step": 2340
|
| 918 |
-
},
|
| 919 |
-
{
|
| 920 |
-
"epoch": 25.93,
|
| 921 |
-
"learning_rate": 0.00021480392156862746,
|
| 922 |
-
"loss": 0.8879,
|
| 923 |
-
"step": 2360
|
| 924 |
-
},
|
| 925 |
-
{
|
| 926 |
-
"epoch": 26.15,
|
| 927 |
-
"learning_rate": 0.00021284313725490196,
|
| 928 |
-
"loss": 0.929,
|
| 929 |
-
"step": 2380
|
| 930 |
-
},
|
| 931 |
-
{
|
| 932 |
-
"epoch": 26.37,
|
| 933 |
-
"learning_rate": 0.00021088235294117647,
|
| 934 |
-
"loss": 0.8942,
|
| 935 |
-
"step": 2400
|
| 936 |
-
},
|
| 937 |
-
{
|
| 938 |
-
"epoch": 26.37,
|
| 939 |
-
"eval_loss": 0.16744859516620636,
|
| 940 |
-
"eval_runtime": 190.6796,
|
| 941 |
-
"eval_samples_per_second": 25.399,
|
| 942 |
-
"eval_steps_per_second": 0.797,
|
| 943 |
-
"eval_wer": 0.14916264547260857,
|
| 944 |
-
"step": 2400
|
| 945 |
-
},
|
| 946 |
-
{
|
| 947 |
-
"epoch": 26.59,
|
| 948 |
-
"learning_rate": 0.000208921568627451,
|
| 949 |
-
"loss": 0.8717,
|
| 950 |
-
"step": 2420
|
| 951 |
-
},
|
| 952 |
-
{
|
| 953 |
-
"epoch": 26.81,
|
| 954 |
-
"learning_rate": 0.0002069607843137255,
|
| 955 |
-
"loss": 0.8952,
|
| 956 |
-
"step": 2440
|
| 957 |
-
},
|
| 958 |
-
{
|
| 959 |
-
"epoch": 27.03,
|
| 960 |
-
"learning_rate": 0.000205,
|
| 961 |
-
"loss": 0.8849,
|
| 962 |
-
"step": 2460
|
| 963 |
-
},
|
| 964 |
-
{
|
| 965 |
-
"epoch": 27.25,
|
| 966 |
-
"learning_rate": 0.00020313725490196078,
|
| 967 |
-
"loss": 0.8781,
|
| 968 |
-
"step": 2480
|
| 969 |
-
},
|
| 970 |
-
{
|
| 971 |
-
"epoch": 27.47,
|
| 972 |
-
"learning_rate": 0.0002011764705882353,
|
| 973 |
-
"loss": 0.8701,
|
| 974 |
-
"step": 2500
|
| 975 |
-
},
|
| 976 |
-
{
|
| 977 |
-
"epoch": 27.47,
|
| 978 |
-
"eval_loss": 0.1738910973072052,
|
| 979 |
-
"eval_runtime": 187.7205,
|
| 980 |
-
"eval_samples_per_second": 25.799,
|
| 981 |
-
"eval_steps_per_second": 0.81,
|
| 982 |
-
"eval_wer": 0.1511779733181947,
|
| 983 |
-
"step": 2500
|
| 984 |
-
},
|
| 985 |
-
{
|
| 986 |
-
"epoch": 27.69,
|
| 987 |
-
"learning_rate": 0.0001992156862745098,
|
| 988 |
-
"loss": 0.8674,
|
| 989 |
-
"step": 2520
|
| 990 |
-
},
|
| 991 |
-
{
|
| 992 |
-
"epoch": 27.91,
|
| 993 |
-
"learning_rate": 0.00019725490196078433,
|
| 994 |
-
"loss": 0.8752,
|
| 995 |
-
"step": 2540
|
| 996 |
-
},
|
| 997 |
-
{
|
| 998 |
-
"epoch": 28.13,
|
| 999 |
-
"learning_rate": 0.0001952941176470588,
|
| 1000 |
-
"loss": 0.91,
|
| 1001 |
-
"step": 2560
|
| 1002 |
-
},
|
| 1003 |
-
{
|
| 1004 |
-
"epoch": 28.35,
|
| 1005 |
-
"learning_rate": 0.00019333333333333333,
|
| 1006 |
-
"loss": 0.8693,
|
| 1007 |
-
"step": 2580
|
| 1008 |
-
},
|
| 1009 |
-
{
|
| 1010 |
-
"epoch": 28.57,
|
| 1011 |
-
"learning_rate": 0.00019137254901960786,
|
| 1012 |
-
"loss": 0.8555,
|
| 1013 |
-
"step": 2600
|
| 1014 |
-
},
|
| 1015 |
-
{
|
| 1016 |
-
"epoch": 28.57,
|
| 1017 |
-
"eval_loss": 0.1689654141664505,
|
| 1018 |
-
"eval_runtime": 196.2665,
|
| 1019 |
-
"eval_samples_per_second": 24.676,
|
| 1020 |
-
"eval_steps_per_second": 0.774,
|
| 1021 |
-
"eval_wer": 0.14459267669599773,
|
| 1022 |
-
"step": 2600
|
| 1023 |
-
},
|
| 1024 |
-
{
|
| 1025 |
-
"epoch": 28.79,
|
| 1026 |
-
"learning_rate": 0.00018941176470588236,
|
| 1027 |
-
"loss": 0.8796,
|
| 1028 |
-
"step": 2620
|
| 1029 |
-
},
|
| 1030 |
-
{
|
| 1031 |
-
"epoch": 29.01,
|
| 1032 |
-
"learning_rate": 0.00018745098039215686,
|
| 1033 |
-
"loss": 0.9045,
|
| 1034 |
-
"step": 2640
|
| 1035 |
-
},
|
| 1036 |
-
{
|
| 1037 |
-
"epoch": 29.23,
|
| 1038 |
-
"learning_rate": 0.00018549019607843137,
|
| 1039 |
-
"loss": 0.8515,
|
| 1040 |
-
"step": 2660
|
| 1041 |
-
},
|
| 1042 |
-
{
|
| 1043 |
-
"epoch": 29.45,
|
| 1044 |
-
"learning_rate": 0.0001835294117647059,
|
| 1045 |
-
"loss": 0.861,
|
| 1046 |
-
"step": 2680
|
| 1047 |
-
},
|
| 1048 |
-
{
|
| 1049 |
-
"epoch": 29.67,
|
| 1050 |
-
"learning_rate": 0.0001815686274509804,
|
| 1051 |
-
"loss": 0.8513,
|
| 1052 |
-
"step": 2700
|
| 1053 |
-
},
|
| 1054 |
-
{
|
| 1055 |
-
"epoch": 29.67,
|
| 1056 |
-
"eval_loss": 0.16488835215568542,
|
| 1057 |
-
"eval_runtime": 189.1938,
|
| 1058 |
-
"eval_samples_per_second": 25.598,
|
| 1059 |
-
"eval_steps_per_second": 0.803,
|
| 1060 |
-
"eval_wer": 0.14774340051092819,
|
| 1061 |
-
"step": 2700
|
| 1062 |
-
},
|
| 1063 |
-
{
|
| 1064 |
-
"epoch": 29.89,
|
| 1065 |
-
"learning_rate": 0.0001796078431372549,
|
| 1066 |
-
"loss": 0.8573,
|
| 1067 |
-
"step": 2720
|
| 1068 |
-
},
|
| 1069 |
-
{
|
| 1070 |
-
"epoch": 30.11,
|
| 1071 |
-
"learning_rate": 0.00017764705882352942,
|
| 1072 |
-
"loss": 0.8686,
|
| 1073 |
-
"step": 2740
|
| 1074 |
-
},
|
| 1075 |
-
{
|
| 1076 |
-
"epoch": 30.33,
|
| 1077 |
-
"learning_rate": 0.00017568627450980392,
|
| 1078 |
-
"loss": 0.8462,
|
| 1079 |
-
"step": 2760
|
| 1080 |
-
},
|
| 1081 |
-
{
|
| 1082 |
-
"epoch": 30.55,
|
| 1083 |
-
"learning_rate": 0.00017372549019607845,
|
| 1084 |
-
"loss": 0.8419,
|
| 1085 |
-
"step": 2780
|
| 1086 |
-
},
|
| 1087 |
-
{
|
| 1088 |
-
"epoch": 30.77,
|
| 1089 |
-
"learning_rate": 0.00017176470588235293,
|
| 1090 |
-
"loss": 0.8659,
|
| 1091 |
-
"step": 2800
|
| 1092 |
-
},
|
| 1093 |
-
{
|
| 1094 |
-
"epoch": 30.77,
|
| 1095 |
-
"eval_loss": 0.1636602133512497,
|
| 1096 |
-
"eval_runtime": 190.245,
|
| 1097 |
-
"eval_samples_per_second": 25.457,
|
| 1098 |
-
"eval_steps_per_second": 0.799,
|
| 1099 |
-
"eval_wer": 0.1422367300596083,
|
| 1100 |
-
"step": 2800
|
| 1101 |
-
},
|
| 1102 |
-
{
|
| 1103 |
-
"epoch": 30.98,
|
| 1104 |
-
"learning_rate": 0.00016980392156862745,
|
| 1105 |
-
"loss": 0.8607,
|
| 1106 |
-
"step": 2820
|
| 1107 |
-
},
|
| 1108 |
-
{
|
| 1109 |
-
"epoch": 31.21,
|
| 1110 |
-
"learning_rate": 0.00016784313725490196,
|
| 1111 |
-
"loss": 0.8484,
|
| 1112 |
-
"step": 2840
|
| 1113 |
-
},
|
| 1114 |
-
{
|
| 1115 |
-
"epoch": 31.43,
|
| 1116 |
-
"learning_rate": 0.00016588235294117648,
|
| 1117 |
-
"loss": 0.8335,
|
| 1118 |
-
"step": 2860
|
| 1119 |
-
},
|
| 1120 |
-
{
|
| 1121 |
-
"epoch": 31.64,
|
| 1122 |
-
"learning_rate": 0.00016392156862745098,
|
| 1123 |
-
"loss": 0.8321,
|
| 1124 |
-
"step": 2880
|
| 1125 |
-
},
|
| 1126 |
-
{
|
| 1127 |
-
"epoch": 31.86,
|
| 1128 |
-
"learning_rate": 0.00016196078431372549,
|
| 1129 |
-
"loss": 0.8419,
|
| 1130 |
-
"step": 2900
|
| 1131 |
-
},
|
| 1132 |
-
{
|
| 1133 |
-
"epoch": 31.86,
|
| 1134 |
-
"eval_loss": 0.16137854754924774,
|
| 1135 |
-
"eval_runtime": 190.2053,
|
| 1136 |
-
"eval_samples_per_second": 25.462,
|
| 1137 |
-
"eval_steps_per_second": 0.799,
|
| 1138 |
-
"eval_wer": 0.1397104740278172,
|
| 1139 |
-
"step": 2900
|
| 1140 |
-
},
|
| 1141 |
-
{
|
| 1142 |
-
"epoch": 32.09,
|
| 1143 |
-
"learning_rate": 0.00016,
|
| 1144 |
-
"loss": 0.8634,
|
| 1145 |
-
"step": 2920
|
| 1146 |
-
},
|
| 1147 |
-
{
|
| 1148 |
-
"epoch": 32.31,
|
| 1149 |
-
"learning_rate": 0.00015803921568627451,
|
| 1150 |
-
"loss": 0.8351,
|
| 1151 |
-
"step": 2940
|
| 1152 |
-
},
|
| 1153 |
-
{
|
| 1154 |
-
"epoch": 32.52,
|
| 1155 |
-
"learning_rate": 0.00015607843137254901,
|
| 1156 |
-
"loss": 0.8427,
|
| 1157 |
-
"step": 2960
|
| 1158 |
-
},
|
| 1159 |
-
{
|
| 1160 |
-
"epoch": 32.74,
|
| 1161 |
-
"learning_rate": 0.00015411764705882352,
|
| 1162 |
-
"loss": 0.8324,
|
| 1163 |
-
"step": 2980
|
| 1164 |
-
},
|
| 1165 |
-
{
|
| 1166 |
-
"epoch": 32.96,
|
| 1167 |
-
"learning_rate": 0.00015215686274509804,
|
| 1168 |
-
"loss": 0.8491,
|
| 1169 |
-
"step": 3000
|
| 1170 |
-
},
|
| 1171 |
-
{
|
| 1172 |
-
"epoch": 32.96,
|
| 1173 |
-
"eval_loss": 0.1594749242067337,
|
| 1174 |
-
"eval_runtime": 191.4631,
|
| 1175 |
-
"eval_samples_per_second": 25.295,
|
| 1176 |
-
"eval_steps_per_second": 0.794,
|
| 1177 |
-
"eval_wer": 0.1401362475163213,
|
| 1178 |
-
"step": 3000
|
| 1179 |
-
},
|
| 1180 |
-
{
|
| 1181 |
-
"epoch": 33.19,
|
| 1182 |
-
"learning_rate": 0.00015019607843137257,
|
| 1183 |
-
"loss": 0.8263,
|
| 1184 |
-
"step": 3020
|
| 1185 |
-
},
|
| 1186 |
-
{
|
| 1187 |
-
"epoch": 33.4,
|
| 1188 |
-
"learning_rate": 0.00014823529411764705,
|
| 1189 |
-
"loss": 0.8203,
|
| 1190 |
-
"step": 3040
|
| 1191 |
-
},
|
| 1192 |
-
{
|
| 1193 |
-
"epoch": 33.62,
|
| 1194 |
-
"learning_rate": 0.00014627450980392157,
|
| 1195 |
-
"loss": 0.8277,
|
| 1196 |
-
"step": 3060
|
| 1197 |
-
},
|
| 1198 |
-
{
|
| 1199 |
-
"epoch": 33.84,
|
| 1200 |
-
"learning_rate": 0.00014431372549019607,
|
| 1201 |
-
"loss": 0.8183,
|
| 1202 |
-
"step": 3080
|
| 1203 |
-
},
|
| 1204 |
-
{
|
| 1205 |
-
"epoch": 34.07,
|
| 1206 |
-
"learning_rate": 0.0001423529411764706,
|
| 1207 |
-
"loss": 0.8395,
|
| 1208 |
-
"step": 3100
|
| 1209 |
-
},
|
| 1210 |
-
{
|
| 1211 |
-
"epoch": 34.07,
|
| 1212 |
-
"eval_loss": 0.1606692373752594,
|
| 1213 |
-
"eval_runtime": 190.8222,
|
| 1214 |
-
"eval_samples_per_second": 25.38,
|
| 1215 |
-
"eval_steps_per_second": 0.797,
|
| 1216 |
-
"eval_wer": 0.13755322168606302,
|
| 1217 |
-
"step": 3100
|
| 1218 |
-
},
|
| 1219 |
-
{
|
| 1220 |
-
"epoch": 34.28,
|
| 1221 |
-
"learning_rate": 0.0001403921568627451,
|
| 1222 |
-
"loss": 0.8207,
|
| 1223 |
-
"step": 3120
|
| 1224 |
-
},
|
| 1225 |
-
{
|
| 1226 |
-
"epoch": 34.5,
|
| 1227 |
-
"learning_rate": 0.0001384313725490196,
|
| 1228 |
-
"loss": 0.8106,
|
| 1229 |
-
"step": 3140
|
| 1230 |
-
},
|
| 1231 |
-
{
|
| 1232 |
-
"epoch": 34.72,
|
| 1233 |
-
"learning_rate": 0.00013647058823529413,
|
| 1234 |
-
"loss": 0.8329,
|
| 1235 |
-
"step": 3160
|
| 1236 |
-
},
|
| 1237 |
-
{
|
| 1238 |
-
"epoch": 34.94,
|
| 1239 |
-
"learning_rate": 0.00013450980392156863,
|
| 1240 |
-
"loss": 0.8048,
|
| 1241 |
-
"step": 3180
|
| 1242 |
-
},
|
| 1243 |
-
{
|
| 1244 |
-
"epoch": 35.16,
|
| 1245 |
-
"learning_rate": 0.00013254901960784313,
|
| 1246 |
-
"loss": 0.83,
|
| 1247 |
-
"step": 3200
|
| 1248 |
-
},
|
| 1249 |
-
{
|
| 1250 |
-
"epoch": 35.16,
|
| 1251 |
-
"eval_loss": 0.15384173393249512,
|
| 1252 |
-
"eval_runtime": 191.1912,
|
| 1253 |
-
"eval_samples_per_second": 25.331,
|
| 1254 |
-
"eval_steps_per_second": 0.795,
|
| 1255 |
-
"eval_wer": 0.1379222253760999,
|
| 1256 |
-
"step": 3200
|
| 1257 |
-
},
|
| 1258 |
-
{
|
| 1259 |
-
"epoch": 35.38,
|
| 1260 |
-
"learning_rate": 0.00013058823529411764,
|
| 1261 |
-
"loss": 0.7963,
|
| 1262 |
-
"step": 3220
|
| 1263 |
-
},
|
| 1264 |
-
{
|
| 1265 |
-
"epoch": 35.6,
|
| 1266 |
-
"learning_rate": 0.00012862745098039216,
|
| 1267 |
-
"loss": 0.7895,
|
| 1268 |
-
"step": 3240
|
| 1269 |
-
},
|
| 1270 |
-
{
|
| 1271 |
-
"epoch": 35.82,
|
| 1272 |
-
"learning_rate": 0.0001266666666666667,
|
| 1273 |
-
"loss": 0.7964,
|
| 1274 |
-
"step": 3260
|
| 1275 |
-
},
|
| 1276 |
-
{
|
| 1277 |
-
"epoch": 36.04,
|
| 1278 |
-
"learning_rate": 0.0001247058823529412,
|
| 1279 |
-
"loss": 0.7931,
|
| 1280 |
-
"step": 3280
|
| 1281 |
-
},
|
| 1282 |
-
{
|
| 1283 |
-
"epoch": 36.26,
|
| 1284 |
-
"learning_rate": 0.0001227450980392157,
|
| 1285 |
-
"loss": 0.7835,
|
| 1286 |
-
"step": 3300
|
| 1287 |
-
},
|
| 1288 |
-
{
|
| 1289 |
-
"epoch": 36.26,
|
| 1290 |
-
"eval_loss": 0.16020993888378143,
|
| 1291 |
-
"eval_runtime": 192.1405,
|
| 1292 |
-
"eval_samples_per_second": 25.206,
|
| 1293 |
-
"eval_steps_per_second": 0.791,
|
| 1294 |
-
"eval_wer": 0.1408458699971615,
|
| 1295 |
-
"step": 3300
|
| 1296 |
-
},
|
| 1297 |
-
{
|
| 1298 |
-
"epoch": 36.48,
|
| 1299 |
-
"learning_rate": 0.00012078431372549021,
|
| 1300 |
-
"loss": 0.772,
|
| 1301 |
-
"step": 3320
|
| 1302 |
-
},
|
| 1303 |
-
{
|
| 1304 |
-
"epoch": 36.7,
|
| 1305 |
-
"learning_rate": 0.00011882352941176471,
|
| 1306 |
-
"loss": 0.7818,
|
| 1307 |
-
"step": 3340
|
| 1308 |
-
},
|
| 1309 |
-
{
|
| 1310 |
-
"epoch": 36.92,
|
| 1311 |
-
"learning_rate": 0.00011696078431372549,
|
| 1312 |
-
"loss": 0.8016,
|
| 1313 |
-
"step": 3360
|
| 1314 |
-
},
|
| 1315 |
-
{
|
| 1316 |
-
"epoch": 37.14,
|
| 1317 |
-
"learning_rate": 0.000115,
|
| 1318 |
-
"loss": 0.8061,
|
| 1319 |
-
"step": 3380
|
| 1320 |
-
},
|
| 1321 |
-
{
|
| 1322 |
-
"epoch": 37.36,
|
| 1323 |
-
"learning_rate": 0.0001130392156862745,
|
| 1324 |
-
"loss": 0.7703,
|
| 1325 |
-
"step": 3400
|
| 1326 |
-
},
|
| 1327 |
-
{
|
| 1328 |
-
"epoch": 37.36,
|
| 1329 |
-
"eval_loss": 0.16011376678943634,
|
| 1330 |
-
"eval_runtime": 187.5367,
|
| 1331 |
-
"eval_samples_per_second": 25.824,
|
| 1332 |
-
"eval_steps_per_second": 0.811,
|
| 1333 |
-
"eval_wer": 0.13692875390292364,
|
| 1334 |
-
"step": 3400
|
| 1335 |
-
},
|
| 1336 |
-
{
|
| 1337 |
-
"epoch": 37.58,
|
| 1338 |
-
"learning_rate": 0.00011107843137254903,
|
| 1339 |
-
"loss": 0.7713,
|
| 1340 |
-
"step": 3420
|
| 1341 |
-
},
|
| 1342 |
-
{
|
| 1343 |
-
"epoch": 37.8,
|
| 1344 |
-
"learning_rate": 0.00010911764705882353,
|
| 1345 |
-
"loss": 0.7712,
|
| 1346 |
-
"step": 3440
|
| 1347 |
-
},
|
| 1348 |
-
{
|
| 1349 |
-
"epoch": 38.02,
|
| 1350 |
-
"learning_rate": 0.00010715686274509805,
|
| 1351 |
-
"loss": 0.8149,
|
| 1352 |
-
"step": 3460
|
| 1353 |
-
},
|
| 1354 |
-
{
|
| 1355 |
-
"epoch": 38.24,
|
| 1356 |
-
"learning_rate": 0.00010519607843137255,
|
| 1357 |
-
"loss": 0.7885,
|
| 1358 |
-
"step": 3480
|
| 1359 |
-
},
|
| 1360 |
-
{
|
| 1361 |
-
"epoch": 38.46,
|
| 1362 |
-
"learning_rate": 0.00010323529411764706,
|
| 1363 |
-
"loss": 0.7474,
|
| 1364 |
-
"step": 3500
|
| 1365 |
-
},
|
| 1366 |
-
{
|
| 1367 |
-
"epoch": 38.46,
|
| 1368 |
-
"eval_loss": 0.1514146625995636,
|
| 1369 |
-
"eval_runtime": 190.8228,
|
| 1370 |
-
"eval_samples_per_second": 25.38,
|
| 1371 |
-
"eval_steps_per_second": 0.797,
|
| 1372 |
-
"eval_wer": 0.1342321884757309,
|
| 1373 |
-
"step": 3500
|
| 1374 |
-
},
|
| 1375 |
-
{
|
| 1376 |
-
"epoch": 38.68,
|
| 1377 |
-
"learning_rate": 0.00010127450980392156,
|
| 1378 |
-
"loss": 0.785,
|
| 1379 |
-
"step": 3520
|
| 1380 |
-
},
|
| 1381 |
-
{
|
| 1382 |
-
"epoch": 38.9,
|
| 1383 |
-
"learning_rate": 9.931372549019609e-05,
|
| 1384 |
-
"loss": 0.7677,
|
| 1385 |
-
"step": 3540
|
| 1386 |
-
},
|
| 1387 |
-
{
|
| 1388 |
-
"epoch": 39.12,
|
| 1389 |
-
"learning_rate": 9.73529411764706e-05,
|
| 1390 |
-
"loss": 0.7849,
|
| 1391 |
-
"step": 3560
|
| 1392 |
-
},
|
| 1393 |
-
{
|
| 1394 |
-
"epoch": 39.34,
|
| 1395 |
-
"learning_rate": 9.539215686274511e-05,
|
| 1396 |
-
"loss": 0.7637,
|
| 1397 |
-
"step": 3580
|
| 1398 |
-
},
|
| 1399 |
-
{
|
| 1400 |
-
"epoch": 39.56,
|
| 1401 |
-
"learning_rate": 9.343137254901961e-05,
|
| 1402 |
-
"loss": 0.7719,
|
| 1403 |
-
"step": 3600
|
| 1404 |
-
},
|
| 1405 |
-
{
|
| 1406 |
-
"epoch": 39.56,
|
| 1407 |
-
"eval_loss": 0.15932896733283997,
|
| 1408 |
-
"eval_runtime": 189.6806,
|
| 1409 |
-
"eval_samples_per_second": 25.532,
|
| 1410 |
-
"eval_steps_per_second": 0.801,
|
| 1411 |
-
"eval_wer": 0.1352540448481408,
|
| 1412 |
-
"step": 3600
|
| 1413 |
-
},
|
| 1414 |
-
{
|
| 1415 |
-
"epoch": 39.78,
|
| 1416 |
-
"learning_rate": 9.147058823529412e-05,
|
| 1417 |
-
"loss": 0.7591,
|
| 1418 |
-
"step": 3620
|
| 1419 |
-
},
|
| 1420 |
-
{
|
| 1421 |
-
"epoch": 39.99,
|
| 1422 |
-
"learning_rate": 8.950980392156862e-05,
|
| 1423 |
-
"loss": 0.7706,
|
| 1424 |
-
"step": 3640
|
| 1425 |
-
},
|
| 1426 |
-
{
|
| 1427 |
-
"epoch": 40.22,
|
| 1428 |
-
"learning_rate": 8.754901960784314e-05,
|
| 1429 |
-
"loss": 0.7805,
|
| 1430 |
-
"step": 3660
|
| 1431 |
-
},
|
| 1432 |
-
{
|
| 1433 |
-
"epoch": 40.44,
|
| 1434 |
-
"learning_rate": 8.558823529411765e-05,
|
| 1435 |
-
"loss": 0.7753,
|
| 1436 |
-
"step": 3680
|
| 1437 |
-
},
|
| 1438 |
-
{
|
| 1439 |
-
"epoch": 40.66,
|
| 1440 |
-
"learning_rate": 8.362745098039217e-05,
|
| 1441 |
-
"loss": 0.7638,
|
| 1442 |
-
"step": 3700
|
| 1443 |
-
},
|
| 1444 |
-
{
|
| 1445 |
-
"epoch": 40.66,
|
| 1446 |
-
"eval_loss": 0.15362653136253357,
|
| 1447 |
-
"eval_runtime": 192.1118,
|
| 1448 |
-
"eval_samples_per_second": 25.209,
|
| 1449 |
-
"eval_steps_per_second": 0.791,
|
| 1450 |
-
"eval_wer": 0.13380641498722678,
|
| 1451 |
-
"step": 3700
|
| 1452 |
-
},
|
| 1453 |
-
{
|
| 1454 |
-
"epoch": 40.87,
|
| 1455 |
-
"learning_rate": 8.166666666666667e-05,
|
| 1456 |
-
"loss": 0.7791,
|
| 1457 |
-
"step": 3720
|
| 1458 |
-
},
|
| 1459 |
-
{
|
| 1460 |
-
"epoch": 41.1,
|
| 1461 |
-
"learning_rate": 7.970588235294118e-05,
|
| 1462 |
-
"loss": 0.7931,
|
| 1463 |
-
"step": 3740
|
| 1464 |
-
},
|
| 1465 |
-
{
|
| 1466 |
-
"epoch": 41.32,
|
| 1467 |
-
"learning_rate": 7.774509803921568e-05,
|
| 1468 |
-
"loss": 0.741,
|
| 1469 |
-
"step": 3760
|
| 1470 |
-
},
|
| 1471 |
-
{
|
| 1472 |
-
"epoch": 41.54,
|
| 1473 |
-
"learning_rate": 7.57843137254902e-05,
|
| 1474 |
-
"loss": 0.7434,
|
| 1475 |
-
"step": 3780
|
| 1476 |
-
},
|
| 1477 |
-
{
|
| 1478 |
-
"epoch": 41.75,
|
| 1479 |
-
"learning_rate": 7.38235294117647e-05,
|
| 1480 |
-
"loss": 0.771,
|
| 1481 |
-
"step": 3800
|
| 1482 |
-
},
|
| 1483 |
-
{
|
| 1484 |
-
"epoch": 41.75,
|
| 1485 |
-
"eval_loss": 0.1530592143535614,
|
| 1486 |
-
"eval_runtime": 186.414,
|
| 1487 |
-
"eval_samples_per_second": 25.98,
|
| 1488 |
-
"eval_steps_per_second": 0.815,
|
| 1489 |
-
"eval_wer": 0.13170593244393983,
|
| 1490 |
-
"step": 3800
|
| 1491 |
-
},
|
| 1492 |
-
{
|
| 1493 |
-
"epoch": 41.97,
|
| 1494 |
-
"learning_rate": 7.186274509803923e-05,
|
| 1495 |
-
"loss": 0.7765,
|
| 1496 |
-
"step": 3820
|
| 1497 |
-
},
|
| 1498 |
-
{
|
| 1499 |
-
"epoch": 42.2,
|
| 1500 |
-
"learning_rate": 6.990196078431373e-05,
|
| 1501 |
-
"loss": 0.7599,
|
| 1502 |
-
"step": 3840
|
| 1503 |
-
},
|
| 1504 |
-
{
|
| 1505 |
-
"epoch": 42.42,
|
| 1506 |
-
"learning_rate": 6.794117647058824e-05,
|
| 1507 |
-
"loss": 0.7782,
|
| 1508 |
-
"step": 3860
|
| 1509 |
-
},
|
| 1510 |
-
{
|
| 1511 |
-
"epoch": 42.63,
|
| 1512 |
-
"learning_rate": 6.598039215686274e-05,
|
| 1513 |
-
"loss": 0.7395,
|
| 1514 |
-
"step": 3880
|
| 1515 |
-
},
|
| 1516 |
-
{
|
| 1517 |
-
"epoch": 42.85,
|
| 1518 |
-
"learning_rate": 6.401960784313726e-05,
|
| 1519 |
-
"loss": 0.7594,
|
| 1520 |
-
"step": 3900
|
| 1521 |
-
},
|
| 1522 |
-
{
|
| 1523 |
-
"epoch": 42.85,
|
| 1524 |
-
"eval_loss": 0.14983513951301575,
|
| 1525 |
-
"eval_runtime": 190.4439,
|
| 1526 |
-
"eval_samples_per_second": 25.43,
|
| 1527 |
-
"eval_steps_per_second": 0.798,
|
| 1528 |
-
"eval_wer": 0.12883905762134545,
|
| 1529 |
-
"step": 3900
|
| 1530 |
-
},
|
| 1531 |
-
{
|
| 1532 |
-
"epoch": 43.08,
|
| 1533 |
-
"learning_rate": 6.205882352941176e-05,
|
| 1534 |
-
"loss": 0.7691,
|
| 1535 |
-
"step": 3920
|
| 1536 |
-
},
|
| 1537 |
-
{
|
| 1538 |
-
"epoch": 43.3,
|
| 1539 |
-
"learning_rate": 6.0098039215686274e-05,
|
| 1540 |
-
"loss": 0.728,
|
| 1541 |
-
"step": 3940
|
| 1542 |
-
},
|
| 1543 |
-
{
|
| 1544 |
-
"epoch": 43.51,
|
| 1545 |
-
"learning_rate": 5.813725490196078e-05,
|
| 1546 |
-
"loss": 0.7585,
|
| 1547 |
-
"step": 3960
|
| 1548 |
-
},
|
| 1549 |
-
{
|
| 1550 |
-
"epoch": 43.73,
|
| 1551 |
-
"learning_rate": 5.6176470588235296e-05,
|
| 1552 |
-
"loss": 0.7564,
|
| 1553 |
-
"step": 3980
|
| 1554 |
-
},
|
| 1555 |
-
{
|
| 1556 |
-
"epoch": 43.95,
|
| 1557 |
-
"learning_rate": 5.4215686274509804e-05,
|
| 1558 |
-
"loss": 0.7383,
|
| 1559 |
-
"step": 4000
|
| 1560 |
-
},
|
| 1561 |
-
{
|
| 1562 |
-
"epoch": 43.95,
|
| 1563 |
-
"eval_loss": 0.15269021689891815,
|
| 1564 |
-
"eval_runtime": 188.5816,
|
| 1565 |
-
"eval_samples_per_second": 25.681,
|
| 1566 |
-
"eval_steps_per_second": 0.806,
|
| 1567 |
-
"eval_wer": 0.13003122338915696,
|
| 1568 |
-
"step": 4000
|
| 1569 |
-
},
|
| 1570 |
-
{
|
| 1571 |
-
"epoch": 44.17,
|
| 1572 |
-
"learning_rate": 5.225490196078431e-05,
|
| 1573 |
-
"loss": 0.7693,
|
| 1574 |
-
"step": 4020
|
| 1575 |
-
},
|
| 1576 |
-
{
|
| 1577 |
-
"epoch": 44.39,
|
| 1578 |
-
"learning_rate": 5.0294117647058826e-05,
|
| 1579 |
-
"loss": 0.7347,
|
| 1580 |
-
"step": 4040
|
| 1581 |
-
},
|
| 1582 |
-
{
|
| 1583 |
-
"epoch": 44.61,
|
| 1584 |
-
"learning_rate": 4.8333333333333334e-05,
|
| 1585 |
-
"loss": 0.7185,
|
| 1586 |
-
"step": 4060
|
| 1587 |
-
},
|
| 1588 |
-
{
|
| 1589 |
-
"epoch": 44.83,
|
| 1590 |
-
"learning_rate": 4.637254901960784e-05,
|
| 1591 |
-
"loss": 0.7394,
|
| 1592 |
-
"step": 4080
|
| 1593 |
-
},
|
| 1594 |
-
{
|
| 1595 |
-
"epoch": 45.05,
|
| 1596 |
-
"learning_rate": 4.4411764705882356e-05,
|
| 1597 |
-
"loss": 0.7565,
|
| 1598 |
-
"step": 4100
|
| 1599 |
-
},
|
| 1600 |
-
{
|
| 1601 |
-
"epoch": 45.05,
|
| 1602 |
-
"eval_loss": 0.14823457598686218,
|
| 1603 |
-
"eval_runtime": 191.1254,
|
| 1604 |
-
"eval_samples_per_second": 25.339,
|
| 1605 |
-
"eval_steps_per_second": 0.795,
|
| 1606 |
-
"eval_wer": 0.12886744252057905,
|
| 1607 |
-
"step": 4100
|
| 1608 |
-
},
|
| 1609 |
-
{
|
| 1610 |
-
"epoch": 45.27,
|
| 1611 |
-
"learning_rate": 4.2450980392156864e-05,
|
| 1612 |
-
"loss": 0.7535,
|
| 1613 |
-
"step": 4120
|
| 1614 |
-
},
|
| 1615 |
-
{
|
| 1616 |
-
"epoch": 45.49,
|
| 1617 |
-
"learning_rate": 4.049019607843137e-05,
|
| 1618 |
-
"loss": 0.737,
|
| 1619 |
-
"step": 4140
|
| 1620 |
-
},
|
| 1621 |
-
{
|
| 1622 |
-
"epoch": 45.71,
|
| 1623 |
-
"learning_rate": 3.852941176470588e-05,
|
| 1624 |
-
"loss": 0.7501,
|
| 1625 |
-
"step": 4160
|
| 1626 |
-
},
|
| 1627 |
-
{
|
| 1628 |
-
"epoch": 45.93,
|
| 1629 |
-
"learning_rate": 3.6568627450980393e-05,
|
| 1630 |
-
"loss": 0.7285,
|
| 1631 |
-
"step": 4180
|
| 1632 |
-
},
|
| 1633 |
-
{
|
| 1634 |
-
"epoch": 46.15,
|
| 1635 |
-
"learning_rate": 3.46078431372549e-05,
|
| 1636 |
-
"loss": 0.7697,
|
| 1637 |
-
"step": 4200
|
| 1638 |
-
},
|
| 1639 |
-
{
|
| 1640 |
-
"epoch": 46.15,
|
| 1641 |
-
"eval_loss": 0.1494804471731186,
|
| 1642 |
-
"eval_runtime": 190.3868,
|
| 1643 |
-
"eval_samples_per_second": 25.438,
|
| 1644 |
-
"eval_steps_per_second": 0.798,
|
| 1645 |
-
"eval_wer": 0.1271927334657962,
|
| 1646 |
-
"step": 4200
|
| 1647 |
-
},
|
| 1648 |
-
{
|
| 1649 |
-
"epoch": 46.37,
|
| 1650 |
-
"learning_rate": 3.264705882352941e-05,
|
| 1651 |
-
"loss": 0.7283,
|
| 1652 |
-
"step": 4220
|
| 1653 |
-
},
|
| 1654 |
-
{
|
| 1655 |
-
"epoch": 46.59,
|
| 1656 |
-
"learning_rate": 3.0686274509803923e-05,
|
| 1657 |
-
"loss": 0.7218,
|
| 1658 |
-
"step": 4240
|
| 1659 |
-
},
|
| 1660 |
-
{
|
| 1661 |
-
"epoch": 46.81,
|
| 1662 |
-
"learning_rate": 2.872549019607843e-05,
|
| 1663 |
-
"loss": 0.7341,
|
| 1664 |
-
"step": 4260
|
| 1665 |
-
},
|
| 1666 |
-
{
|
| 1667 |
-
"epoch": 47.03,
|
| 1668 |
-
"learning_rate": 2.6764705882352942e-05,
|
| 1669 |
-
"loss": 0.7293,
|
| 1670 |
-
"step": 4280
|
| 1671 |
-
},
|
| 1672 |
-
{
|
| 1673 |
-
"epoch": 47.25,
|
| 1674 |
-
"learning_rate": 2.480392156862745e-05,
|
| 1675 |
-
"loss": 0.7194,
|
| 1676 |
-
"step": 4300
|
| 1677 |
-
},
|
| 1678 |
-
{
|
| 1679 |
-
"epoch": 47.25,
|
| 1680 |
-
"eval_loss": 0.14928147196769714,
|
| 1681 |
-
"eval_runtime": 189.625,
|
| 1682 |
-
"eval_samples_per_second": 25.54,
|
| 1683 |
-
"eval_steps_per_second": 0.802,
|
| 1684 |
-
"eval_wer": 0.12693726937269373,
|
| 1685 |
-
"step": 4300
|
| 1686 |
-
},
|
| 1687 |
-
{
|
| 1688 |
-
"epoch": 47.47,
|
| 1689 |
-
"learning_rate": 2.284313725490196e-05,
|
| 1690 |
-
"loss": 0.7326,
|
| 1691 |
-
"step": 4320
|
| 1692 |
-
},
|
| 1693 |
-
{
|
| 1694 |
-
"epoch": 47.69,
|
| 1695 |
-
"learning_rate": 2.088235294117647e-05,
|
| 1696 |
-
"loss": 0.724,
|
| 1697 |
-
"step": 4340
|
| 1698 |
-
},
|
| 1699 |
-
{
|
| 1700 |
-
"epoch": 47.91,
|
| 1701 |
-
"learning_rate": 1.892156862745098e-05,
|
| 1702 |
-
"loss": 0.7316,
|
| 1703 |
-
"step": 4360
|
| 1704 |
-
},
|
| 1705 |
-
{
|
| 1706 |
-
"epoch": 48.13,
|
| 1707 |
-
"learning_rate": 1.696078431372549e-05,
|
| 1708 |
-
"loss": 0.7566,
|
| 1709 |
-
"step": 4380
|
| 1710 |
-
},
|
| 1711 |
-
{
|
| 1712 |
-
"epoch": 48.35,
|
| 1713 |
-
"learning_rate": 1.5e-05,
|
| 1714 |
-
"loss": 0.7479,
|
| 1715 |
-
"step": 4400
|
| 1716 |
-
},
|
| 1717 |
-
{
|
| 1718 |
-
"epoch": 48.35,
|
| 1719 |
-
"eval_loss": 0.1490115076303482,
|
| 1720 |
-
"eval_runtime": 188.3681,
|
| 1721 |
-
"eval_samples_per_second": 25.71,
|
| 1722 |
-
"eval_steps_per_second": 0.807,
|
| 1723 |
-
"eval_wer": 0.12761850695430033,
|
| 1724 |
-
"step": 4400
|
| 1725 |
-
},
|
| 1726 |
-
{
|
| 1727 |
-
"epoch": 48.57,
|
| 1728 |
-
"learning_rate": 1.303921568627451e-05,
|
| 1729 |
-
"loss": 0.7139,
|
| 1730 |
-
"step": 4420
|
| 1731 |
-
},
|
| 1732 |
-
{
|
| 1733 |
-
"epoch": 48.79,
|
| 1734 |
-
"learning_rate": 1.107843137254902e-05,
|
| 1735 |
-
"loss": 0.7345,
|
| 1736 |
-
"step": 4440
|
| 1737 |
-
},
|
| 1738 |
-
{
|
| 1739 |
-
"epoch": 49.01,
|
| 1740 |
-
"learning_rate": 9.117647058823529e-06,
|
| 1741 |
-
"loss": 0.756,
|
| 1742 |
-
"step": 4460
|
| 1743 |
-
},
|
| 1744 |
-
{
|
| 1745 |
-
"epoch": 49.23,
|
| 1746 |
-
"learning_rate": 7.15686274509804e-06,
|
| 1747 |
-
"loss": 0.7196,
|
| 1748 |
-
"step": 4480
|
| 1749 |
-
},
|
| 1750 |
-
{
|
| 1751 |
-
"epoch": 49.45,
|
| 1752 |
-
"learning_rate": 5.196078431372549e-06,
|
| 1753 |
-
"loss": 0.7132,
|
| 1754 |
-
"step": 4500
|
| 1755 |
-
},
|
| 1756 |
-
{
|
| 1757 |
-
"epoch": 49.45,
|
| 1758 |
-
"eval_loss": 0.1501067876815796,
|
| 1759 |
-
"eval_runtime": 190.1989,
|
| 1760 |
-
"eval_samples_per_second": 25.463,
|
| 1761 |
-
"eval_steps_per_second": 0.799,
|
| 1762 |
-
"eval_wer": 0.1265398807834232,
|
| 1763 |
-
"step": 4500
|
| 1764 |
-
}
|
| 1765 |
-
],
|
| 1766 |
-
"max_steps": 4550,
|
| 1767 |
-
"num_train_epochs": 50,
|
| 1768 |
-
"total_flos": 6.9574010203448205e+19,
|
| 1769 |
-
"trial_name": null,
|
| 1770 |
-
"trial_params": null
|
| 1771 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-4500/training_args.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:61c53503eb760641941023d2dcab7bb32b620bf895e31db5c3910b2760135b07
|
| 3 |
-
size 3055
|
|
|
|
|
|
|
|
|
|
|
|
old_run_asr_ctc.py → join_datasets_asr_ctc.py
RENAMED
|
File without changes
|
old_run.sh → join_datasets_asr_ctc_run.sh
RENAMED
|
File without changes
|