Commit
•
63865a1
1
Parent(s):
1b2d3ae
training
Browse files- all_results.json +10 -10
- checkpoint-1000/config.json +0 -41
- checkpoint-1000/optimizer.pt +0 -3
- checkpoint-1000/preprocessor_config.json +0 -0
- checkpoint-1000/pytorch_model.bin +0 -3
- checkpoint-1000/rng_state.pth +0 -3
- checkpoint-1000/scaler.pt +0 -3
- checkpoint-1000/scheduler.pt +0 -3
- checkpoint-1000/trainer_state.json +0 -265
- checkpoint-1000/training_args.bin +0 -3
- checkpoint-2000/config.json +0 -41
- checkpoint-2000/optimizer.pt +0 -3
- checkpoint-2000/preprocessor_config.json +0 -0
- checkpoint-2000/pytorch_model.bin +0 -3
- checkpoint-2000/rng_state.pth +0 -3
- checkpoint-2000/scaler.pt +0 -3
- checkpoint-2000/scheduler.pt +0 -3
- checkpoint-2000/trainer_state.json +0 -514
- checkpoint-2000/training_args.bin +0 -3
- eval_results.json +6 -6
- pytorch_model.bin +1 -1
- run.sh +9 -11
- run_speech_recognition_seq2seq_streaming.py +217 -38
- train_results.json +5 -5
- trainer_state.json +941 -314
- training_args.bin +1 -1
all_results.json
CHANGED
|
@@ -1,12 +1,12 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"eval_loss": 0.
|
| 4 |
-
"eval_runtime":
|
| 5 |
-
"eval_samples_per_second": 1.
|
| 6 |
-
"eval_steps_per_second": 0.
|
| 7 |
-
"eval_wer":
|
| 8 |
-
"train_loss": 0.
|
| 9 |
-
"train_runtime":
|
| 10 |
-
"train_samples_per_second": 3.
|
| 11 |
-
"train_steps_per_second": 0.
|
| 12 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_loss": 0.24834245443344116,
|
| 4 |
+
"eval_runtime": 2999.4256,
|
| 5 |
+
"eval_samples_per_second": 1.69,
|
| 6 |
+
"eval_steps_per_second": 0.106,
|
| 7 |
+
"eval_wer": 9.891409525857435,
|
| 8 |
+
"train_loss": 0.025400285175442697,
|
| 9 |
+
"train_runtime": 51804.3597,
|
| 10 |
+
"train_samples_per_second": 3.089,
|
| 11 |
+
"train_steps_per_second": 0.097
|
| 12 |
}
|
checkpoint-1000/config.json
DELETED
|
@@ -1,41 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_name_or_path": "marinone94/whisper-medium-nordic",
|
| 3 |
-
"activation_dropout": 0.0,
|
| 4 |
-
"activation_function": "gelu",
|
| 5 |
-
"architectures": [
|
| 6 |
-
"WhisperForConditionalGeneration"
|
| 7 |
-
],
|
| 8 |
-
"attention_dropout": 0.0,
|
| 9 |
-
"begin_suppress_tokens": [
|
| 10 |
-
220,
|
| 11 |
-
50257
|
| 12 |
-
],
|
| 13 |
-
"bos_token_id": 50257,
|
| 14 |
-
"d_model": 1024,
|
| 15 |
-
"decoder_attention_heads": 16,
|
| 16 |
-
"decoder_ffn_dim": 4096,
|
| 17 |
-
"decoder_layerdrop": 0.0,
|
| 18 |
-
"decoder_layers": 24,
|
| 19 |
-
"decoder_start_token_id": 50258,
|
| 20 |
-
"dropout": 0.0,
|
| 21 |
-
"encoder_attention_heads": 16,
|
| 22 |
-
"encoder_ffn_dim": 4096,
|
| 23 |
-
"encoder_layerdrop": 0.0,
|
| 24 |
-
"encoder_layers": 24,
|
| 25 |
-
"eos_token_id": 50257,
|
| 26 |
-
"forced_decoder_ids": null,
|
| 27 |
-
"init_std": 0.02,
|
| 28 |
-
"is_encoder_decoder": true,
|
| 29 |
-
"max_length": 448,
|
| 30 |
-
"max_source_positions": 1500,
|
| 31 |
-
"max_target_positions": 448,
|
| 32 |
-
"model_type": "whisper",
|
| 33 |
-
"num_hidden_layers": 24,
|
| 34 |
-
"num_mel_bins": 80,
|
| 35 |
-
"pad_token_id": 50257,
|
| 36 |
-
"scale_embedding": false,
|
| 37 |
-
"torch_dtype": "float32",
|
| 38 |
-
"transformers_version": "4.26.0.dev0",
|
| 39 |
-
"use_cache": false,
|
| 40 |
-
"vocab_size": 51865
|
| 41 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-1000/optimizer.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:2bfff8704bd6704fbfd4aa9975e9d6d8362b15cbf7113f7f23f0eaefcad253fc
|
| 3 |
-
size 6111428695
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-1000/preprocessor_config.json
DELETED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-1000/pytorch_model.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:52c4f954de9065deba6c2d5e400f27179e3eb7070e6948847c74977d9c96a8c4
|
| 3 |
-
size 3055754841
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-1000/rng_state.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:9d2e0f4427086b473c2d8b746cd9a0c73b8d6aa2bbde1faf6cb2428e78028a09
|
| 3 |
-
size 14575
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-1000/scaler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:26dbe439de3dd73741374b46df74ee981e9ad3b04a0c5c8a564c625a65b8aee5
|
| 3 |
-
size 557
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-1000/scheduler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:58c1ed193257a9c4b6ec19ec7a1a09d3fba65f6437077e3c0f61b03457bba09d
|
| 3 |
-
size 627
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-1000/trainer_state.json
DELETED
|
@@ -1,265 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"best_metric": 11.942264272658853,
|
| 3 |
-
"best_model_checkpoint": "./checkpoint-1000",
|
| 4 |
-
"epoch": 3.0208,
|
| 5 |
-
"global_step": 1000,
|
| 6 |
-
"is_hyper_param_search": false,
|
| 7 |
-
"is_local_process_zero": true,
|
| 8 |
-
"is_world_process_zero": true,
|
| 9 |
-
"log_history": [
|
| 10 |
-
{
|
| 11 |
-
"epoch": 0.01,
|
| 12 |
-
"learning_rate": 9.200000000000001e-07,
|
| 13 |
-
"loss": 0.5204,
|
| 14 |
-
"step": 25
|
| 15 |
-
},
|
| 16 |
-
{
|
| 17 |
-
"epoch": 0.02,
|
| 18 |
-
"learning_rate": 1.9200000000000003e-06,
|
| 19 |
-
"loss": 0.0582,
|
| 20 |
-
"step": 50
|
| 21 |
-
},
|
| 22 |
-
{
|
| 23 |
-
"epoch": 0.03,
|
| 24 |
-
"learning_rate": 2.92e-06,
|
| 25 |
-
"loss": 0.0527,
|
| 26 |
-
"step": 75
|
| 27 |
-
},
|
| 28 |
-
{
|
| 29 |
-
"epoch": 0.04,
|
| 30 |
-
"learning_rate": 3.920000000000001e-06,
|
| 31 |
-
"loss": 0.0587,
|
| 32 |
-
"step": 100
|
| 33 |
-
},
|
| 34 |
-
{
|
| 35 |
-
"epoch": 0.05,
|
| 36 |
-
"learning_rate": 4.92e-06,
|
| 37 |
-
"loss": 0.0659,
|
| 38 |
-
"step": 125
|
| 39 |
-
},
|
| 40 |
-
{
|
| 41 |
-
"epoch": 0.06,
|
| 42 |
-
"learning_rate": 5.92e-06,
|
| 43 |
-
"loss": 0.0681,
|
| 44 |
-
"step": 150
|
| 45 |
-
},
|
| 46 |
-
{
|
| 47 |
-
"epoch": 0.07,
|
| 48 |
-
"learning_rate": 6.92e-06,
|
| 49 |
-
"loss": 0.07,
|
| 50 |
-
"step": 175
|
| 51 |
-
},
|
| 52 |
-
{
|
| 53 |
-
"epoch": 0.08,
|
| 54 |
-
"learning_rate": 7.92e-06,
|
| 55 |
-
"loss": 0.0856,
|
| 56 |
-
"step": 200
|
| 57 |
-
},
|
| 58 |
-
{
|
| 59 |
-
"epoch": 0.09,
|
| 60 |
-
"learning_rate": 8.920000000000001e-06,
|
| 61 |
-
"loss": 0.0894,
|
| 62 |
-
"step": 225
|
| 63 |
-
},
|
| 64 |
-
{
|
| 65 |
-
"epoch": 0.1,
|
| 66 |
-
"learning_rate": 9.920000000000002e-06,
|
| 67 |
-
"loss": 0.0751,
|
| 68 |
-
"step": 250
|
| 69 |
-
},
|
| 70 |
-
{
|
| 71 |
-
"epoch": 0.11,
|
| 72 |
-
"learning_rate": 9.89777777777778e-06,
|
| 73 |
-
"loss": 0.0772,
|
| 74 |
-
"step": 275
|
| 75 |
-
},
|
| 76 |
-
{
|
| 77 |
-
"epoch": 0.12,
|
| 78 |
-
"learning_rate": 9.786666666666667e-06,
|
| 79 |
-
"loss": 0.0698,
|
| 80 |
-
"step": 300
|
| 81 |
-
},
|
| 82 |
-
{
|
| 83 |
-
"epoch": 1.0,
|
| 84 |
-
"learning_rate": 9.675555555555555e-06,
|
| 85 |
-
"loss": 0.0893,
|
| 86 |
-
"step": 325
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"epoch": 1.01,
|
| 90 |
-
"learning_rate": 9.564444444444445e-06,
|
| 91 |
-
"loss": 0.0782,
|
| 92 |
-
"step": 350
|
| 93 |
-
},
|
| 94 |
-
{
|
| 95 |
-
"epoch": 1.02,
|
| 96 |
-
"learning_rate": 9.453333333333335e-06,
|
| 97 |
-
"loss": 0.0595,
|
| 98 |
-
"step": 375
|
| 99 |
-
},
|
| 100 |
-
{
|
| 101 |
-
"epoch": 1.03,
|
| 102 |
-
"learning_rate": 9.342222222222223e-06,
|
| 103 |
-
"loss": 0.0537,
|
| 104 |
-
"step": 400
|
| 105 |
-
},
|
| 106 |
-
{
|
| 107 |
-
"epoch": 1.04,
|
| 108 |
-
"learning_rate": 9.231111111111111e-06,
|
| 109 |
-
"loss": 0.052,
|
| 110 |
-
"step": 425
|
| 111 |
-
},
|
| 112 |
-
{
|
| 113 |
-
"epoch": 1.05,
|
| 114 |
-
"learning_rate": 9.12e-06,
|
| 115 |
-
"loss": 0.05,
|
| 116 |
-
"step": 450
|
| 117 |
-
},
|
| 118 |
-
{
|
| 119 |
-
"epoch": 1.06,
|
| 120 |
-
"learning_rate": 9.008888888888889e-06,
|
| 121 |
-
"loss": 0.0522,
|
| 122 |
-
"step": 475
|
| 123 |
-
},
|
| 124 |
-
{
|
| 125 |
-
"epoch": 1.07,
|
| 126 |
-
"learning_rate": 8.897777777777779e-06,
|
| 127 |
-
"loss": 0.0467,
|
| 128 |
-
"step": 500
|
| 129 |
-
},
|
| 130 |
-
{
|
| 131 |
-
"epoch": 1.08,
|
| 132 |
-
"learning_rate": 8.786666666666668e-06,
|
| 133 |
-
"loss": 0.0586,
|
| 134 |
-
"step": 525
|
| 135 |
-
},
|
| 136 |
-
{
|
| 137 |
-
"epoch": 1.09,
|
| 138 |
-
"learning_rate": 8.675555555555556e-06,
|
| 139 |
-
"loss": 0.0437,
|
| 140 |
-
"step": 550
|
| 141 |
-
},
|
| 142 |
-
{
|
| 143 |
-
"epoch": 1.1,
|
| 144 |
-
"learning_rate": 8.564444444444445e-06,
|
| 145 |
-
"loss": 0.0395,
|
| 146 |
-
"step": 575
|
| 147 |
-
},
|
| 148 |
-
{
|
| 149 |
-
"epoch": 1.11,
|
| 150 |
-
"learning_rate": 8.453333333333334e-06,
|
| 151 |
-
"loss": 0.0371,
|
| 152 |
-
"step": 600
|
| 153 |
-
},
|
| 154 |
-
{
|
| 155 |
-
"epoch": 1.12,
|
| 156 |
-
"learning_rate": 8.342222222222222e-06,
|
| 157 |
-
"loss": 0.0316,
|
| 158 |
-
"step": 625
|
| 159 |
-
},
|
| 160 |
-
{
|
| 161 |
-
"epoch": 2.01,
|
| 162 |
-
"learning_rate": 8.231111111111112e-06,
|
| 163 |
-
"loss": 0.0412,
|
| 164 |
-
"step": 650
|
| 165 |
-
},
|
| 166 |
-
{
|
| 167 |
-
"epoch": 2.02,
|
| 168 |
-
"learning_rate": 8.120000000000002e-06,
|
| 169 |
-
"loss": 0.0335,
|
| 170 |
-
"step": 675
|
| 171 |
-
},
|
| 172 |
-
{
|
| 173 |
-
"epoch": 2.03,
|
| 174 |
-
"learning_rate": 8.00888888888889e-06,
|
| 175 |
-
"loss": 0.0314,
|
| 176 |
-
"step": 700
|
| 177 |
-
},
|
| 178 |
-
{
|
| 179 |
-
"epoch": 2.04,
|
| 180 |
-
"learning_rate": 7.897777777777778e-06,
|
| 181 |
-
"loss": 0.0329,
|
| 182 |
-
"step": 725
|
| 183 |
-
},
|
| 184 |
-
{
|
| 185 |
-
"epoch": 2.05,
|
| 186 |
-
"learning_rate": 7.786666666666666e-06,
|
| 187 |
-
"loss": 0.0275,
|
| 188 |
-
"step": 750
|
| 189 |
-
},
|
| 190 |
-
{
|
| 191 |
-
"epoch": 2.06,
|
| 192 |
-
"learning_rate": 7.675555555555556e-06,
|
| 193 |
-
"loss": 0.028,
|
| 194 |
-
"step": 775
|
| 195 |
-
},
|
| 196 |
-
{
|
| 197 |
-
"epoch": 2.07,
|
| 198 |
-
"learning_rate": 7.564444444444446e-06,
|
| 199 |
-
"loss": 0.0256,
|
| 200 |
-
"step": 800
|
| 201 |
-
},
|
| 202 |
-
{
|
| 203 |
-
"epoch": 2.08,
|
| 204 |
-
"learning_rate": 7.453333333333334e-06,
|
| 205 |
-
"loss": 0.026,
|
| 206 |
-
"step": 825
|
| 207 |
-
},
|
| 208 |
-
{
|
| 209 |
-
"epoch": 2.09,
|
| 210 |
-
"learning_rate": 7.342222222222223e-06,
|
| 211 |
-
"loss": 0.0295,
|
| 212 |
-
"step": 850
|
| 213 |
-
},
|
| 214 |
-
{
|
| 215 |
-
"epoch": 2.1,
|
| 216 |
-
"learning_rate": 7.231111111111112e-06,
|
| 217 |
-
"loss": 0.0259,
|
| 218 |
-
"step": 875
|
| 219 |
-
},
|
| 220 |
-
{
|
| 221 |
-
"epoch": 2.11,
|
| 222 |
-
"learning_rate": 7.1200000000000004e-06,
|
| 223 |
-
"loss": 0.0173,
|
| 224 |
-
"step": 900
|
| 225 |
-
},
|
| 226 |
-
{
|
| 227 |
-
"epoch": 2.12,
|
| 228 |
-
"learning_rate": 7.008888888888889e-06,
|
| 229 |
-
"loss": 0.014,
|
| 230 |
-
"step": 925
|
| 231 |
-
},
|
| 232 |
-
{
|
| 233 |
-
"epoch": 3.0,
|
| 234 |
-
"learning_rate": 6.897777777777779e-06,
|
| 235 |
-
"loss": 0.0158,
|
| 236 |
-
"step": 950
|
| 237 |
-
},
|
| 238 |
-
{
|
| 239 |
-
"epoch": 3.01,
|
| 240 |
-
"learning_rate": 6.786666666666667e-06,
|
| 241 |
-
"loss": 0.0193,
|
| 242 |
-
"step": 975
|
| 243 |
-
},
|
| 244 |
-
{
|
| 245 |
-
"epoch": 3.02,
|
| 246 |
-
"learning_rate": 6.675555555555556e-06,
|
| 247 |
-
"loss": 0.0146,
|
| 248 |
-
"step": 1000
|
| 249 |
-
},
|
| 250 |
-
{
|
| 251 |
-
"epoch": 3.02,
|
| 252 |
-
"eval_loss": 0.2546452581882477,
|
| 253 |
-
"eval_runtime": 2882.0567,
|
| 254 |
-
"eval_samples_per_second": 1.759,
|
| 255 |
-
"eval_steps_per_second": 0.11,
|
| 256 |
-
"eval_wer": 11.942264272658853,
|
| 257 |
-
"step": 1000
|
| 258 |
-
}
|
| 259 |
-
],
|
| 260 |
-
"max_steps": 2500,
|
| 261 |
-
"num_train_epochs": 9223372036854775807,
|
| 262 |
-
"total_flos": 3.263486252285952e+19,
|
| 263 |
-
"trial_name": null,
|
| 264 |
-
"trial_params": null
|
| 265 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-1000/training_args.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:6e13de8ddcc61da424ec8e8d050cd43e412a6f52aa9be94507835652edf1dedf
|
| 3 |
-
size 3579
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-2000/config.json
DELETED
|
@@ -1,41 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"_name_or_path": "marinone94/whisper-medium-nordic",
|
| 3 |
-
"activation_dropout": 0.0,
|
| 4 |
-
"activation_function": "gelu",
|
| 5 |
-
"architectures": [
|
| 6 |
-
"WhisperForConditionalGeneration"
|
| 7 |
-
],
|
| 8 |
-
"attention_dropout": 0.0,
|
| 9 |
-
"begin_suppress_tokens": [
|
| 10 |
-
220,
|
| 11 |
-
50257
|
| 12 |
-
],
|
| 13 |
-
"bos_token_id": 50257,
|
| 14 |
-
"d_model": 1024,
|
| 15 |
-
"decoder_attention_heads": 16,
|
| 16 |
-
"decoder_ffn_dim": 4096,
|
| 17 |
-
"decoder_layerdrop": 0.0,
|
| 18 |
-
"decoder_layers": 24,
|
| 19 |
-
"decoder_start_token_id": 50258,
|
| 20 |
-
"dropout": 0.0,
|
| 21 |
-
"encoder_attention_heads": 16,
|
| 22 |
-
"encoder_ffn_dim": 4096,
|
| 23 |
-
"encoder_layerdrop": 0.0,
|
| 24 |
-
"encoder_layers": 24,
|
| 25 |
-
"eos_token_id": 50257,
|
| 26 |
-
"forced_decoder_ids": null,
|
| 27 |
-
"init_std": 0.02,
|
| 28 |
-
"is_encoder_decoder": true,
|
| 29 |
-
"max_length": 448,
|
| 30 |
-
"max_source_positions": 1500,
|
| 31 |
-
"max_target_positions": 448,
|
| 32 |
-
"model_type": "whisper",
|
| 33 |
-
"num_hidden_layers": 24,
|
| 34 |
-
"num_mel_bins": 80,
|
| 35 |
-
"pad_token_id": 50257,
|
| 36 |
-
"scale_embedding": false,
|
| 37 |
-
"torch_dtype": "float32",
|
| 38 |
-
"transformers_version": "4.26.0.dev0",
|
| 39 |
-
"use_cache": false,
|
| 40 |
-
"vocab_size": 51865
|
| 41 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-2000/optimizer.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:259711c4e6978872a098051416766fdc14fdd8fac16daedeb90809ff04b25acd
|
| 3 |
-
size 6111428695
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-2000/preprocessor_config.json
DELETED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-2000/pytorch_model.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:523e5675a34a184a100528b8c77ec2ff6d5c5fddb30ef2d034ca8b7b1cdae7e6
|
| 3 |
-
size 3055754841
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-2000/rng_state.pth
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:a0ac7588e56b2f3605270c18191b85d12253a2aa8ac37011f5dd7ee58036e40c
|
| 3 |
-
size 14575
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-2000/scaler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:93dcbb27405421069b5f31003e3f224646e375ff4c9d7ba4b0d265614f660d0f
|
| 3 |
-
size 557
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-2000/scheduler.pt
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:37492482747b6e8c7f9b2ddf6b8c932454afa931831c235a6f136e9ca901a936
|
| 3 |
-
size 627
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-2000/trainer_state.json
DELETED
|
@@ -1,514 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"best_metric": 11.37780883775938,
|
| 3 |
-
"best_model_checkpoint": "./checkpoint-2000",
|
| 4 |
-
"epoch": 6.0416,
|
| 5 |
-
"global_step": 2000,
|
| 6 |
-
"is_hyper_param_search": false,
|
| 7 |
-
"is_local_process_zero": true,
|
| 8 |
-
"is_world_process_zero": true,
|
| 9 |
-
"log_history": [
|
| 10 |
-
{
|
| 11 |
-
"epoch": 0.01,
|
| 12 |
-
"learning_rate": 9.200000000000001e-07,
|
| 13 |
-
"loss": 0.5204,
|
| 14 |
-
"step": 25
|
| 15 |
-
},
|
| 16 |
-
{
|
| 17 |
-
"epoch": 0.02,
|
| 18 |
-
"learning_rate": 1.9200000000000003e-06,
|
| 19 |
-
"loss": 0.0582,
|
| 20 |
-
"step": 50
|
| 21 |
-
},
|
| 22 |
-
{
|
| 23 |
-
"epoch": 0.03,
|
| 24 |
-
"learning_rate": 2.92e-06,
|
| 25 |
-
"loss": 0.0527,
|
| 26 |
-
"step": 75
|
| 27 |
-
},
|
| 28 |
-
{
|
| 29 |
-
"epoch": 0.04,
|
| 30 |
-
"learning_rate": 3.920000000000001e-06,
|
| 31 |
-
"loss": 0.0587,
|
| 32 |
-
"step": 100
|
| 33 |
-
},
|
| 34 |
-
{
|
| 35 |
-
"epoch": 0.05,
|
| 36 |
-
"learning_rate": 4.92e-06,
|
| 37 |
-
"loss": 0.0659,
|
| 38 |
-
"step": 125
|
| 39 |
-
},
|
| 40 |
-
{
|
| 41 |
-
"epoch": 0.06,
|
| 42 |
-
"learning_rate": 5.92e-06,
|
| 43 |
-
"loss": 0.0681,
|
| 44 |
-
"step": 150
|
| 45 |
-
},
|
| 46 |
-
{
|
| 47 |
-
"epoch": 0.07,
|
| 48 |
-
"learning_rate": 6.92e-06,
|
| 49 |
-
"loss": 0.07,
|
| 50 |
-
"step": 175
|
| 51 |
-
},
|
| 52 |
-
{
|
| 53 |
-
"epoch": 0.08,
|
| 54 |
-
"learning_rate": 7.92e-06,
|
| 55 |
-
"loss": 0.0856,
|
| 56 |
-
"step": 200
|
| 57 |
-
},
|
| 58 |
-
{
|
| 59 |
-
"epoch": 0.09,
|
| 60 |
-
"learning_rate": 8.920000000000001e-06,
|
| 61 |
-
"loss": 0.0894,
|
| 62 |
-
"step": 225
|
| 63 |
-
},
|
| 64 |
-
{
|
| 65 |
-
"epoch": 0.1,
|
| 66 |
-
"learning_rate": 9.920000000000002e-06,
|
| 67 |
-
"loss": 0.0751,
|
| 68 |
-
"step": 250
|
| 69 |
-
},
|
| 70 |
-
{
|
| 71 |
-
"epoch": 0.11,
|
| 72 |
-
"learning_rate": 9.89777777777778e-06,
|
| 73 |
-
"loss": 0.0772,
|
| 74 |
-
"step": 275
|
| 75 |
-
},
|
| 76 |
-
{
|
| 77 |
-
"epoch": 0.12,
|
| 78 |
-
"learning_rate": 9.786666666666667e-06,
|
| 79 |
-
"loss": 0.0698,
|
| 80 |
-
"step": 300
|
| 81 |
-
},
|
| 82 |
-
{
|
| 83 |
-
"epoch": 1.0,
|
| 84 |
-
"learning_rate": 9.675555555555555e-06,
|
| 85 |
-
"loss": 0.0893,
|
| 86 |
-
"step": 325
|
| 87 |
-
},
|
| 88 |
-
{
|
| 89 |
-
"epoch": 1.01,
|
| 90 |
-
"learning_rate": 9.564444444444445e-06,
|
| 91 |
-
"loss": 0.0782,
|
| 92 |
-
"step": 350
|
| 93 |
-
},
|
| 94 |
-
{
|
| 95 |
-
"epoch": 1.02,
|
| 96 |
-
"learning_rate": 9.453333333333335e-06,
|
| 97 |
-
"loss": 0.0595,
|
| 98 |
-
"step": 375
|
| 99 |
-
},
|
| 100 |
-
{
|
| 101 |
-
"epoch": 1.03,
|
| 102 |
-
"learning_rate": 9.342222222222223e-06,
|
| 103 |
-
"loss": 0.0537,
|
| 104 |
-
"step": 400
|
| 105 |
-
},
|
| 106 |
-
{
|
| 107 |
-
"epoch": 1.04,
|
| 108 |
-
"learning_rate": 9.231111111111111e-06,
|
| 109 |
-
"loss": 0.052,
|
| 110 |
-
"step": 425
|
| 111 |
-
},
|
| 112 |
-
{
|
| 113 |
-
"epoch": 1.05,
|
| 114 |
-
"learning_rate": 9.12e-06,
|
| 115 |
-
"loss": 0.05,
|
| 116 |
-
"step": 450
|
| 117 |
-
},
|
| 118 |
-
{
|
| 119 |
-
"epoch": 1.06,
|
| 120 |
-
"learning_rate": 9.008888888888889e-06,
|
| 121 |
-
"loss": 0.0522,
|
| 122 |
-
"step": 475
|
| 123 |
-
},
|
| 124 |
-
{
|
| 125 |
-
"epoch": 1.07,
|
| 126 |
-
"learning_rate": 8.897777777777779e-06,
|
| 127 |
-
"loss": 0.0467,
|
| 128 |
-
"step": 500
|
| 129 |
-
},
|
| 130 |
-
{
|
| 131 |
-
"epoch": 1.08,
|
| 132 |
-
"learning_rate": 8.786666666666668e-06,
|
| 133 |
-
"loss": 0.0586,
|
| 134 |
-
"step": 525
|
| 135 |
-
},
|
| 136 |
-
{
|
| 137 |
-
"epoch": 1.09,
|
| 138 |
-
"learning_rate": 8.675555555555556e-06,
|
| 139 |
-
"loss": 0.0437,
|
| 140 |
-
"step": 550
|
| 141 |
-
},
|
| 142 |
-
{
|
| 143 |
-
"epoch": 1.1,
|
| 144 |
-
"learning_rate": 8.564444444444445e-06,
|
| 145 |
-
"loss": 0.0395,
|
| 146 |
-
"step": 575
|
| 147 |
-
},
|
| 148 |
-
{
|
| 149 |
-
"epoch": 1.11,
|
| 150 |
-
"learning_rate": 8.453333333333334e-06,
|
| 151 |
-
"loss": 0.0371,
|
| 152 |
-
"step": 600
|
| 153 |
-
},
|
| 154 |
-
{
|
| 155 |
-
"epoch": 1.12,
|
| 156 |
-
"learning_rate": 8.342222222222222e-06,
|
| 157 |
-
"loss": 0.0316,
|
| 158 |
-
"step": 625
|
| 159 |
-
},
|
| 160 |
-
{
|
| 161 |
-
"epoch": 2.01,
|
| 162 |
-
"learning_rate": 8.231111111111112e-06,
|
| 163 |
-
"loss": 0.0412,
|
| 164 |
-
"step": 650
|
| 165 |
-
},
|
| 166 |
-
{
|
| 167 |
-
"epoch": 2.02,
|
| 168 |
-
"learning_rate": 8.120000000000002e-06,
|
| 169 |
-
"loss": 0.0335,
|
| 170 |
-
"step": 675
|
| 171 |
-
},
|
| 172 |
-
{
|
| 173 |
-
"epoch": 2.03,
|
| 174 |
-
"learning_rate": 8.00888888888889e-06,
|
| 175 |
-
"loss": 0.0314,
|
| 176 |
-
"step": 700
|
| 177 |
-
},
|
| 178 |
-
{
|
| 179 |
-
"epoch": 2.04,
|
| 180 |
-
"learning_rate": 7.897777777777778e-06,
|
| 181 |
-
"loss": 0.0329,
|
| 182 |
-
"step": 725
|
| 183 |
-
},
|
| 184 |
-
{
|
| 185 |
-
"epoch": 2.05,
|
| 186 |
-
"learning_rate": 7.786666666666666e-06,
|
| 187 |
-
"loss": 0.0275,
|
| 188 |
-
"step": 750
|
| 189 |
-
},
|
| 190 |
-
{
|
| 191 |
-
"epoch": 2.06,
|
| 192 |
-
"learning_rate": 7.675555555555556e-06,
|
| 193 |
-
"loss": 0.028,
|
| 194 |
-
"step": 775
|
| 195 |
-
},
|
| 196 |
-
{
|
| 197 |
-
"epoch": 2.07,
|
| 198 |
-
"learning_rate": 7.564444444444446e-06,
|
| 199 |
-
"loss": 0.0256,
|
| 200 |
-
"step": 800
|
| 201 |
-
},
|
| 202 |
-
{
|
| 203 |
-
"epoch": 2.08,
|
| 204 |
-
"learning_rate": 7.453333333333334e-06,
|
| 205 |
-
"loss": 0.026,
|
| 206 |
-
"step": 825
|
| 207 |
-
},
|
| 208 |
-
{
|
| 209 |
-
"epoch": 2.09,
|
| 210 |
-
"learning_rate": 7.342222222222223e-06,
|
| 211 |
-
"loss": 0.0295,
|
| 212 |
-
"step": 850
|
| 213 |
-
},
|
| 214 |
-
{
|
| 215 |
-
"epoch": 2.1,
|
| 216 |
-
"learning_rate": 7.231111111111112e-06,
|
| 217 |
-
"loss": 0.0259,
|
| 218 |
-
"step": 875
|
| 219 |
-
},
|
| 220 |
-
{
|
| 221 |
-
"epoch": 2.11,
|
| 222 |
-
"learning_rate": 7.1200000000000004e-06,
|
| 223 |
-
"loss": 0.0173,
|
| 224 |
-
"step": 900
|
| 225 |
-
},
|
| 226 |
-
{
|
| 227 |
-
"epoch": 2.12,
|
| 228 |
-
"learning_rate": 7.008888888888889e-06,
|
| 229 |
-
"loss": 0.014,
|
| 230 |
-
"step": 925
|
| 231 |
-
},
|
| 232 |
-
{
|
| 233 |
-
"epoch": 3.0,
|
| 234 |
-
"learning_rate": 6.897777777777779e-06,
|
| 235 |
-
"loss": 0.0158,
|
| 236 |
-
"step": 950
|
| 237 |
-
},
|
| 238 |
-
{
|
| 239 |
-
"epoch": 3.01,
|
| 240 |
-
"learning_rate": 6.786666666666667e-06,
|
| 241 |
-
"loss": 0.0193,
|
| 242 |
-
"step": 975
|
| 243 |
-
},
|
| 244 |
-
{
|
| 245 |
-
"epoch": 3.02,
|
| 246 |
-
"learning_rate": 6.675555555555556e-06,
|
| 247 |
-
"loss": 0.0146,
|
| 248 |
-
"step": 1000
|
| 249 |
-
},
|
| 250 |
-
{
|
| 251 |
-
"epoch": 3.02,
|
| 252 |
-
"eval_loss": 0.2546452581882477,
|
| 253 |
-
"eval_runtime": 2882.0567,
|
| 254 |
-
"eval_samples_per_second": 1.759,
|
| 255 |
-
"eval_steps_per_second": 0.11,
|
| 256 |
-
"eval_wer": 11.942264272658853,
|
| 257 |
-
"step": 1000
|
| 258 |
-
},
|
| 259 |
-
{
|
| 260 |
-
"epoch": 3.03,
|
| 261 |
-
"learning_rate": 6.564444444444446e-06,
|
| 262 |
-
"loss": 0.0129,
|
| 263 |
-
"step": 1025
|
| 264 |
-
},
|
| 265 |
-
{
|
| 266 |
-
"epoch": 3.04,
|
| 267 |
-
"learning_rate": 6.453333333333334e-06,
|
| 268 |
-
"loss": 0.0111,
|
| 269 |
-
"step": 1050
|
| 270 |
-
},
|
| 271 |
-
{
|
| 272 |
-
"epoch": 3.05,
|
| 273 |
-
"learning_rate": 6.342222222222223e-06,
|
| 274 |
-
"loss": 0.0143,
|
| 275 |
-
"step": 1075
|
| 276 |
-
},
|
| 277 |
-
{
|
| 278 |
-
"epoch": 3.06,
|
| 279 |
-
"learning_rate": 6.231111111111111e-06,
|
| 280 |
-
"loss": 0.0128,
|
| 281 |
-
"step": 1100
|
| 282 |
-
},
|
| 283 |
-
{
|
| 284 |
-
"epoch": 3.07,
|
| 285 |
-
"learning_rate": 6.120000000000001e-06,
|
| 286 |
-
"loss": 0.0105,
|
| 287 |
-
"step": 1125
|
| 288 |
-
},
|
| 289 |
-
{
|
| 290 |
-
"epoch": 3.08,
|
| 291 |
-
"learning_rate": 6.00888888888889e-06,
|
| 292 |
-
"loss": 0.012,
|
| 293 |
-
"step": 1150
|
| 294 |
-
},
|
| 295 |
-
{
|
| 296 |
-
"epoch": 3.09,
|
| 297 |
-
"learning_rate": 5.897777777777778e-06,
|
| 298 |
-
"loss": 0.0133,
|
| 299 |
-
"step": 1175
|
| 300 |
-
},
|
| 301 |
-
{
|
| 302 |
-
"epoch": 3.1,
|
| 303 |
-
"learning_rate": 5.7866666666666674e-06,
|
| 304 |
-
"loss": 0.0093,
|
| 305 |
-
"step": 1200
|
| 306 |
-
},
|
| 307 |
-
{
|
| 308 |
-
"epoch": 3.11,
|
| 309 |
-
"learning_rate": 5.675555555555556e-06,
|
| 310 |
-
"loss": 0.0097,
|
| 311 |
-
"step": 1225
|
| 312 |
-
},
|
| 313 |
-
{
|
| 314 |
-
"epoch": 3.12,
|
| 315 |
-
"learning_rate": 5.5644444444444444e-06,
|
| 316 |
-
"loss": 0.007,
|
| 317 |
-
"step": 1250
|
| 318 |
-
},
|
| 319 |
-
{
|
| 320 |
-
"epoch": 4.0,
|
| 321 |
-
"learning_rate": 5.453333333333334e-06,
|
| 322 |
-
"loss": 0.0059,
|
| 323 |
-
"step": 1275
|
| 324 |
-
},
|
| 325 |
-
{
|
| 326 |
-
"epoch": 4.01,
|
| 327 |
-
"learning_rate": 5.342222222222223e-06,
|
| 328 |
-
"loss": 0.01,
|
| 329 |
-
"step": 1300
|
| 330 |
-
},
|
| 331 |
-
{
|
| 332 |
-
"epoch": 4.02,
|
| 333 |
-
"learning_rate": 5.231111111111111e-06,
|
| 334 |
-
"loss": 0.0086,
|
| 335 |
-
"step": 1325
|
| 336 |
-
},
|
| 337 |
-
{
|
| 338 |
-
"epoch": 4.03,
|
| 339 |
-
"learning_rate": 5.12e-06,
|
| 340 |
-
"loss": 0.0048,
|
| 341 |
-
"step": 1350
|
| 342 |
-
},
|
| 343 |
-
{
|
| 344 |
-
"epoch": 4.04,
|
| 345 |
-
"learning_rate": 5.00888888888889e-06,
|
| 346 |
-
"loss": 0.0058,
|
| 347 |
-
"step": 1375
|
| 348 |
-
},
|
| 349 |
-
{
|
| 350 |
-
"epoch": 4.05,
|
| 351 |
-
"learning_rate": 4.897777777777778e-06,
|
| 352 |
-
"loss": 0.0056,
|
| 353 |
-
"step": 1400
|
| 354 |
-
},
|
| 355 |
-
{
|
| 356 |
-
"epoch": 4.06,
|
| 357 |
-
"learning_rate": 4.786666666666667e-06,
|
| 358 |
-
"loss": 0.0059,
|
| 359 |
-
"step": 1425
|
| 360 |
-
},
|
| 361 |
-
{
|
| 362 |
-
"epoch": 4.07,
|
| 363 |
-
"learning_rate": 4.675555555555556e-06,
|
| 364 |
-
"loss": 0.0038,
|
| 365 |
-
"step": 1450
|
| 366 |
-
},
|
| 367 |
-
{
|
| 368 |
-
"epoch": 4.08,
|
| 369 |
-
"learning_rate": 4.564444444444445e-06,
|
| 370 |
-
"loss": 0.0049,
|
| 371 |
-
"step": 1475
|
| 372 |
-
},
|
| 373 |
-
{
|
| 374 |
-
"epoch": 4.09,
|
| 375 |
-
"learning_rate": 4.453333333333334e-06,
|
| 376 |
-
"loss": 0.0051,
|
| 377 |
-
"step": 1500
|
| 378 |
-
},
|
| 379 |
-
{
|
| 380 |
-
"epoch": 4.1,
|
| 381 |
-
"learning_rate": 4.3422222222222225e-06,
|
| 382 |
-
"loss": 0.0044,
|
| 383 |
-
"step": 1525
|
| 384 |
-
},
|
| 385 |
-
{
|
| 386 |
-
"epoch": 4.11,
|
| 387 |
-
"learning_rate": 4.2311111111111114e-06,
|
| 388 |
-
"loss": 0.0026,
|
| 389 |
-
"step": 1550
|
| 390 |
-
},
|
| 391 |
-
{
|
| 392 |
-
"epoch": 4.12,
|
| 393 |
-
"learning_rate": 4.12e-06,
|
| 394 |
-
"loss": 0.0036,
|
| 395 |
-
"step": 1575
|
| 396 |
-
},
|
| 397 |
-
{
|
| 398 |
-
"epoch": 5.01,
|
| 399 |
-
"learning_rate": 4.008888888888889e-06,
|
| 400 |
-
"loss": 0.0035,
|
| 401 |
-
"step": 1600
|
| 402 |
-
},
|
| 403 |
-
{
|
| 404 |
-
"epoch": 5.02,
|
| 405 |
-
"learning_rate": 3.897777777777778e-06,
|
| 406 |
-
"loss": 0.0024,
|
| 407 |
-
"step": 1625
|
| 408 |
-
},
|
| 409 |
-
{
|
| 410 |
-
"epoch": 5.03,
|
| 411 |
-
"learning_rate": 3.7866666666666667e-06,
|
| 412 |
-
"loss": 0.0026,
|
| 413 |
-
"step": 1650
|
| 414 |
-
},
|
| 415 |
-
{
|
| 416 |
-
"epoch": 5.04,
|
| 417 |
-
"learning_rate": 3.675555555555556e-06,
|
| 418 |
-
"loss": 0.0031,
|
| 419 |
-
"step": 1675
|
| 420 |
-
},
|
| 421 |
-
{
|
| 422 |
-
"epoch": 5.05,
|
| 423 |
-
"learning_rate": 3.564444444444445e-06,
|
| 424 |
-
"loss": 0.0045,
|
| 425 |
-
"step": 1700
|
| 426 |
-
},
|
| 427 |
-
{
|
| 428 |
-
"epoch": 5.06,
|
| 429 |
-
"learning_rate": 3.4533333333333334e-06,
|
| 430 |
-
"loss": 0.0019,
|
| 431 |
-
"step": 1725
|
| 432 |
-
},
|
| 433 |
-
{
|
| 434 |
-
"epoch": 5.07,
|
| 435 |
-
"learning_rate": 3.3422222222222224e-06,
|
| 436 |
-
"loss": 0.0018,
|
| 437 |
-
"step": 1750
|
| 438 |
-
},
|
| 439 |
-
{
|
| 440 |
-
"epoch": 5.08,
|
| 441 |
-
"learning_rate": 3.2311111111111117e-06,
|
| 442 |
-
"loss": 0.0033,
|
| 443 |
-
"step": 1775
|
| 444 |
-
},
|
| 445 |
-
{
|
| 446 |
-
"epoch": 5.09,
|
| 447 |
-
"learning_rate": 3.12e-06,
|
| 448 |
-
"loss": 0.0048,
|
| 449 |
-
"step": 1800
|
| 450 |
-
},
|
| 451 |
-
{
|
| 452 |
-
"epoch": 5.1,
|
| 453 |
-
"learning_rate": 3.008888888888889e-06,
|
| 454 |
-
"loss": 0.0019,
|
| 455 |
-
"step": 1825
|
| 456 |
-
},
|
| 457 |
-
{
|
| 458 |
-
"epoch": 5.11,
|
| 459 |
-
"learning_rate": 2.8977777777777785e-06,
|
| 460 |
-
"loss": 0.0015,
|
| 461 |
-
"step": 1850
|
| 462 |
-
},
|
| 463 |
-
{
|
| 464 |
-
"epoch": 5.12,
|
| 465 |
-
"learning_rate": 2.786666666666667e-06,
|
| 466 |
-
"loss": 0.0018,
|
| 467 |
-
"step": 1875
|
| 468 |
-
},
|
| 469 |
-
{
|
| 470 |
-
"epoch": 6.0,
|
| 471 |
-
"learning_rate": 2.675555555555556e-06,
|
| 472 |
-
"loss": 0.0017,
|
| 473 |
-
"step": 1900
|
| 474 |
-
},
|
| 475 |
-
{
|
| 476 |
-
"epoch": 6.01,
|
| 477 |
-
"learning_rate": 2.5644444444444444e-06,
|
| 478 |
-
"loss": 0.0017,
|
| 479 |
-
"step": 1925
|
| 480 |
-
},
|
| 481 |
-
{
|
| 482 |
-
"epoch": 6.02,
|
| 483 |
-
"learning_rate": 2.4533333333333333e-06,
|
| 484 |
-
"loss": 0.0009,
|
| 485 |
-
"step": 1950
|
| 486 |
-
},
|
| 487 |
-
{
|
| 488 |
-
"epoch": 6.03,
|
| 489 |
-
"learning_rate": 2.342222222222222e-06,
|
| 490 |
-
"loss": 0.0011,
|
| 491 |
-
"step": 1975
|
| 492 |
-
},
|
| 493 |
-
{
|
| 494 |
-
"epoch": 6.04,
|
| 495 |
-
"learning_rate": 2.2311111111111115e-06,
|
| 496 |
-
"loss": 0.0017,
|
| 497 |
-
"step": 2000
|
| 498 |
-
},
|
| 499 |
-
{
|
| 500 |
-
"epoch": 6.04,
|
| 501 |
-
"eval_loss": 0.29696929454803467,
|
| 502 |
-
"eval_runtime": 2899.6818,
|
| 503 |
-
"eval_samples_per_second": 1.748,
|
| 504 |
-
"eval_steps_per_second": 0.109,
|
| 505 |
-
"eval_wer": 11.37780883775938,
|
| 506 |
-
"step": 2000
|
| 507 |
-
}
|
| 508 |
-
],
|
| 509 |
-
"max_steps": 2500,
|
| 510 |
-
"num_train_epochs": 9223372036854775807,
|
| 511 |
-
"total_flos": 6.526972504571904e+19,
|
| 512 |
-
"trial_name": null,
|
| 513 |
-
"trial_params": null
|
| 514 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
checkpoint-2000/training_args.bin
DELETED
|
@@ -1,3 +0,0 @@
|
|
| 1 |
-
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:6e13de8ddcc61da424ec8e8d050cd43e412a6f52aa9be94507835652edf1dedf
|
| 3 |
-
size 3579
|
|
|
|
|
|
|
|
|
|
|
|
eval_results.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"eval_loss": 0.
|
| 4 |
-
"eval_runtime":
|
| 5 |
-
"eval_samples_per_second": 1.
|
| 6 |
-
"eval_steps_per_second": 0.
|
| 7 |
-
"eval_wer":
|
| 8 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"eval_loss": 0.24834245443344116,
|
| 4 |
+
"eval_runtime": 2999.4256,
|
| 5 |
+
"eval_samples_per_second": 1.69,
|
| 6 |
+
"eval_steps_per_second": 0.106,
|
| 7 |
+
"eval_wer": 9.891409525857435
|
| 8 |
}
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3055754841
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce7f783491edc671deb55f5ce6701ebc753df5b565204547bdda4e4c24df4477
|
| 3 |
size 3055754841
|
run.sh
CHANGED
|
@@ -1,12 +1,14 @@
|
|
| 1 |
python run_speech_recognition_seq2seq_streaming.py \
|
| 2 |
--model_name_or_path="marinone94/whisper-medium-nordic" \
|
| 3 |
-
--
|
| 4 |
-
--
|
| 5 |
--language="swedish" \
|
| 6 |
-
--train_split_name="train+validation" \
|
|
|
|
|
|
|
| 7 |
--eval_split_name="test" \
|
| 8 |
--model_index_name="Whisper Medium Swedish" \
|
| 9 |
-
--max_steps="
|
| 10 |
--output_dir="./" \
|
| 11 |
--per_device_train_batch_size="32" \
|
| 12 |
--per_device_eval_batch_size="16" \
|
|
@@ -20,19 +22,15 @@ python run_speech_recognition_seq2seq_streaming.py \
|
|
| 20 |
--generation_max_length="225" \
|
| 21 |
--length_column_name="input_length" \
|
| 22 |
--max_duration_in_seconds="30" \
|
| 23 |
-
--text_column_name="sentence" \
|
| 24 |
--freeze_feature_encoder="False" \
|
| 25 |
-
--report_to="
|
| 26 |
--metric_for_best_model="wer" \
|
| 27 |
--greater_is_better="False" \
|
| 28 |
--load_best_model_at_end \
|
| 29 |
--gradient_checkpointing \
|
| 30 |
--fp16 \
|
| 31 |
-
--overwrite_output_dir \
|
| 32 |
-
--do_train \
|
| 33 |
-
--do_eval \
|
| 34 |
--predict_with_generate \
|
| 35 |
--do_normalize_eval \
|
| 36 |
--streaming \
|
| 37 |
-
--use_auth_token
|
| 38 |
-
--push_to_hub
|
|
|
|
| 1 |
python run_speech_recognition_seq2seq_streaming.py \
|
| 2 |
--model_name_or_path="marinone94/whisper-medium-nordic" \
|
| 3 |
+
--dataset_train_name="mozilla-foundation/common_voice_11_0,babelbox/babelbox_voice,google/fleurs" \
|
| 4 |
+
--dataset_train_config_name="sv-SE,nst,sv_se" \
|
| 5 |
--language="swedish" \
|
| 6 |
+
--train_split_name="train+validation,train,train+validation+test" \
|
| 7 |
+
--dataset_eval_name="mozilla-foundation/common_voice_11_0" \
|
| 8 |
+
--dataset_eval_config_name="sv-SE" \
|
| 9 |
--eval_split_name="test" \
|
| 10 |
--model_index_name="Whisper Medium Swedish" \
|
| 11 |
+
--max_steps="5000" \
|
| 12 |
--output_dir="./" \
|
| 13 |
--per_device_train_batch_size="32" \
|
| 14 |
--per_device_eval_batch_size="16" \
|
|
|
|
| 22 |
--generation_max_length="225" \
|
| 23 |
--length_column_name="input_length" \
|
| 24 |
--max_duration_in_seconds="30" \
|
| 25 |
+
--text_column_name="sentence,raw_transcription" \
|
| 26 |
--freeze_feature_encoder="False" \
|
| 27 |
+
--report_to="wandb" \
|
| 28 |
--metric_for_best_model="wer" \
|
| 29 |
--greater_is_better="False" \
|
| 30 |
--load_best_model_at_end \
|
| 31 |
--gradient_checkpointing \
|
| 32 |
--fp16 \
|
|
|
|
|
|
|
|
|
|
| 33 |
--predict_with_generate \
|
| 34 |
--do_normalize_eval \
|
| 35 |
--streaming \
|
| 36 |
+
--use_auth_token
|
|
|
run_speech_recognition_seq2seq_streaming.py
CHANGED
|
@@ -20,6 +20,7 @@ with 🤗 Datasets' streaming mode.
|
|
| 20 |
# You can also adapt this script for your own sequence to sequence speech
|
| 21 |
# recognition task. Pointers for this are left as comments.
|
| 22 |
|
|
|
|
| 23 |
import logging
|
| 24 |
import os
|
| 25 |
import sys
|
|
@@ -28,6 +29,7 @@ from typing import Any, Dict, List, Optional, Union
|
|
| 28 |
|
| 29 |
import datasets
|
| 30 |
import torch
|
|
|
|
| 31 |
from datasets import DatasetDict, IterableDatasetDict, interleave_datasets, load_dataset
|
| 32 |
from torch.utils.data import IterableDataset
|
| 33 |
|
|
@@ -60,6 +62,42 @@ require_version("datasets>=1.18.2", "To fix: pip install -r examples/pytorch/spe
|
|
| 60 |
logger = logging.getLogger(__name__)
|
| 61 |
|
| 62 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
@dataclass
|
| 64 |
class ModelArguments:
|
| 65 |
"""
|
|
@@ -127,10 +165,16 @@ class DataTrainingArguments:
|
|
| 127 |
Arguments pertaining to what data we are going to input our model for training and eval.
|
| 128 |
"""
|
| 129 |
|
| 130 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 131 |
default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."}
|
| 132 |
)
|
| 133 |
-
|
| 134 |
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
|
| 135 |
)
|
| 136 |
text_column: Optional[str] = field(
|
|
@@ -265,27 +309,131 @@ class DataCollatorSpeechSeq2SeqWithPadding:
|
|
| 265 |
return batch
|
| 266 |
|
| 267 |
|
| 268 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 269 |
"""
|
| 270 |
Utility function to load a dataset in streaming mode. For datasets with multiple splits,
|
| 271 |
each split is loaded individually and then splits combined by taking alternating examples from
|
| 272 |
each (interleaving).
|
| 273 |
"""
|
| 274 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 275 |
# load multiple splits separated by the `+` symbol with streaming mode
|
| 276 |
-
dataset_splits = [
|
| 277 |
-
|
| 278 |
-
|
| 279 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 280 |
# interleave multiple splits to form one dataset
|
| 281 |
-
interleaved_dataset = interleave_datasets(dataset_splits)
|
| 282 |
return interleaved_dataset
|
| 283 |
else:
|
| 284 |
# load a single split *with* streaming mode
|
| 285 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 286 |
return dataset
|
| 287 |
|
| 288 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 289 |
def main():
|
| 290 |
# 1. Parse input arguments
|
| 291 |
# See all possible arguments in src/transformers/training_args.py
|
|
@@ -349,39 +497,55 @@ def main():
|
|
| 349 |
# Set seed before initializing model.
|
| 350 |
set_seed(training_args.seed)
|
| 351 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 352 |
# 4. Load dataset
|
| 353 |
raw_datasets = IterableDatasetDict() if data_args.streaming else DatasetDict()
|
| 354 |
|
| 355 |
if training_args.do_train:
|
| 356 |
raw_datasets["train"] = load_maybe_streaming_dataset(
|
| 357 |
-
data_args.
|
| 358 |
-
data_args.
|
| 359 |
split=data_args.train_split_name,
|
| 360 |
-
use_auth_token=
|
| 361 |
streaming=data_args.streaming,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 362 |
)
|
| 363 |
|
| 364 |
if training_args.do_eval:
|
| 365 |
raw_datasets["eval"] = load_maybe_streaming_dataset(
|
| 366 |
-
data_args.
|
| 367 |
-
data_args.
|
| 368 |
split=data_args.eval_split_name,
|
| 369 |
-
use_auth_token=
|
| 370 |
streaming=data_args.streaming,
|
|
|
|
|
|
|
|
|
|
|
|
|
| 371 |
)
|
| 372 |
|
| 373 |
raw_datasets_features = list(next(iter(raw_datasets.values())).features.keys())
|
| 374 |
-
|
| 375 |
if data_args.audio_column_name not in raw_datasets_features:
|
| 376 |
raise ValueError(
|
| 377 |
-
f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.
|
| 378 |
"Make sure to set `--audio_column_name` to the correct audio column - one of "
|
| 379 |
f"{', '.join(raw_datasets_features)}."
|
| 380 |
)
|
| 381 |
|
| 382 |
-
if
|
| 383 |
raise ValueError(
|
| 384 |
-
f"--text_column_name {
|
| 385 |
"Make sure to set `--text_column_name` to the correct text column - one of "
|
| 386 |
f"{', '.join(raw_datasets_features)}."
|
| 387 |
)
|
|
@@ -394,7 +558,7 @@ def main():
|
|
| 394 |
model_args.config_name if model_args.config_name else model_args.model_name_or_path,
|
| 395 |
cache_dir=model_args.cache_dir,
|
| 396 |
revision=model_args.model_revision,
|
| 397 |
-
use_auth_token=
|
| 398 |
)
|
| 399 |
|
| 400 |
config.update({"forced_decoder_ids": model_args.forced_decoder_ids, "suppress_tokens": model_args.suppress_tokens})
|
|
@@ -402,25 +566,19 @@ def main():
|
|
| 402 |
if training_args.gradient_checkpointing:
|
| 403 |
config.update({"use_cache": False})
|
| 404 |
|
| 405 |
-
feature_extractor = AutoFeatureExtractor.from_pretrained(
|
| 406 |
-
model_args.feature_extractor_name if model_args.feature_extractor_name else model_args.model_name_or_path,
|
| 407 |
-
cache_dir=model_args.cache_dir,
|
| 408 |
-
revision=model_args.model_revision,
|
| 409 |
-
use_auth_token=True if model_args.use_auth_token else None,
|
| 410 |
-
)
|
| 411 |
tokenizer = AutoTokenizer.from_pretrained(
|
| 412 |
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path,
|
| 413 |
cache_dir=model_args.cache_dir,
|
| 414 |
use_fast=model_args.use_fast_tokenizer,
|
| 415 |
revision=model_args.model_revision,
|
| 416 |
-
use_auth_token=
|
| 417 |
)
|
| 418 |
model = AutoModelForSpeechSeq2Seq.from_pretrained(
|
| 419 |
model_args.model_name_or_path,
|
| 420 |
config=config,
|
| 421 |
cache_dir=model_args.cache_dir,
|
| 422 |
revision=model_args.model_revision,
|
| 423 |
-
use_auth_token=
|
| 424 |
)
|
| 425 |
|
| 426 |
if model.config.decoder_start_token_id is None:
|
|
@@ -448,7 +606,6 @@ def main():
|
|
| 448 |
max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
|
| 449 |
min_input_length = data_args.min_duration_in_seconds * feature_extractor.sampling_rate
|
| 450 |
audio_column_name = data_args.audio_column_name
|
| 451 |
-
text_column_name = data_args.text_column_name
|
| 452 |
model_input_name = feature_extractor.model_input_names[0]
|
| 453 |
do_lower_case = data_args.do_lower_case
|
| 454 |
do_remove_punctuation = data_args.do_remove_punctuation
|
|
@@ -568,6 +725,9 @@ def main():
|
|
| 568 |
callbacks=[ShuffleCallback()] if data_args.streaming else None,
|
| 569 |
)
|
| 570 |
|
|
|
|
|
|
|
|
|
|
| 571 |
# 12. Training
|
| 572 |
if training_args.do_train:
|
| 573 |
checkpoint = None
|
|
@@ -606,24 +766,43 @@ def main():
|
|
| 606 |
"tasks": "automatic-speech-recognition",
|
| 607 |
"tags": "whisper-event",
|
| 608 |
}
|
| 609 |
-
if data_args.
|
| 610 |
-
kwargs["dataset_tags"] = data_args.
|
| 611 |
-
if data_args.
|
| 612 |
-
kwargs["dataset"] = f"{data_args.
|
| 613 |
else:
|
| 614 |
-
kwargs["dataset"] = data_args.
|
| 615 |
-
if "common_voice" in data_args.
|
| 616 |
-
kwargs["language"] = data_args.
|
| 617 |
if model_args.model_index_name is not None:
|
| 618 |
kwargs["model_name"] = model_args.model_index_name
|
| 619 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 620 |
if training_args.push_to_hub:
|
| 621 |
trainer.push_to_hub(**kwargs)
|
| 622 |
else:
|
| 623 |
trainer.create_model_card(**kwargs)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 624 |
|
| 625 |
return results
|
| 626 |
|
| 627 |
|
| 628 |
if __name__ == "__main__":
|
| 629 |
-
main()
|
|
|
|
| 20 |
# You can also adapt this script for your own sequence to sequence speech
|
| 21 |
# recognition task. Pointers for this are left as comments.
|
| 22 |
|
| 23 |
+
import json
|
| 24 |
import logging
|
| 25 |
import os
|
| 26 |
import sys
|
|
|
|
| 29 |
|
| 30 |
import datasets
|
| 31 |
import torch
|
| 32 |
+
import wandb
|
| 33 |
from datasets import DatasetDict, IterableDatasetDict, interleave_datasets, load_dataset
|
| 34 |
from torch.utils.data import IterableDataset
|
| 35 |
|
|
|
|
| 62 |
logger = logging.getLogger(__name__)
|
| 63 |
|
| 64 |
|
| 65 |
+
SENDING_NOTIFICATION = "*** Sending notification to email ***"
|
| 66 |
+
RECIPIENT_ADDRESS = "marinone94@gmail.com"
|
| 67 |
+
|
| 68 |
+
wandb_token = os.environ.get("WANDB_TOKEN", "None")
|
| 69 |
+
hf_token = os.environ.get("HF_TOKEN", None)
|
| 70 |
+
if (hf_token is None or wandb_token == "None") and os.path.exists("./creds.txt"):
|
| 71 |
+
with open("./creds.txt", "r") as f:
|
| 72 |
+
lines = f.readlines()
|
| 73 |
+
for line in lines:
|
| 74 |
+
key, value = line.split("=")
|
| 75 |
+
if key == "HF_TOKEN":
|
| 76 |
+
hf_token = value.strip()
|
| 77 |
+
if key == "WANDB_TOKEN":
|
| 78 |
+
wandb_token = value.strip()
|
| 79 |
+
if key == "EMAIL_ADDRESS":
|
| 80 |
+
os.environ["EMAIL_ADDRESS"] = value.strip()
|
| 81 |
+
if key == "EMAIL_PASSWORD":
|
| 82 |
+
os.environ["EMAIL_PASSWORD"] = value.strip()
|
| 83 |
+
|
| 84 |
+
if hf_token is not None:
|
| 85 |
+
try:
|
| 86 |
+
os.makedirs("/root/.huggingface", exist_ok=True)
|
| 87 |
+
with open("/root/.huggingface/token", "w") as f:
|
| 88 |
+
f.write(hf_token)
|
| 89 |
+
logger.info("Huggingface API key set")
|
| 90 |
+
except (PermissionError, OSError):
|
| 91 |
+
logger.warning("Huggingface API key not set, relying on ~/.huggingface/token")
|
| 92 |
+
else:
|
| 93 |
+
logger.warning("Huggingface API key not set, relying on ~/.huggingface/token")
|
| 94 |
+
|
| 95 |
+
wandb.login(key=wandb_token, relogin=True, timeout=5)
|
| 96 |
+
wandb.init(project="whisper", entity="pn-aa")
|
| 97 |
+
|
| 98 |
+
logger.info("Wandb API key set, logging to wandb")
|
| 99 |
+
|
| 100 |
+
|
| 101 |
@dataclass
|
| 102 |
class ModelArguments:
|
| 103 |
"""
|
|
|
|
| 165 |
Arguments pertaining to what data we are going to input our model for training and eval.
|
| 166 |
"""
|
| 167 |
|
| 168 |
+
dataset_train_name: str = field(
|
| 169 |
+
default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."}
|
| 170 |
+
)
|
| 171 |
+
dataset_train_config_name: Optional[str] = field(
|
| 172 |
+
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
|
| 173 |
+
)
|
| 174 |
+
dataset_eval_name: str = field(
|
| 175 |
default=None, metadata={"help": "The name of the dataset to use (via the datasets library)."}
|
| 176 |
)
|
| 177 |
+
dataset_eval_config_name: Optional[str] = field(
|
| 178 |
default=None, metadata={"help": "The configuration name of the dataset to use (via the datasets library)."}
|
| 179 |
)
|
| 180 |
text_column: Optional[str] = field(
|
|
|
|
| 309 |
return batch
|
| 310 |
|
| 311 |
|
| 312 |
+
def rename_col_and_resample(dataset, dataset_name, text_column_names, text_col_name_ref, audio_column_name, sampling_rate):
|
| 313 |
+
raw_datasets_features = list(dataset.features.keys())
|
| 314 |
+
logger.info(f"Dataset {dataset_name} - Features: {raw_datasets_features}")
|
| 315 |
+
|
| 316 |
+
if text_col_name_ref not in raw_datasets_features:
|
| 317 |
+
if len(text_column_names) == 1:
|
| 318 |
+
raise ValueError("None of the text column names provided found in dataset."
|
| 319 |
+
f"Text columns: {text_column_names}"
|
| 320 |
+
f"Dataset columns: {raw_datasets_features}")
|
| 321 |
+
flag = False
|
| 322 |
+
for text_column_name in text_column_names:
|
| 323 |
+
if text_column_name in raw_datasets_features:
|
| 324 |
+
logger.info(f"Renaming text column {text_column_name} to {text_col_name_ref}")
|
| 325 |
+
dataset = dataset.rename_column(text_column_name, text_col_name_ref)
|
| 326 |
+
flag = True
|
| 327 |
+
break
|
| 328 |
+
if flag is False:
|
| 329 |
+
raise ValueError("None of the text column names provided found in dataset."
|
| 330 |
+
f"Text columns: {text_column_names}"
|
| 331 |
+
f"Dataset columns: {raw_datasets_features}")
|
| 332 |
+
if audio_column_name is not None and sampling_rate is not None:
|
| 333 |
+
ds_sr = int(dataset.features[audio_column_name].sampling_rate)
|
| 334 |
+
if ds_sr != sampling_rate:
|
| 335 |
+
dataset = dataset.cast_column(
|
| 336 |
+
audio_column_name, datasets.features.Audio(sampling_rate=sampling_rate)
|
| 337 |
+
)
|
| 338 |
+
|
| 339 |
+
raw_datasets_features = list(dataset.features.keys())
|
| 340 |
+
raw_datasets_features.remove(audio_column_name)
|
| 341 |
+
raw_datasets_features.remove(text_col_name_ref)
|
| 342 |
+
# Keep only audio and sentence
|
| 343 |
+
dataset = dataset.remove_columns(column_names=raw_datasets_features)
|
| 344 |
+
return dataset
|
| 345 |
+
|
| 346 |
+
|
| 347 |
+
def load_maybe_streaming_dataset(
|
| 348 |
+
dataset_names,
|
| 349 |
+
dataset_config_names,
|
| 350 |
+
split="train",
|
| 351 |
+
streaming=True,
|
| 352 |
+
audio_column_name=None,
|
| 353 |
+
sampling_rate=None,
|
| 354 |
+
**kwargs
|
| 355 |
+
):
|
| 356 |
"""
|
| 357 |
Utility function to load a dataset in streaming mode. For datasets with multiple splits,
|
| 358 |
each split is loaded individually and then splits combined by taking alternating examples from
|
| 359 |
each (interleaving).
|
| 360 |
"""
|
| 361 |
+
text_column_names = None
|
| 362 |
+
if "text_column_name" in kwargs:
|
| 363 |
+
text_column_names = kwargs.pop("text_column_name").split(",")
|
| 364 |
+
text_col_name_ref = text_column_names[0]
|
| 365 |
+
|
| 366 |
+
if "," in dataset_names or "+" in split:
|
| 367 |
# load multiple splits separated by the `+` symbol with streaming mode
|
| 368 |
+
dataset_splits = []
|
| 369 |
+
for dataset_name, dataset_config_name, split_names in zip(
|
| 370 |
+
dataset_names.split(","), dataset_config_names.split(","), split.split(",")
|
| 371 |
+
):
|
| 372 |
+
for split_name in split_names.split("+"):
|
| 373 |
+
if dataset_config_name:
|
| 374 |
+
dataset = load_dataset(dataset_name, dataset_config_name, split=split_name, streaming=streaming, **kwargs)
|
| 375 |
+
else:
|
| 376 |
+
dataset = load_dataset(dataset_name, split=split_name, streaming=streaming, **kwargs)
|
| 377 |
+
|
| 378 |
+
dataset = rename_col_and_resample(
|
| 379 |
+
dataset,
|
| 380 |
+
dataset_name,
|
| 381 |
+
text_column_names,
|
| 382 |
+
text_col_name_ref,
|
| 383 |
+
audio_column_name,
|
| 384 |
+
sampling_rate
|
| 385 |
+
)
|
| 386 |
+
|
| 387 |
+
dataset_splits.append(dataset)
|
| 388 |
+
|
| 389 |
# interleave multiple splits to form one dataset
|
| 390 |
+
interleaved_dataset = interleave_datasets(dataset_splits, stopping_strategy="all_exhausted")
|
| 391 |
return interleaved_dataset
|
| 392 |
else:
|
| 393 |
# load a single split *with* streaming mode
|
| 394 |
+
|
| 395 |
+
dataset = load_dataset(dataset_names, dataset_config_names, split=split, streaming=streaming, **kwargs)
|
| 396 |
+
dataset = rename_col_and_resample(
|
| 397 |
+
dataset,
|
| 398 |
+
dataset_names,
|
| 399 |
+
text_column_names,
|
| 400 |
+
text_col_name_ref,
|
| 401 |
+
audio_column_name,
|
| 402 |
+
sampling_rate
|
| 403 |
+
)
|
| 404 |
return dataset
|
| 405 |
|
| 406 |
|
| 407 |
+
def notify_me(recipient, message=None):
|
| 408 |
+
"""
|
| 409 |
+
Send an email to the specified address with the specified message
|
| 410 |
+
"""
|
| 411 |
+
sender = os.environ.get("EMAIL_ADDRESS", None)
|
| 412 |
+
password = os.environ.get("EMAIL_PASSWORD", None)
|
| 413 |
+
if sender is None:
|
| 414 |
+
logging.warning("No email address specified, not sending notification")
|
| 415 |
+
if password is None:
|
| 416 |
+
logging.warning("No email password specified, not sending notification")
|
| 417 |
+
if message is None:
|
| 418 |
+
message = "Training is finished!"
|
| 419 |
+
|
| 420 |
+
if sender is not None:
|
| 421 |
+
import smtplib
|
| 422 |
+
from email.mime.text import MIMEText
|
| 423 |
+
|
| 424 |
+
msg = MIMEText(message)
|
| 425 |
+
msg["Subject"] = "Training updates..."
|
| 426 |
+
msg["From"] = "marinone.auto@gmail.com"
|
| 427 |
+
msg["To"] = recipient
|
| 428 |
+
|
| 429 |
+
# send the email
|
| 430 |
+
smtp_obj = smtplib.SMTP("smtp.gmail.com", 587)
|
| 431 |
+
smtp_obj.starttls()
|
| 432 |
+
smtp_obj.login(sender, password)
|
| 433 |
+
smtp_obj.sendmail(sender, recipient, msg.as_string())
|
| 434 |
+
smtp_obj.quit()
|
| 435 |
+
|
| 436 |
+
|
| 437 |
def main():
|
| 438 |
# 1. Parse input arguments
|
| 439 |
# See all possible arguments in src/transformers/training_args.py
|
|
|
|
| 497 |
# Set seed before initializing model.
|
| 498 |
set_seed(training_args.seed)
|
| 499 |
|
| 500 |
+
# Load feature extractor
|
| 501 |
+
feature_extractor = AutoFeatureExtractor.from_pretrained(
|
| 502 |
+
model_args.feature_extractor_name if model_args.feature_extractor_name else model_args.model_name_or_path,
|
| 503 |
+
cache_dir=model_args.cache_dir,
|
| 504 |
+
revision=model_args.model_revision,
|
| 505 |
+
use_auth_token=hf_token if model_args.use_auth_token else None,
|
| 506 |
+
)
|
| 507 |
+
|
| 508 |
# 4. Load dataset
|
| 509 |
raw_datasets = IterableDatasetDict() if data_args.streaming else DatasetDict()
|
| 510 |
|
| 511 |
if training_args.do_train:
|
| 512 |
raw_datasets["train"] = load_maybe_streaming_dataset(
|
| 513 |
+
data_args.dataset_train_name,
|
| 514 |
+
data_args.dataset_train_config_name,
|
| 515 |
split=data_args.train_split_name,
|
| 516 |
+
use_auth_token=hf_token if model_args.use_auth_token else None,
|
| 517 |
streaming=data_args.streaming,
|
| 518 |
+
text_column_name=data_args.text_column_name,
|
| 519 |
+
audio_column_name=data_args.audio_column_name,
|
| 520 |
+
sampling_rate=int(feature_extractor.sampling_rate),
|
| 521 |
+
# language=data_args.language_train
|
| 522 |
)
|
| 523 |
|
| 524 |
if training_args.do_eval:
|
| 525 |
raw_datasets["eval"] = load_maybe_streaming_dataset(
|
| 526 |
+
data_args.dataset_eval_name,
|
| 527 |
+
data_args.dataset_eval_config_name,
|
| 528 |
split=data_args.eval_split_name,
|
| 529 |
+
use_auth_token=hf_token if model_args.use_auth_token else None,
|
| 530 |
streaming=data_args.streaming,
|
| 531 |
+
text_column_name=data_args.text_column_name,
|
| 532 |
+
audio_column_name=data_args.audio_column_name,
|
| 533 |
+
sampling_rate=int(feature_extractor.sampling_rate),
|
| 534 |
+
# language=data_args.language_eval
|
| 535 |
)
|
| 536 |
|
| 537 |
raw_datasets_features = list(next(iter(raw_datasets.values())).features.keys())
|
| 538 |
+
text_column_name = data_args.text_column_name.split(",")[0]
|
| 539 |
if data_args.audio_column_name not in raw_datasets_features:
|
| 540 |
raise ValueError(
|
| 541 |
+
f"--audio_column_name '{data_args.audio_column_name}' not found in dataset '{data_args.dataset_train_name}'. "
|
| 542 |
"Make sure to set `--audio_column_name` to the correct audio column - one of "
|
| 543 |
f"{', '.join(raw_datasets_features)}."
|
| 544 |
)
|
| 545 |
|
| 546 |
+
if text_column_name not in raw_datasets_features:
|
| 547 |
raise ValueError(
|
| 548 |
+
f"--text_column_name {text_column_name} not found in dataset '{data_args.dataset_train_name}'. "
|
| 549 |
"Make sure to set `--text_column_name` to the correct text column - one of "
|
| 550 |
f"{', '.join(raw_datasets_features)}."
|
| 551 |
)
|
|
|
|
| 558 |
model_args.config_name if model_args.config_name else model_args.model_name_or_path,
|
| 559 |
cache_dir=model_args.cache_dir,
|
| 560 |
revision=model_args.model_revision,
|
| 561 |
+
use_auth_token=hf_token if model_args.use_auth_token else None,
|
| 562 |
)
|
| 563 |
|
| 564 |
config.update({"forced_decoder_ids": model_args.forced_decoder_ids, "suppress_tokens": model_args.suppress_tokens})
|
|
|
|
| 566 |
if training_args.gradient_checkpointing:
|
| 567 |
config.update({"use_cache": False})
|
| 568 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 569 |
tokenizer = AutoTokenizer.from_pretrained(
|
| 570 |
model_args.tokenizer_name if model_args.tokenizer_name else model_args.model_name_or_path,
|
| 571 |
cache_dir=model_args.cache_dir,
|
| 572 |
use_fast=model_args.use_fast_tokenizer,
|
| 573 |
revision=model_args.model_revision,
|
| 574 |
+
use_auth_token=hf_token if model_args.use_auth_token else None,
|
| 575 |
)
|
| 576 |
model = AutoModelForSpeechSeq2Seq.from_pretrained(
|
| 577 |
model_args.model_name_or_path,
|
| 578 |
config=config,
|
| 579 |
cache_dir=model_args.cache_dir,
|
| 580 |
revision=model_args.model_revision,
|
| 581 |
+
use_auth_token=hf_token if model_args.use_auth_token else None,
|
| 582 |
)
|
| 583 |
|
| 584 |
if model.config.decoder_start_token_id is None:
|
|
|
|
| 606 |
max_input_length = data_args.max_duration_in_seconds * feature_extractor.sampling_rate
|
| 607 |
min_input_length = data_args.min_duration_in_seconds * feature_extractor.sampling_rate
|
| 608 |
audio_column_name = data_args.audio_column_name
|
|
|
|
| 609 |
model_input_name = feature_extractor.model_input_names[0]
|
| 610 |
do_lower_case = data_args.do_lower_case
|
| 611 |
do_remove_punctuation = data_args.do_remove_punctuation
|
|
|
|
| 725 |
callbacks=[ShuffleCallback()] if data_args.streaming else None,
|
| 726 |
)
|
| 727 |
|
| 728 |
+
orig_push_to_hub = trainer.args.push_to_hub
|
| 729 |
+
trainer.args.push_to_hub = False
|
| 730 |
+
|
| 731 |
# 12. Training
|
| 732 |
if training_args.do_train:
|
| 733 |
checkpoint = None
|
|
|
|
| 766 |
"tasks": "automatic-speech-recognition",
|
| 767 |
"tags": "whisper-event",
|
| 768 |
}
|
| 769 |
+
if data_args.dataset_train_name is not None:
|
| 770 |
+
kwargs["dataset_tags"] = data_args.dataset_train_name
|
| 771 |
+
if data_args.dataset_train_config_name is not None:
|
| 772 |
+
kwargs["dataset"] = f"{data_args.dataset_train_name} {data_args.dataset_train_config_name}"
|
| 773 |
else:
|
| 774 |
+
kwargs["dataset"] = data_args.dataset_train_name
|
| 775 |
+
if "common_voice" in data_args.dataset_train_name:
|
| 776 |
+
kwargs["language"] = data_args.dataset_train_config_name[:2]
|
| 777 |
if model_args.model_index_name is not None:
|
| 778 |
kwargs["model_name"] = model_args.model_index_name
|
| 779 |
|
| 780 |
+
logger.info("*** Training stats written ***")
|
| 781 |
+
logger.info(json.dumps(kwargs, indent=4))
|
| 782 |
+
|
| 783 |
+
# Training complete notification
|
| 784 |
+
logger.info("*** Training and eval complete ***")
|
| 785 |
+
logger.info(SENDING_NOTIFICATION)
|
| 786 |
+
with open(os.path.join(training_args.output_dir, "train_results.json"), "r") as f:
|
| 787 |
+
train_results = json.load(f)
|
| 788 |
+
with open(os.path.join(training_args.output_dir, "eval_results.json"), "r") as f:
|
| 789 |
+
eval_results = json.load(f)
|
| 790 |
+
notify_me(recipient=RECIPIENT_ADDRESS,
|
| 791 |
+
message=f"Training complete! {train_results = } {eval_results = }")
|
| 792 |
+
|
| 793 |
+
trainer.args.push_to_hub = orig_push_to_hub
|
| 794 |
if training_args.push_to_hub:
|
| 795 |
trainer.push_to_hub(**kwargs)
|
| 796 |
else:
|
| 797 |
trainer.create_model_card(**kwargs)
|
| 798 |
+
|
| 799 |
+
with open(os.path.join(training_args.output_dir, "README.md"), "r") as f:
|
| 800 |
+
readme = f.read()
|
| 801 |
+
notify_me(recipient=RECIPIENT_ADDRESS,
|
| 802 |
+
message=f"Model pushed to hub! {readme = }")
|
| 803 |
|
| 804 |
return results
|
| 805 |
|
| 806 |
|
| 807 |
if __name__ == "__main__":
|
| 808 |
+
main()
|
train_results.json
CHANGED
|
@@ -1,7 +1,7 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
-
"train_loss": 0.
|
| 4 |
-
"train_runtime":
|
| 5 |
-
"train_samples_per_second": 3.
|
| 6 |
-
"train_steps_per_second": 0.
|
| 7 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 1.0,
|
| 3 |
+
"train_loss": 0.025400285175442697,
|
| 4 |
+
"train_runtime": 51804.3597,
|
| 5 |
+
"train_samples_per_second": 3.089,
|
| 6 |
+
"train_steps_per_second": 0.097
|
| 7 |
}
|
trainer_state.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric":
|
| 3 |
-
"best_model_checkpoint": "./checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -10,634 +10,1261 @@
|
|
| 10 |
{
|
| 11 |
"epoch": 0.01,
|
| 12 |
"learning_rate": 9.200000000000001e-07,
|
| 13 |
-
"loss": 0.
|
| 14 |
"step": 25
|
| 15 |
},
|
| 16 |
{
|
| 17 |
-
"epoch": 0.
|
| 18 |
"learning_rate": 1.9200000000000003e-06,
|
| 19 |
-
"loss": 0.
|
| 20 |
"step": 50
|
| 21 |
},
|
| 22 |
{
|
| 23 |
-
"epoch": 0.
|
| 24 |
"learning_rate": 2.92e-06,
|
| 25 |
-
"loss": 0.
|
| 26 |
"step": 75
|
| 27 |
},
|
| 28 |
{
|
| 29 |
-
"epoch": 0.
|
| 30 |
"learning_rate": 3.920000000000001e-06,
|
| 31 |
-
"loss": 0.
|
| 32 |
"step": 100
|
| 33 |
},
|
| 34 |
{
|
| 35 |
-
"epoch": 0.
|
| 36 |
"learning_rate": 4.92e-06,
|
| 37 |
-
"loss": 0.
|
| 38 |
"step": 125
|
| 39 |
},
|
| 40 |
{
|
| 41 |
-
"epoch": 0.
|
| 42 |
"learning_rate": 5.92e-06,
|
| 43 |
-
"loss": 0.
|
| 44 |
"step": 150
|
| 45 |
},
|
| 46 |
{
|
| 47 |
-
"epoch": 0.
|
| 48 |
"learning_rate": 6.92e-06,
|
| 49 |
-
"loss": 0.
|
| 50 |
"step": 175
|
| 51 |
},
|
| 52 |
{
|
| 53 |
-
"epoch": 0.
|
| 54 |
"learning_rate": 7.92e-06,
|
| 55 |
-
"loss": 0.
|
| 56 |
"step": 200
|
| 57 |
},
|
| 58 |
{
|
| 59 |
-
"epoch": 0.
|
| 60 |
"learning_rate": 8.920000000000001e-06,
|
| 61 |
-
"loss": 0.
|
| 62 |
"step": 225
|
| 63 |
},
|
| 64 |
{
|
| 65 |
-
"epoch": 0.
|
| 66 |
"learning_rate": 9.920000000000002e-06,
|
| 67 |
-
"loss": 0.
|
| 68 |
"step": 250
|
| 69 |
},
|
| 70 |
{
|
| 71 |
-
"epoch": 0.
|
| 72 |
-
"learning_rate": 9.
|
| 73 |
-
"loss": 0.
|
| 74 |
"step": 275
|
| 75 |
},
|
| 76 |
{
|
| 77 |
-
"epoch": 0.
|
| 78 |
-
"learning_rate": 9.
|
| 79 |
-
"loss": 0.
|
| 80 |
"step": 300
|
| 81 |
},
|
| 82 |
{
|
| 83 |
-
"epoch":
|
| 84 |
-
"learning_rate": 9.
|
| 85 |
-
"loss": 0.
|
| 86 |
"step": 325
|
| 87 |
},
|
| 88 |
{
|
| 89 |
-
"epoch":
|
| 90 |
-
"learning_rate": 9.
|
| 91 |
-
"loss": 0.
|
| 92 |
"step": 350
|
| 93 |
},
|
| 94 |
{
|
| 95 |
-
"epoch":
|
| 96 |
-
"learning_rate": 9.
|
| 97 |
-
"loss": 0.
|
| 98 |
"step": 375
|
| 99 |
},
|
| 100 |
{
|
| 101 |
-
"epoch":
|
| 102 |
-
"learning_rate": 9.
|
| 103 |
-
"loss": 0.
|
| 104 |
"step": 400
|
| 105 |
},
|
| 106 |
{
|
| 107 |
-
"epoch":
|
| 108 |
-
"learning_rate": 9.
|
| 109 |
-
"loss": 0.
|
| 110 |
"step": 425
|
| 111 |
},
|
| 112 |
{
|
| 113 |
-
"epoch":
|
| 114 |
-
"learning_rate": 9.
|
| 115 |
-
"loss": 0.
|
| 116 |
"step": 450
|
| 117 |
},
|
| 118 |
{
|
| 119 |
-
"epoch": 1
|
| 120 |
-
"learning_rate": 9.
|
| 121 |
-
"loss": 0.
|
| 122 |
"step": 475
|
| 123 |
},
|
| 124 |
{
|
| 125 |
-
"epoch": 1
|
| 126 |
-
"learning_rate":
|
| 127 |
-
"loss": 0.
|
| 128 |
"step": 500
|
| 129 |
},
|
| 130 |
{
|
| 131 |
-
"epoch": 1
|
| 132 |
-
"learning_rate":
|
| 133 |
-
"loss": 0.
|
| 134 |
"step": 525
|
| 135 |
},
|
| 136 |
{
|
| 137 |
-
"epoch":
|
| 138 |
-
"learning_rate":
|
| 139 |
-
"loss": 0.
|
| 140 |
"step": 550
|
| 141 |
},
|
| 142 |
{
|
| 143 |
-
"epoch":
|
| 144 |
-
"learning_rate":
|
| 145 |
-
"loss": 0.
|
| 146 |
"step": 575
|
| 147 |
},
|
| 148 |
{
|
| 149 |
-
"epoch":
|
| 150 |
-
"learning_rate":
|
| 151 |
-
"loss": 0.
|
| 152 |
"step": 600
|
| 153 |
},
|
| 154 |
{
|
| 155 |
-
"epoch":
|
| 156 |
-
"learning_rate":
|
| 157 |
-
"loss": 0.
|
| 158 |
"step": 625
|
| 159 |
},
|
| 160 |
{
|
| 161 |
-
"epoch":
|
| 162 |
-
"learning_rate":
|
| 163 |
-
"loss": 0.
|
| 164 |
"step": 650
|
| 165 |
},
|
| 166 |
{
|
| 167 |
-
"epoch":
|
| 168 |
-
"learning_rate":
|
| 169 |
-
"loss": 0.
|
| 170 |
"step": 675
|
| 171 |
},
|
| 172 |
{
|
| 173 |
-
"epoch":
|
| 174 |
-
"learning_rate":
|
| 175 |
-
"loss": 0.
|
| 176 |
"step": 700
|
| 177 |
},
|
| 178 |
{
|
| 179 |
-
"epoch":
|
| 180 |
-
"learning_rate":
|
| 181 |
-
"loss": 0.
|
| 182 |
"step": 725
|
| 183 |
},
|
| 184 |
{
|
| 185 |
-
"epoch":
|
| 186 |
-
"learning_rate":
|
| 187 |
-
"loss": 0.
|
| 188 |
"step": 750
|
| 189 |
},
|
| 190 |
{
|
| 191 |
-
"epoch":
|
| 192 |
-
"learning_rate":
|
| 193 |
-
"loss": 0.
|
| 194 |
"step": 775
|
| 195 |
},
|
| 196 |
{
|
| 197 |
-
"epoch":
|
| 198 |
-
"learning_rate":
|
| 199 |
-
"loss": 0.
|
| 200 |
"step": 800
|
| 201 |
},
|
| 202 |
{
|
| 203 |
-
"epoch":
|
| 204 |
-
"learning_rate":
|
| 205 |
-
"loss": 0.
|
| 206 |
"step": 825
|
| 207 |
},
|
| 208 |
{
|
| 209 |
-
"epoch":
|
| 210 |
-
"learning_rate":
|
| 211 |
-
"loss": 0.
|
| 212 |
"step": 850
|
| 213 |
},
|
| 214 |
{
|
| 215 |
-
"epoch":
|
| 216 |
-
"learning_rate":
|
| 217 |
-
"loss": 0.
|
| 218 |
"step": 875
|
| 219 |
},
|
| 220 |
{
|
| 221 |
-
"epoch":
|
| 222 |
-
"learning_rate":
|
| 223 |
-
"loss": 0.
|
| 224 |
"step": 900
|
| 225 |
},
|
| 226 |
{
|
| 227 |
-
"epoch":
|
| 228 |
-
"learning_rate":
|
| 229 |
-
"loss": 0.
|
| 230 |
"step": 925
|
| 231 |
},
|
| 232 |
{
|
| 233 |
-
"epoch":
|
| 234 |
-
"learning_rate":
|
| 235 |
-
"loss": 0.
|
| 236 |
"step": 950
|
| 237 |
},
|
| 238 |
{
|
| 239 |
-
"epoch":
|
| 240 |
-
"learning_rate":
|
| 241 |
-
"loss": 0.
|
| 242 |
"step": 975
|
| 243 |
},
|
| 244 |
{
|
| 245 |
-
"epoch":
|
| 246 |
-
"learning_rate":
|
| 247 |
-
"loss": 0.
|
| 248 |
"step": 1000
|
| 249 |
},
|
| 250 |
{
|
| 251 |
-
"epoch":
|
| 252 |
-
"eval_loss": 0.
|
| 253 |
-
"eval_runtime":
|
| 254 |
-
"eval_samples_per_second": 1.
|
| 255 |
-
"eval_steps_per_second": 0.
|
| 256 |
-
"eval_wer":
|
| 257 |
"step": 1000
|
| 258 |
},
|
| 259 |
{
|
| 260 |
-
"epoch":
|
| 261 |
-
"learning_rate":
|
| 262 |
-
"loss": 0.
|
| 263 |
"step": 1025
|
| 264 |
},
|
| 265 |
{
|
| 266 |
-
"epoch":
|
| 267 |
-
"learning_rate":
|
| 268 |
-
"loss": 0.
|
| 269 |
"step": 1050
|
| 270 |
},
|
| 271 |
{
|
| 272 |
-
"epoch":
|
| 273 |
-
"learning_rate":
|
| 274 |
-
"loss": 0.
|
| 275 |
"step": 1075
|
| 276 |
},
|
| 277 |
{
|
| 278 |
-
"epoch":
|
| 279 |
-
"learning_rate":
|
| 280 |
-
"loss": 0.
|
| 281 |
"step": 1100
|
| 282 |
},
|
| 283 |
{
|
| 284 |
-
"epoch":
|
| 285 |
-
"learning_rate":
|
| 286 |
-
"loss": 0.
|
| 287 |
"step": 1125
|
| 288 |
},
|
| 289 |
{
|
| 290 |
-
"epoch":
|
| 291 |
-
"learning_rate":
|
| 292 |
-
"loss": 0.
|
| 293 |
"step": 1150
|
| 294 |
},
|
| 295 |
{
|
| 296 |
-
"epoch":
|
| 297 |
-
"learning_rate":
|
| 298 |
-
"loss": 0.
|
| 299 |
"step": 1175
|
| 300 |
},
|
| 301 |
{
|
| 302 |
-
"epoch":
|
| 303 |
-
"learning_rate":
|
| 304 |
-
"loss": 0.
|
| 305 |
"step": 1200
|
| 306 |
},
|
| 307 |
{
|
| 308 |
-
"epoch":
|
| 309 |
-
"learning_rate":
|
| 310 |
-
"loss": 0.
|
| 311 |
"step": 1225
|
| 312 |
},
|
| 313 |
{
|
| 314 |
-
"epoch":
|
| 315 |
-
"learning_rate":
|
| 316 |
-
"loss": 0.
|
| 317 |
"step": 1250
|
| 318 |
},
|
| 319 |
{
|
| 320 |
-
"epoch":
|
| 321 |
-
"learning_rate":
|
| 322 |
-
"loss": 0.
|
| 323 |
"step": 1275
|
| 324 |
},
|
| 325 |
{
|
| 326 |
-
"epoch":
|
| 327 |
-
"learning_rate":
|
| 328 |
-
"loss": 0.
|
| 329 |
"step": 1300
|
| 330 |
},
|
| 331 |
{
|
| 332 |
-
"epoch":
|
| 333 |
-
"learning_rate":
|
| 334 |
-
"loss": 0.
|
| 335 |
"step": 1325
|
| 336 |
},
|
| 337 |
{
|
| 338 |
-
"epoch":
|
| 339 |
-
"learning_rate":
|
| 340 |
-
"loss": 0.
|
| 341 |
"step": 1350
|
| 342 |
},
|
| 343 |
{
|
| 344 |
-
"epoch":
|
| 345 |
-
"learning_rate":
|
| 346 |
-
"loss": 0.
|
| 347 |
"step": 1375
|
| 348 |
},
|
| 349 |
{
|
| 350 |
-
"epoch":
|
| 351 |
-
"learning_rate":
|
| 352 |
-
"loss": 0.
|
| 353 |
"step": 1400
|
| 354 |
},
|
| 355 |
{
|
| 356 |
-
"epoch":
|
| 357 |
-
"learning_rate":
|
| 358 |
-
"loss": 0.
|
| 359 |
"step": 1425
|
| 360 |
},
|
| 361 |
{
|
| 362 |
-
"epoch":
|
| 363 |
-
"learning_rate":
|
| 364 |
-
"loss": 0.
|
| 365 |
"step": 1450
|
| 366 |
},
|
| 367 |
{
|
| 368 |
-
"epoch":
|
| 369 |
-
"learning_rate":
|
| 370 |
-
"loss": 0.
|
| 371 |
"step": 1475
|
| 372 |
},
|
| 373 |
{
|
| 374 |
-
"epoch":
|
| 375 |
-
"learning_rate":
|
| 376 |
-
"loss": 0.
|
| 377 |
"step": 1500
|
| 378 |
},
|
| 379 |
{
|
| 380 |
-
"epoch":
|
| 381 |
-
"learning_rate":
|
| 382 |
-
"loss": 0.
|
| 383 |
"step": 1525
|
| 384 |
},
|
| 385 |
{
|
| 386 |
-
"epoch":
|
| 387 |
-
"learning_rate":
|
| 388 |
-
"loss": 0.
|
| 389 |
"step": 1550
|
| 390 |
},
|
| 391 |
{
|
| 392 |
-
"epoch":
|
| 393 |
-
"learning_rate":
|
| 394 |
-
"loss": 0.
|
| 395 |
"step": 1575
|
| 396 |
},
|
| 397 |
{
|
| 398 |
-
"epoch":
|
| 399 |
-
"learning_rate":
|
| 400 |
-
"loss": 0.
|
| 401 |
"step": 1600
|
| 402 |
},
|
| 403 |
{
|
| 404 |
-
"epoch":
|
| 405 |
-
"learning_rate":
|
| 406 |
-
"loss": 0.
|
| 407 |
"step": 1625
|
| 408 |
},
|
| 409 |
{
|
| 410 |
-
"epoch":
|
| 411 |
-
"learning_rate":
|
| 412 |
-
"loss": 0.
|
| 413 |
"step": 1650
|
| 414 |
},
|
| 415 |
{
|
| 416 |
-
"epoch":
|
| 417 |
-
"learning_rate":
|
| 418 |
-
"loss": 0.
|
| 419 |
"step": 1675
|
| 420 |
},
|
| 421 |
{
|
| 422 |
-
"epoch":
|
| 423 |
-
"learning_rate":
|
| 424 |
-
"loss": 0.
|
| 425 |
"step": 1700
|
| 426 |
},
|
| 427 |
{
|
| 428 |
-
"epoch":
|
| 429 |
-
"learning_rate":
|
| 430 |
-
"loss": 0.
|
| 431 |
"step": 1725
|
| 432 |
},
|
| 433 |
{
|
| 434 |
-
"epoch":
|
| 435 |
-
"learning_rate":
|
| 436 |
-
"loss": 0.
|
| 437 |
"step": 1750
|
| 438 |
},
|
| 439 |
{
|
| 440 |
-
"epoch":
|
| 441 |
-
"learning_rate":
|
| 442 |
-
"loss": 0.
|
| 443 |
"step": 1775
|
| 444 |
},
|
| 445 |
{
|
| 446 |
-
"epoch":
|
| 447 |
-
"learning_rate":
|
| 448 |
-
"loss": 0.
|
| 449 |
"step": 1800
|
| 450 |
},
|
| 451 |
{
|
| 452 |
-
"epoch":
|
| 453 |
-
"learning_rate":
|
| 454 |
-
"loss": 0.
|
| 455 |
"step": 1825
|
| 456 |
},
|
| 457 |
{
|
| 458 |
-
"epoch":
|
| 459 |
-
"learning_rate":
|
| 460 |
-
"loss": 0.
|
| 461 |
"step": 1850
|
| 462 |
},
|
| 463 |
{
|
| 464 |
-
"epoch":
|
| 465 |
-
"learning_rate":
|
| 466 |
-
"loss": 0.
|
| 467 |
"step": 1875
|
| 468 |
},
|
| 469 |
{
|
| 470 |
-
"epoch":
|
| 471 |
-
"learning_rate":
|
| 472 |
-
"loss": 0.
|
| 473 |
"step": 1900
|
| 474 |
},
|
| 475 |
{
|
| 476 |
-
"epoch":
|
| 477 |
-
"learning_rate":
|
| 478 |
-
"loss": 0.
|
| 479 |
"step": 1925
|
| 480 |
},
|
| 481 |
{
|
| 482 |
-
"epoch":
|
| 483 |
-
"learning_rate":
|
| 484 |
-
"loss": 0.
|
| 485 |
"step": 1950
|
| 486 |
},
|
| 487 |
{
|
| 488 |
-
"epoch":
|
| 489 |
-
"learning_rate":
|
| 490 |
-
"loss": 0.
|
| 491 |
"step": 1975
|
| 492 |
},
|
| 493 |
{
|
| 494 |
-
"epoch":
|
| 495 |
-
"learning_rate":
|
| 496 |
-
"loss": 0.
|
| 497 |
"step": 2000
|
| 498 |
},
|
| 499 |
{
|
| 500 |
-
"epoch":
|
| 501 |
-
"eval_loss": 0.
|
| 502 |
-
"eval_runtime":
|
| 503 |
-
"eval_samples_per_second": 1.
|
| 504 |
-
"eval_steps_per_second": 0.
|
| 505 |
-
"eval_wer": 11.
|
| 506 |
"step": 2000
|
| 507 |
},
|
| 508 |
{
|
| 509 |
-
"epoch":
|
| 510 |
-
"learning_rate":
|
| 511 |
-
"loss": 0.
|
| 512 |
"step": 2025
|
| 513 |
},
|
| 514 |
{
|
| 515 |
-
"epoch":
|
| 516 |
-
"learning_rate":
|
| 517 |
-
"loss": 0.
|
| 518 |
"step": 2050
|
| 519 |
},
|
| 520 |
{
|
| 521 |
-
"epoch":
|
| 522 |
-
"learning_rate":
|
| 523 |
-
"loss": 0.
|
| 524 |
"step": 2075
|
| 525 |
},
|
| 526 |
{
|
| 527 |
-
"epoch":
|
| 528 |
-
"learning_rate":
|
| 529 |
-
"loss": 0.
|
| 530 |
"step": 2100
|
| 531 |
},
|
| 532 |
{
|
| 533 |
-
"epoch":
|
| 534 |
-
"learning_rate":
|
| 535 |
-
"loss": 0.
|
| 536 |
"step": 2125
|
| 537 |
},
|
| 538 |
{
|
| 539 |
-
"epoch":
|
| 540 |
-
"learning_rate":
|
| 541 |
-
"loss": 0.
|
| 542 |
"step": 2150
|
| 543 |
},
|
| 544 |
{
|
| 545 |
-
"epoch":
|
| 546 |
-
"learning_rate":
|
| 547 |
-
"loss": 0.
|
| 548 |
"step": 2175
|
| 549 |
},
|
| 550 |
{
|
| 551 |
-
"epoch":
|
| 552 |
-
"learning_rate":
|
| 553 |
-
"loss": 0.
|
| 554 |
"step": 2200
|
| 555 |
},
|
| 556 |
{
|
| 557 |
-
"epoch":
|
| 558 |
-
"learning_rate":
|
| 559 |
-
"loss": 0.
|
| 560 |
"step": 2225
|
| 561 |
},
|
| 562 |
{
|
| 563 |
-
"epoch":
|
| 564 |
-
"learning_rate":
|
| 565 |
-
"loss": 0.
|
| 566 |
"step": 2250
|
| 567 |
},
|
| 568 |
{
|
| 569 |
-
"epoch":
|
| 570 |
-
"learning_rate":
|
| 571 |
-
"loss": 0.
|
| 572 |
"step": 2275
|
| 573 |
},
|
| 574 |
{
|
| 575 |
-
"epoch":
|
| 576 |
-
"learning_rate":
|
| 577 |
-
"loss": 0.
|
| 578 |
"step": 2300
|
| 579 |
},
|
| 580 |
{
|
| 581 |
-
"epoch":
|
| 582 |
-
"learning_rate":
|
| 583 |
-
"loss": 0.
|
| 584 |
"step": 2325
|
| 585 |
},
|
| 586 |
{
|
| 587 |
-
"epoch":
|
| 588 |
-
"learning_rate":
|
| 589 |
-
"loss": 0.
|
| 590 |
"step": 2350
|
| 591 |
},
|
| 592 |
{
|
| 593 |
-
"epoch":
|
| 594 |
-
"learning_rate": 5.
|
| 595 |
-
"loss": 0.
|
| 596 |
"step": 2375
|
| 597 |
},
|
| 598 |
{
|
| 599 |
-
"epoch":
|
| 600 |
-
"learning_rate":
|
| 601 |
-
"loss": 0.
|
| 602 |
"step": 2400
|
| 603 |
},
|
| 604 |
{
|
| 605 |
-
"epoch":
|
| 606 |
-
"learning_rate":
|
| 607 |
-
"loss": 0.
|
| 608 |
"step": 2425
|
| 609 |
},
|
| 610 |
{
|
| 611 |
-
"epoch":
|
| 612 |
-
"learning_rate":
|
| 613 |
-
"loss": 0.
|
| 614 |
"step": 2450
|
| 615 |
},
|
| 616 |
{
|
| 617 |
-
"epoch":
|
| 618 |
-
"learning_rate":
|
| 619 |
-
"loss": 0.
|
| 620 |
"step": 2475
|
| 621 |
},
|
| 622 |
{
|
| 623 |
-
"epoch":
|
| 624 |
-
"learning_rate":
|
| 625 |
-
"loss": 0.
|
| 626 |
"step": 2500
|
| 627 |
},
|
| 628 |
{
|
| 629 |
-
"epoch":
|
| 630 |
-
"
|
| 631 |
-
"
|
| 632 |
-
"
|
| 633 |
-
|
| 634 |
-
|
| 635 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 636 |
}
|
| 637 |
],
|
| 638 |
-
"max_steps":
|
| 639 |
"num_train_epochs": 9223372036854775807,
|
| 640 |
-
"total_flos":
|
| 641 |
"trial_name": null,
|
| 642 |
"trial_params": null
|
| 643 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 9.891409525857435,
|
| 3 |
+
"best_model_checkpoint": "./checkpoint-5000",
|
| 4 |
+
"epoch": 1.0,
|
| 5 |
+
"global_step": 5000,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 10 |
{
|
| 11 |
"epoch": 0.01,
|
| 12 |
"learning_rate": 9.200000000000001e-07,
|
| 13 |
+
"loss": 0.4442,
|
| 14 |
"step": 25
|
| 15 |
},
|
| 16 |
{
|
| 17 |
+
"epoch": 0.01,
|
| 18 |
"learning_rate": 1.9200000000000003e-06,
|
| 19 |
+
"loss": 0.1027,
|
| 20 |
"step": 50
|
| 21 |
},
|
| 22 |
{
|
| 23 |
+
"epoch": 0.01,
|
| 24 |
"learning_rate": 2.92e-06,
|
| 25 |
+
"loss": 0.0877,
|
| 26 |
"step": 75
|
| 27 |
},
|
| 28 |
{
|
| 29 |
+
"epoch": 0.02,
|
| 30 |
"learning_rate": 3.920000000000001e-06,
|
| 31 |
+
"loss": 0.0788,
|
| 32 |
"step": 100
|
| 33 |
},
|
| 34 |
{
|
| 35 |
+
"epoch": 0.03,
|
| 36 |
"learning_rate": 4.92e-06,
|
| 37 |
+
"loss": 0.0815,
|
| 38 |
"step": 125
|
| 39 |
},
|
| 40 |
{
|
| 41 |
+
"epoch": 0.03,
|
| 42 |
"learning_rate": 5.92e-06,
|
| 43 |
+
"loss": 0.0817,
|
| 44 |
"step": 150
|
| 45 |
},
|
| 46 |
{
|
| 47 |
+
"epoch": 0.04,
|
| 48 |
"learning_rate": 6.92e-06,
|
| 49 |
+
"loss": 0.0614,
|
| 50 |
"step": 175
|
| 51 |
},
|
| 52 |
{
|
| 53 |
+
"epoch": 0.04,
|
| 54 |
"learning_rate": 7.92e-06,
|
| 55 |
+
"loss": 0.049,
|
| 56 |
"step": 200
|
| 57 |
},
|
| 58 |
{
|
| 59 |
+
"epoch": 0.04,
|
| 60 |
"learning_rate": 8.920000000000001e-06,
|
| 61 |
+
"loss": 0.0471,
|
| 62 |
"step": 225
|
| 63 |
},
|
| 64 |
{
|
| 65 |
+
"epoch": 0.05,
|
| 66 |
"learning_rate": 9.920000000000002e-06,
|
| 67 |
+
"loss": 0.0456,
|
| 68 |
"step": 250
|
| 69 |
},
|
| 70 |
{
|
| 71 |
+
"epoch": 0.06,
|
| 72 |
+
"learning_rate": 9.951578947368423e-06,
|
| 73 |
+
"loss": 0.0483,
|
| 74 |
"step": 275
|
| 75 |
},
|
| 76 |
{
|
| 77 |
+
"epoch": 0.06,
|
| 78 |
+
"learning_rate": 9.898947368421054e-06,
|
| 79 |
+
"loss": 0.051,
|
| 80 |
"step": 300
|
| 81 |
},
|
| 82 |
{
|
| 83 |
+
"epoch": 0.07,
|
| 84 |
+
"learning_rate": 9.846315789473684e-06,
|
| 85 |
+
"loss": 0.0526,
|
| 86 |
"step": 325
|
| 87 |
},
|
| 88 |
{
|
| 89 |
+
"epoch": 0.07,
|
| 90 |
+
"learning_rate": 9.793684210526316e-06,
|
| 91 |
+
"loss": 0.0609,
|
| 92 |
"step": 350
|
| 93 |
},
|
| 94 |
{
|
| 95 |
+
"epoch": 0.07,
|
| 96 |
+
"learning_rate": 9.741052631578947e-06,
|
| 97 |
+
"loss": 0.052,
|
| 98 |
"step": 375
|
| 99 |
},
|
| 100 |
{
|
| 101 |
+
"epoch": 0.08,
|
| 102 |
+
"learning_rate": 9.68842105263158e-06,
|
| 103 |
+
"loss": 0.0476,
|
| 104 |
"step": 400
|
| 105 |
},
|
| 106 |
{
|
| 107 |
+
"epoch": 0.09,
|
| 108 |
+
"learning_rate": 9.635789473684212e-06,
|
| 109 |
+
"loss": 0.0433,
|
| 110 |
"step": 425
|
| 111 |
},
|
| 112 |
{
|
| 113 |
+
"epoch": 0.09,
|
| 114 |
+
"learning_rate": 9.583157894736842e-06,
|
| 115 |
+
"loss": 0.0508,
|
| 116 |
"step": 450
|
| 117 |
},
|
| 118 |
{
|
| 119 |
+
"epoch": 0.1,
|
| 120 |
+
"learning_rate": 9.530526315789474e-06,
|
| 121 |
+
"loss": 0.0551,
|
| 122 |
"step": 475
|
| 123 |
},
|
| 124 |
{
|
| 125 |
+
"epoch": 0.1,
|
| 126 |
+
"learning_rate": 9.477894736842106e-06,
|
| 127 |
+
"loss": 0.0413,
|
| 128 |
"step": 500
|
| 129 |
},
|
| 130 |
{
|
| 131 |
+
"epoch": 0.1,
|
| 132 |
+
"learning_rate": 9.425263157894737e-06,
|
| 133 |
+
"loss": 0.0313,
|
| 134 |
"step": 525
|
| 135 |
},
|
| 136 |
{
|
| 137 |
+
"epoch": 0.11,
|
| 138 |
+
"learning_rate": 9.372631578947369e-06,
|
| 139 |
+
"loss": 0.0411,
|
| 140 |
"step": 550
|
| 141 |
},
|
| 142 |
{
|
| 143 |
+
"epoch": 0.12,
|
| 144 |
+
"learning_rate": 9.32e-06,
|
| 145 |
+
"loss": 0.0496,
|
| 146 |
"step": 575
|
| 147 |
},
|
| 148 |
{
|
| 149 |
+
"epoch": 0.12,
|
| 150 |
+
"learning_rate": 9.267368421052632e-06,
|
| 151 |
+
"loss": 0.0442,
|
| 152 |
"step": 600
|
| 153 |
},
|
| 154 |
{
|
| 155 |
+
"epoch": 0.12,
|
| 156 |
+
"learning_rate": 9.214736842105264e-06,
|
| 157 |
+
"loss": 0.0445,
|
| 158 |
"step": 625
|
| 159 |
},
|
| 160 |
{
|
| 161 |
+
"epoch": 0.13,
|
| 162 |
+
"learning_rate": 9.162105263157895e-06,
|
| 163 |
+
"loss": 0.0536,
|
| 164 |
"step": 650
|
| 165 |
},
|
| 166 |
{
|
| 167 |
+
"epoch": 0.14,
|
| 168 |
+
"learning_rate": 9.109473684210527e-06,
|
| 169 |
+
"loss": 0.036,
|
| 170 |
"step": 675
|
| 171 |
},
|
| 172 |
{
|
| 173 |
+
"epoch": 0.14,
|
| 174 |
+
"learning_rate": 9.056842105263159e-06,
|
| 175 |
+
"loss": 0.034,
|
| 176 |
"step": 700
|
| 177 |
},
|
| 178 |
{
|
| 179 |
+
"epoch": 0.14,
|
| 180 |
+
"learning_rate": 9.00421052631579e-06,
|
| 181 |
+
"loss": 0.0305,
|
| 182 |
"step": 725
|
| 183 |
},
|
| 184 |
{
|
| 185 |
+
"epoch": 0.15,
|
| 186 |
+
"learning_rate": 8.951578947368422e-06,
|
| 187 |
+
"loss": 0.0301,
|
| 188 |
"step": 750
|
| 189 |
},
|
| 190 |
{
|
| 191 |
+
"epoch": 0.15,
|
| 192 |
+
"learning_rate": 8.898947368421054e-06,
|
| 193 |
+
"loss": 0.0362,
|
| 194 |
"step": 775
|
| 195 |
},
|
| 196 |
{
|
| 197 |
+
"epoch": 0.16,
|
| 198 |
+
"learning_rate": 8.846315789473685e-06,
|
| 199 |
+
"loss": 0.0323,
|
| 200 |
"step": 800
|
| 201 |
},
|
| 202 |
{
|
| 203 |
+
"epoch": 0.17,
|
| 204 |
+
"learning_rate": 8.793684210526317e-06,
|
| 205 |
+
"loss": 0.0416,
|
| 206 |
"step": 825
|
| 207 |
},
|
| 208 |
{
|
| 209 |
+
"epoch": 0.17,
|
| 210 |
+
"learning_rate": 8.741052631578949e-06,
|
| 211 |
+
"loss": 0.033,
|
| 212 |
"step": 850
|
| 213 |
},
|
| 214 |
{
|
| 215 |
+
"epoch": 0.17,
|
| 216 |
+
"learning_rate": 8.688421052631579e-06,
|
| 217 |
+
"loss": 0.0372,
|
| 218 |
"step": 875
|
| 219 |
},
|
| 220 |
{
|
| 221 |
+
"epoch": 0.18,
|
| 222 |
+
"learning_rate": 8.63578947368421e-06,
|
| 223 |
+
"loss": 0.0332,
|
| 224 |
"step": 900
|
| 225 |
},
|
| 226 |
{
|
| 227 |
+
"epoch": 0.18,
|
| 228 |
+
"learning_rate": 8.583157894736843e-06,
|
| 229 |
+
"loss": 0.0319,
|
| 230 |
"step": 925
|
| 231 |
},
|
| 232 |
{
|
| 233 |
+
"epoch": 0.19,
|
| 234 |
+
"learning_rate": 8.530526315789475e-06,
|
| 235 |
+
"loss": 0.037,
|
| 236 |
"step": 950
|
| 237 |
},
|
| 238 |
{
|
| 239 |
+
"epoch": 0.2,
|
| 240 |
+
"learning_rate": 8.477894736842107e-06,
|
| 241 |
+
"loss": 0.0437,
|
| 242 |
"step": 975
|
| 243 |
},
|
| 244 |
{
|
| 245 |
+
"epoch": 0.2,
|
| 246 |
+
"learning_rate": 8.425263157894737e-06,
|
| 247 |
+
"loss": 0.0383,
|
| 248 |
"step": 1000
|
| 249 |
},
|
| 250 |
{
|
| 251 |
+
"epoch": 0.2,
|
| 252 |
+
"eval_loss": 0.22878967225551605,
|
| 253 |
+
"eval_runtime": 2943.9063,
|
| 254 |
+
"eval_samples_per_second": 1.722,
|
| 255 |
+
"eval_steps_per_second": 0.108,
|
| 256 |
+
"eval_wer": 12.202988925921945,
|
| 257 |
"step": 1000
|
| 258 |
},
|
| 259 |
{
|
| 260 |
+
"epoch": 0.2,
|
| 261 |
+
"learning_rate": 8.372631578947368e-06,
|
| 262 |
+
"loss": 0.0308,
|
| 263 |
"step": 1025
|
| 264 |
},
|
| 265 |
{
|
| 266 |
+
"epoch": 0.21,
|
| 267 |
+
"learning_rate": 8.32e-06,
|
| 268 |
+
"loss": 0.0329,
|
| 269 |
"step": 1050
|
| 270 |
},
|
| 271 |
{
|
| 272 |
+
"epoch": 0.21,
|
| 273 |
+
"learning_rate": 8.267368421052632e-06,
|
| 274 |
+
"loss": 0.0351,
|
| 275 |
"step": 1075
|
| 276 |
},
|
| 277 |
{
|
| 278 |
+
"epoch": 0.22,
|
| 279 |
+
"learning_rate": 8.214736842105265e-06,
|
| 280 |
+
"loss": 0.0333,
|
| 281 |
"step": 1100
|
| 282 |
},
|
| 283 |
{
|
| 284 |
+
"epoch": 0.23,
|
| 285 |
+
"learning_rate": 8.162105263157895e-06,
|
| 286 |
+
"loss": 0.0349,
|
| 287 |
"step": 1125
|
| 288 |
},
|
| 289 |
{
|
| 290 |
+
"epoch": 0.23,
|
| 291 |
+
"learning_rate": 8.109473684210527e-06,
|
| 292 |
+
"loss": 0.0339,
|
| 293 |
"step": 1150
|
| 294 |
},
|
| 295 |
{
|
| 296 |
+
"epoch": 0.23,
|
| 297 |
+
"learning_rate": 8.056842105263158e-06,
|
| 298 |
+
"loss": 0.0276,
|
| 299 |
"step": 1175
|
| 300 |
},
|
| 301 |
{
|
| 302 |
+
"epoch": 0.24,
|
| 303 |
+
"learning_rate": 8.00421052631579e-06,
|
| 304 |
+
"loss": 0.0344,
|
| 305 |
"step": 1200
|
| 306 |
},
|
| 307 |
{
|
| 308 |
+
"epoch": 0.24,
|
| 309 |
+
"learning_rate": 7.951578947368421e-06,
|
| 310 |
+
"loss": 0.0368,
|
| 311 |
"step": 1225
|
| 312 |
},
|
| 313 |
{
|
| 314 |
+
"epoch": 0.25,
|
| 315 |
+
"learning_rate": 7.898947368421053e-06,
|
| 316 |
+
"loss": 0.0331,
|
| 317 |
"step": 1250
|
| 318 |
},
|
| 319 |
{
|
| 320 |
+
"epoch": 0.26,
|
| 321 |
+
"learning_rate": 7.846315789473685e-06,
|
| 322 |
+
"loss": 0.028,
|
| 323 |
"step": 1275
|
| 324 |
},
|
| 325 |
{
|
| 326 |
+
"epoch": 0.26,
|
| 327 |
+
"learning_rate": 7.793684210526316e-06,
|
| 328 |
+
"loss": 0.0333,
|
| 329 |
"step": 1300
|
| 330 |
},
|
| 331 |
{
|
| 332 |
+
"epoch": 0.27,
|
| 333 |
+
"learning_rate": 7.741052631578948e-06,
|
| 334 |
+
"loss": 0.0299,
|
| 335 |
"step": 1325
|
| 336 |
},
|
| 337 |
{
|
| 338 |
+
"epoch": 0.27,
|
| 339 |
+
"learning_rate": 7.68842105263158e-06,
|
| 340 |
+
"loss": 0.0252,
|
| 341 |
"step": 1350
|
| 342 |
},
|
| 343 |
{
|
| 344 |
+
"epoch": 0.28,
|
| 345 |
+
"learning_rate": 7.635789473684211e-06,
|
| 346 |
+
"loss": 0.0288,
|
| 347 |
"step": 1375
|
| 348 |
},
|
| 349 |
{
|
| 350 |
+
"epoch": 0.28,
|
| 351 |
+
"learning_rate": 7.583157894736842e-06,
|
| 352 |
+
"loss": 0.0294,
|
| 353 |
"step": 1400
|
| 354 |
},
|
| 355 |
{
|
| 356 |
+
"epoch": 0.28,
|
| 357 |
+
"learning_rate": 7.5305263157894745e-06,
|
| 358 |
+
"loss": 0.0265,
|
| 359 |
"step": 1425
|
| 360 |
},
|
| 361 |
{
|
| 362 |
+
"epoch": 0.29,
|
| 363 |
+
"learning_rate": 7.477894736842106e-06,
|
| 364 |
+
"loss": 0.0239,
|
| 365 |
"step": 1450
|
| 366 |
},
|
| 367 |
{
|
| 368 |
+
"epoch": 0.29,
|
| 369 |
+
"learning_rate": 7.425263157894738e-06,
|
| 370 |
+
"loss": 0.0248,
|
| 371 |
"step": 1475
|
| 372 |
},
|
| 373 |
{
|
| 374 |
+
"epoch": 0.3,
|
| 375 |
+
"learning_rate": 7.3726315789473694e-06,
|
| 376 |
+
"loss": 0.0306,
|
| 377 |
"step": 1500
|
| 378 |
},
|
| 379 |
{
|
| 380 |
+
"epoch": 0.3,
|
| 381 |
+
"learning_rate": 7.32e-06,
|
| 382 |
+
"loss": 0.0262,
|
| 383 |
"step": 1525
|
| 384 |
},
|
| 385 |
{
|
| 386 |
+
"epoch": 0.31,
|
| 387 |
+
"learning_rate": 7.267368421052632e-06,
|
| 388 |
+
"loss": 0.0257,
|
| 389 |
"step": 1550
|
| 390 |
},
|
| 391 |
{
|
| 392 |
+
"epoch": 0.32,
|
| 393 |
+
"learning_rate": 7.2147368421052635e-06,
|
| 394 |
+
"loss": 0.0229,
|
| 395 |
"step": 1575
|
| 396 |
},
|
| 397 |
{
|
| 398 |
+
"epoch": 0.32,
|
| 399 |
+
"learning_rate": 7.162105263157896e-06,
|
| 400 |
+
"loss": 0.0238,
|
| 401 |
"step": 1600
|
| 402 |
},
|
| 403 |
{
|
| 404 |
+
"epoch": 0.33,
|
| 405 |
+
"learning_rate": 7.109473684210528e-06,
|
| 406 |
+
"loss": 0.0187,
|
| 407 |
"step": 1625
|
| 408 |
},
|
| 409 |
{
|
| 410 |
+
"epoch": 0.33,
|
| 411 |
+
"learning_rate": 7.056842105263158e-06,
|
| 412 |
+
"loss": 0.0232,
|
| 413 |
"step": 1650
|
| 414 |
},
|
| 415 |
{
|
| 416 |
+
"epoch": 0.34,
|
| 417 |
+
"learning_rate": 7.00421052631579e-06,
|
| 418 |
+
"loss": 0.0273,
|
| 419 |
"step": 1675
|
| 420 |
},
|
| 421 |
{
|
| 422 |
+
"epoch": 0.34,
|
| 423 |
+
"learning_rate": 6.951578947368422e-06,
|
| 424 |
+
"loss": 0.0326,
|
| 425 |
"step": 1700
|
| 426 |
},
|
| 427 |
{
|
| 428 |
+
"epoch": 0.34,
|
| 429 |
+
"learning_rate": 6.8989473684210525e-06,
|
| 430 |
+
"loss": 0.0266,
|
| 431 |
"step": 1725
|
| 432 |
},
|
| 433 |
{
|
| 434 |
+
"epoch": 0.35,
|
| 435 |
+
"learning_rate": 6.846315789473684e-06,
|
| 436 |
+
"loss": 0.0245,
|
| 437 |
"step": 1750
|
| 438 |
},
|
| 439 |
{
|
| 440 |
+
"epoch": 0.35,
|
| 441 |
+
"learning_rate": 6.793684210526317e-06,
|
| 442 |
+
"loss": 0.0345,
|
| 443 |
"step": 1775
|
| 444 |
},
|
| 445 |
{
|
| 446 |
+
"epoch": 0.36,
|
| 447 |
+
"learning_rate": 6.741052631578948e-06,
|
| 448 |
+
"loss": 0.033,
|
| 449 |
"step": 1800
|
| 450 |
},
|
| 451 |
{
|
| 452 |
+
"epoch": 0.36,
|
| 453 |
+
"learning_rate": 6.68842105263158e-06,
|
| 454 |
+
"loss": 0.0293,
|
| 455 |
"step": 1825
|
| 456 |
},
|
| 457 |
{
|
| 458 |
+
"epoch": 0.37,
|
| 459 |
+
"learning_rate": 6.635789473684211e-06,
|
| 460 |
+
"loss": 0.0187,
|
| 461 |
"step": 1850
|
| 462 |
},
|
| 463 |
{
|
| 464 |
+
"epoch": 0.38,
|
| 465 |
+
"learning_rate": 6.583157894736842e-06,
|
| 466 |
+
"loss": 0.0181,
|
| 467 |
"step": 1875
|
| 468 |
},
|
| 469 |
{
|
| 470 |
+
"epoch": 0.38,
|
| 471 |
+
"learning_rate": 6.530526315789474e-06,
|
| 472 |
+
"loss": 0.0242,
|
| 473 |
"step": 1900
|
| 474 |
},
|
| 475 |
{
|
| 476 |
+
"epoch": 0.39,
|
| 477 |
+
"learning_rate": 6.477894736842106e-06,
|
| 478 |
+
"loss": 0.0214,
|
| 479 |
"step": 1925
|
| 480 |
},
|
| 481 |
{
|
| 482 |
+
"epoch": 0.39,
|
| 483 |
+
"learning_rate": 6.425263157894738e-06,
|
| 484 |
+
"loss": 0.0238,
|
| 485 |
"step": 1950
|
| 486 |
},
|
| 487 |
{
|
| 488 |
+
"epoch": 0.4,
|
| 489 |
+
"learning_rate": 6.372631578947369e-06,
|
| 490 |
+
"loss": 0.0168,
|
| 491 |
"step": 1975
|
| 492 |
},
|
| 493 |
{
|
| 494 |
+
"epoch": 0.4,
|
| 495 |
+
"learning_rate": 6.3200000000000005e-06,
|
| 496 |
+
"loss": 0.0239,
|
| 497 |
"step": 2000
|
| 498 |
},
|
| 499 |
{
|
| 500 |
+
"epoch": 0.4,
|
| 501 |
+
"eval_loss": 0.2214207649230957,
|
| 502 |
+
"eval_runtime": 2933.0533,
|
| 503 |
+
"eval_samples_per_second": 1.728,
|
| 504 |
+
"eval_steps_per_second": 0.108,
|
| 505 |
+
"eval_wer": 11.38049672078271,
|
| 506 |
"step": 2000
|
| 507 |
},
|
| 508 |
{
|
| 509 |
+
"epoch": 0.41,
|
| 510 |
+
"learning_rate": 6.267368421052632e-06,
|
| 511 |
+
"loss": 0.0201,
|
| 512 |
"step": 2025
|
| 513 |
},
|
| 514 |
{
|
| 515 |
+
"epoch": 0.41,
|
| 516 |
+
"learning_rate": 6.214736842105264e-06,
|
| 517 |
+
"loss": 0.0149,
|
| 518 |
"step": 2050
|
| 519 |
},
|
| 520 |
{
|
| 521 |
+
"epoch": 0.41,
|
| 522 |
+
"learning_rate": 6.1621052631578945e-06,
|
| 523 |
+
"loss": 0.0167,
|
| 524 |
"step": 2075
|
| 525 |
},
|
| 526 |
{
|
| 527 |
+
"epoch": 0.42,
|
| 528 |
+
"learning_rate": 6.109473684210527e-06,
|
| 529 |
+
"loss": 0.0169,
|
| 530 |
"step": 2100
|
| 531 |
},
|
| 532 |
{
|
| 533 |
+
"epoch": 0.42,
|
| 534 |
+
"learning_rate": 6.056842105263159e-06,
|
| 535 |
+
"loss": 0.0157,
|
| 536 |
"step": 2125
|
| 537 |
},
|
| 538 |
{
|
| 539 |
+
"epoch": 0.43,
|
| 540 |
+
"learning_rate": 6.00421052631579e-06,
|
| 541 |
+
"loss": 0.0168,
|
| 542 |
"step": 2150
|
| 543 |
},
|
| 544 |
{
|
| 545 |
+
"epoch": 0.43,
|
| 546 |
+
"learning_rate": 5.951578947368422e-06,
|
| 547 |
+
"loss": 0.0173,
|
| 548 |
"step": 2175
|
| 549 |
},
|
| 550 |
{
|
| 551 |
+
"epoch": 0.44,
|
| 552 |
+
"learning_rate": 5.898947368421053e-06,
|
| 553 |
+
"loss": 0.0167,
|
| 554 |
"step": 2200
|
| 555 |
},
|
| 556 |
{
|
| 557 |
+
"epoch": 0.45,
|
| 558 |
+
"learning_rate": 5.846315789473684e-06,
|
| 559 |
+
"loss": 0.015,
|
| 560 |
"step": 2225
|
| 561 |
},
|
| 562 |
{
|
| 563 |
+
"epoch": 0.45,
|
| 564 |
+
"learning_rate": 5.793684210526316e-06,
|
| 565 |
+
"loss": 0.0212,
|
| 566 |
"step": 2250
|
| 567 |
},
|
| 568 |
{
|
| 569 |
+
"epoch": 0.46,
|
| 570 |
+
"learning_rate": 5.7410526315789485e-06,
|
| 571 |
+
"loss": 0.0208,
|
| 572 |
"step": 2275
|
| 573 |
},
|
| 574 |
{
|
| 575 |
+
"epoch": 0.46,
|
| 576 |
+
"learning_rate": 5.68842105263158e-06,
|
| 577 |
+
"loss": 0.0132,
|
| 578 |
"step": 2300
|
| 579 |
},
|
| 580 |
{
|
| 581 |
+
"epoch": 0.47,
|
| 582 |
+
"learning_rate": 5.635789473684211e-06,
|
| 583 |
+
"loss": 0.0133,
|
| 584 |
"step": 2325
|
| 585 |
},
|
| 586 |
{
|
| 587 |
+
"epoch": 0.47,
|
| 588 |
+
"learning_rate": 5.5831578947368425e-06,
|
| 589 |
+
"loss": 0.0146,
|
| 590 |
"step": 2350
|
| 591 |
},
|
| 592 |
{
|
| 593 |
+
"epoch": 0.47,
|
| 594 |
+
"learning_rate": 5.530526315789474e-06,
|
| 595 |
+
"loss": 0.0163,
|
| 596 |
"step": 2375
|
| 597 |
},
|
| 598 |
{
|
| 599 |
+
"epoch": 0.48,
|
| 600 |
+
"learning_rate": 5.477894736842105e-06,
|
| 601 |
+
"loss": 0.0141,
|
| 602 |
"step": 2400
|
| 603 |
},
|
| 604 |
{
|
| 605 |
+
"epoch": 0.48,
|
| 606 |
+
"learning_rate": 5.425263157894737e-06,
|
| 607 |
+
"loss": 0.0168,
|
| 608 |
"step": 2425
|
| 609 |
},
|
| 610 |
{
|
| 611 |
+
"epoch": 0.49,
|
| 612 |
+
"learning_rate": 5.372631578947369e-06,
|
| 613 |
+
"loss": 0.0189,
|
| 614 |
"step": 2450
|
| 615 |
},
|
| 616 |
{
|
| 617 |
+
"epoch": 0.49,
|
| 618 |
+
"learning_rate": 5.320000000000001e-06,
|
| 619 |
+
"loss": 0.0204,
|
| 620 |
"step": 2475
|
| 621 |
},
|
| 622 |
{
|
| 623 |
+
"epoch": 0.5,
|
| 624 |
+
"learning_rate": 5.267368421052632e-06,
|
| 625 |
+
"loss": 0.0165,
|
| 626 |
"step": 2500
|
| 627 |
},
|
| 628 |
{
|
| 629 |
+
"epoch": 0.51,
|
| 630 |
+
"learning_rate": 5.214736842105263e-06,
|
| 631 |
+
"loss": 0.0192,
|
| 632 |
+
"step": 2525
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 0.51,
|
| 636 |
+
"learning_rate": 5.162105263157895e-06,
|
| 637 |
+
"loss": 0.0164,
|
| 638 |
+
"step": 2550
|
| 639 |
+
},
|
| 640 |
+
{
|
| 641 |
+
"epoch": 0.52,
|
| 642 |
+
"learning_rate": 5.1094736842105264e-06,
|
| 643 |
+
"loss": 0.0123,
|
| 644 |
+
"step": 2575
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 0.52,
|
| 648 |
+
"learning_rate": 5.056842105263158e-06,
|
| 649 |
+
"loss": 0.0174,
|
| 650 |
+
"step": 2600
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 0.53,
|
| 654 |
+
"learning_rate": 5.0042105263157906e-06,
|
| 655 |
+
"loss": 0.0152,
|
| 656 |
+
"step": 2625
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 0.53,
|
| 660 |
+
"learning_rate": 4.951578947368421e-06,
|
| 661 |
+
"loss": 0.0156,
|
| 662 |
+
"step": 2650
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 0.54,
|
| 666 |
+
"learning_rate": 4.898947368421053e-06,
|
| 667 |
+
"loss": 0.0138,
|
| 668 |
+
"step": 2675
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 0.54,
|
| 672 |
+
"learning_rate": 4.846315789473685e-06,
|
| 673 |
+
"loss": 0.0138,
|
| 674 |
+
"step": 2700
|
| 675 |
+
},
|
| 676 |
+
{
|
| 677 |
+
"epoch": 0.55,
|
| 678 |
+
"learning_rate": 4.793684210526316e-06,
|
| 679 |
+
"loss": 0.0142,
|
| 680 |
+
"step": 2725
|
| 681 |
+
},
|
| 682 |
+
{
|
| 683 |
+
"epoch": 0.55,
|
| 684 |
+
"learning_rate": 4.741052631578948e-06,
|
| 685 |
+
"loss": 0.0119,
|
| 686 |
+
"step": 2750
|
| 687 |
+
},
|
| 688 |
+
{
|
| 689 |
+
"epoch": 0.56,
|
| 690 |
+
"learning_rate": 4.6884210526315795e-06,
|
| 691 |
+
"loss": 0.0154,
|
| 692 |
+
"step": 2775
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 0.56,
|
| 696 |
+
"learning_rate": 4.63578947368421e-06,
|
| 697 |
+
"loss": 0.0137,
|
| 698 |
+
"step": 2800
|
| 699 |
+
},
|
| 700 |
+
{
|
| 701 |
+
"epoch": 0.56,
|
| 702 |
+
"learning_rate": 4.583157894736843e-06,
|
| 703 |
+
"loss": 0.0167,
|
| 704 |
+
"step": 2825
|
| 705 |
+
},
|
| 706 |
+
{
|
| 707 |
+
"epoch": 0.57,
|
| 708 |
+
"learning_rate": 4.5305263157894744e-06,
|
| 709 |
+
"loss": 0.0172,
|
| 710 |
+
"step": 2850
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"epoch": 0.57,
|
| 714 |
+
"learning_rate": 4.477894736842105e-06,
|
| 715 |
+
"loss": 0.013,
|
| 716 |
+
"step": 2875
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 0.58,
|
| 720 |
+
"learning_rate": 4.425263157894737e-06,
|
| 721 |
+
"loss": 0.0141,
|
| 722 |
+
"step": 2900
|
| 723 |
+
},
|
| 724 |
+
{
|
| 725 |
+
"epoch": 0.58,
|
| 726 |
+
"learning_rate": 4.3726315789473685e-06,
|
| 727 |
+
"loss": 0.0118,
|
| 728 |
+
"step": 2925
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 0.59,
|
| 732 |
+
"learning_rate": 4.32e-06,
|
| 733 |
+
"loss": 0.0179,
|
| 734 |
+
"step": 2950
|
| 735 |
+
},
|
| 736 |
+
{
|
| 737 |
+
"epoch": 0.59,
|
| 738 |
+
"learning_rate": 4.267368421052632e-06,
|
| 739 |
+
"loss": 0.0143,
|
| 740 |
+
"step": 2975
|
| 741 |
+
},
|
| 742 |
+
{
|
| 743 |
+
"epoch": 0.6,
|
| 744 |
+
"learning_rate": 4.214736842105263e-06,
|
| 745 |
+
"loss": 0.0125,
|
| 746 |
+
"step": 3000
|
| 747 |
+
},
|
| 748 |
+
{
|
| 749 |
+
"epoch": 0.6,
|
| 750 |
+
"eval_loss": 0.22999617457389832,
|
| 751 |
+
"eval_runtime": 2921.4658,
|
| 752 |
+
"eval_samples_per_second": 1.735,
|
| 753 |
+
"eval_steps_per_second": 0.109,
|
| 754 |
+
"eval_wer": 10.98269003332975,
|
| 755 |
+
"step": 3000
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 0.6,
|
| 759 |
+
"learning_rate": 4.162105263157895e-06,
|
| 760 |
+
"loss": 0.0089,
|
| 761 |
+
"step": 3025
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 0.61,
|
| 765 |
+
"learning_rate": 4.109473684210527e-06,
|
| 766 |
+
"loss": 0.0113,
|
| 767 |
+
"step": 3050
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 0.61,
|
| 771 |
+
"learning_rate": 4.056842105263158e-06,
|
| 772 |
+
"loss": 0.0097,
|
| 773 |
+
"step": 3075
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 0.62,
|
| 777 |
+
"learning_rate": 4.00421052631579e-06,
|
| 778 |
+
"loss": 0.0135,
|
| 779 |
+
"step": 3100
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"epoch": 0.62,
|
| 783 |
+
"learning_rate": 3.951578947368422e-06,
|
| 784 |
+
"loss": 0.0142,
|
| 785 |
+
"step": 3125
|
| 786 |
+
},
|
| 787 |
+
{
|
| 788 |
+
"epoch": 0.63,
|
| 789 |
+
"learning_rate": 3.898947368421052e-06,
|
| 790 |
+
"loss": 0.0132,
|
| 791 |
+
"step": 3150
|
| 792 |
+
},
|
| 793 |
+
{
|
| 794 |
+
"epoch": 0.64,
|
| 795 |
+
"learning_rate": 3.846315789473685e-06,
|
| 796 |
+
"loss": 0.0161,
|
| 797 |
+
"step": 3175
|
| 798 |
+
},
|
| 799 |
+
{
|
| 800 |
+
"epoch": 0.64,
|
| 801 |
+
"learning_rate": 3.793684210526316e-06,
|
| 802 |
+
"loss": 0.0141,
|
| 803 |
+
"step": 3200
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 0.65,
|
| 807 |
+
"learning_rate": 3.7410526315789473e-06,
|
| 808 |
+
"loss": 0.0189,
|
| 809 |
+
"step": 3225
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"epoch": 0.65,
|
| 813 |
+
"learning_rate": 3.6884210526315794e-06,
|
| 814 |
+
"loss": 0.0143,
|
| 815 |
+
"step": 3250
|
| 816 |
+
},
|
| 817 |
+
{
|
| 818 |
+
"epoch": 0.66,
|
| 819 |
+
"learning_rate": 3.635789473684211e-06,
|
| 820 |
+
"loss": 0.0165,
|
| 821 |
+
"step": 3275
|
| 822 |
+
},
|
| 823 |
+
{
|
| 824 |
+
"epoch": 0.66,
|
| 825 |
+
"learning_rate": 3.5831578947368422e-06,
|
| 826 |
+
"loss": 0.0104,
|
| 827 |
+
"step": 3300
|
| 828 |
+
},
|
| 829 |
+
{
|
| 830 |
+
"epoch": 0.67,
|
| 831 |
+
"learning_rate": 3.5305263157894743e-06,
|
| 832 |
+
"loss": 0.011,
|
| 833 |
+
"step": 3325
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 0.67,
|
| 837 |
+
"learning_rate": 3.4778947368421055e-06,
|
| 838 |
+
"loss": 0.0146,
|
| 839 |
+
"step": 3350
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 0.68,
|
| 843 |
+
"learning_rate": 3.425263157894737e-06,
|
| 844 |
+
"loss": 0.0196,
|
| 845 |
+
"step": 3375
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 0.68,
|
| 849 |
+
"learning_rate": 3.3726315789473683e-06,
|
| 850 |
+
"loss": 0.0139,
|
| 851 |
+
"step": 3400
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 0.69,
|
| 855 |
+
"learning_rate": 3.3200000000000004e-06,
|
| 856 |
+
"loss": 0.0127,
|
| 857 |
+
"step": 3425
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 0.69,
|
| 861 |
+
"learning_rate": 3.267368421052632e-06,
|
| 862 |
+
"loss": 0.0124,
|
| 863 |
+
"step": 3450
|
| 864 |
+
},
|
| 865 |
+
{
|
| 866 |
+
"epoch": 0.69,
|
| 867 |
+
"learning_rate": 3.2147368421052633e-06,
|
| 868 |
+
"loss": 0.0112,
|
| 869 |
+
"step": 3475
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 0.7,
|
| 873 |
+
"learning_rate": 3.1621052631578953e-06,
|
| 874 |
+
"loss": 0.0149,
|
| 875 |
+
"step": 3500
|
| 876 |
+
},
|
| 877 |
+
{
|
| 878 |
+
"epoch": 0.7,
|
| 879 |
+
"learning_rate": 3.1094736842105265e-06,
|
| 880 |
+
"loss": 0.0155,
|
| 881 |
+
"step": 3525
|
| 882 |
+
},
|
| 883 |
+
{
|
| 884 |
+
"epoch": 0.71,
|
| 885 |
+
"learning_rate": 3.056842105263158e-06,
|
| 886 |
+
"loss": 0.0215,
|
| 887 |
+
"step": 3550
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 0.71,
|
| 891 |
+
"learning_rate": 3.0042105263157894e-06,
|
| 892 |
+
"loss": 0.0171,
|
| 893 |
+
"step": 3575
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"epoch": 0.72,
|
| 897 |
+
"learning_rate": 2.9515789473684214e-06,
|
| 898 |
+
"loss": 0.0205,
|
| 899 |
+
"step": 3600
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 0.72,
|
| 903 |
+
"learning_rate": 2.8989473684210526e-06,
|
| 904 |
+
"loss": 0.0156,
|
| 905 |
+
"step": 3625
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 0.73,
|
| 909 |
+
"learning_rate": 2.8463157894736843e-06,
|
| 910 |
+
"loss": 0.0151,
|
| 911 |
+
"step": 3650
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 0.73,
|
| 915 |
+
"learning_rate": 2.7936842105263163e-06,
|
| 916 |
+
"loss": 0.0155,
|
| 917 |
+
"step": 3675
|
| 918 |
+
},
|
| 919 |
+
{
|
| 920 |
+
"epoch": 0.74,
|
| 921 |
+
"learning_rate": 2.7410526315789476e-06,
|
| 922 |
+
"loss": 0.0109,
|
| 923 |
+
"step": 3700
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"epoch": 0.74,
|
| 927 |
+
"learning_rate": 2.688421052631579e-06,
|
| 928 |
+
"loss": 0.0183,
|
| 929 |
+
"step": 3725
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 0.75,
|
| 933 |
+
"learning_rate": 2.635789473684211e-06,
|
| 934 |
+
"loss": 0.0125,
|
| 935 |
+
"step": 3750
|
| 936 |
+
},
|
| 937 |
+
{
|
| 938 |
+
"epoch": 0.76,
|
| 939 |
+
"learning_rate": 2.5831578947368425e-06,
|
| 940 |
+
"loss": 0.0108,
|
| 941 |
+
"step": 3775
|
| 942 |
+
},
|
| 943 |
+
{
|
| 944 |
+
"epoch": 0.76,
|
| 945 |
+
"learning_rate": 2.5305263157894737e-06,
|
| 946 |
+
"loss": 0.0141,
|
| 947 |
+
"step": 3800
|
| 948 |
+
},
|
| 949 |
+
{
|
| 950 |
+
"epoch": 0.77,
|
| 951 |
+
"learning_rate": 2.4778947368421053e-06,
|
| 952 |
+
"loss": 0.012,
|
| 953 |
+
"step": 3825
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"epoch": 0.77,
|
| 957 |
+
"learning_rate": 2.425263157894737e-06,
|
| 958 |
+
"loss": 0.0104,
|
| 959 |
+
"step": 3850
|
| 960 |
+
},
|
| 961 |
+
{
|
| 962 |
+
"epoch": 0.78,
|
| 963 |
+
"learning_rate": 2.3726315789473686e-06,
|
| 964 |
+
"loss": 0.0099,
|
| 965 |
+
"step": 3875
|
| 966 |
+
},
|
| 967 |
+
{
|
| 968 |
+
"epoch": 0.78,
|
| 969 |
+
"learning_rate": 2.3200000000000002e-06,
|
| 970 |
+
"loss": 0.0196,
|
| 971 |
+
"step": 3900
|
| 972 |
+
},
|
| 973 |
+
{
|
| 974 |
+
"epoch": 0.79,
|
| 975 |
+
"learning_rate": 2.267368421052632e-06,
|
| 976 |
+
"loss": 0.0201,
|
| 977 |
+
"step": 3925
|
| 978 |
+
},
|
| 979 |
+
{
|
| 980 |
+
"epoch": 0.79,
|
| 981 |
+
"learning_rate": 2.2147368421052635e-06,
|
| 982 |
+
"loss": 0.0127,
|
| 983 |
+
"step": 3950
|
| 984 |
+
},
|
| 985 |
+
{
|
| 986 |
+
"epoch": 0.8,
|
| 987 |
+
"learning_rate": 2.1621052631578947e-06,
|
| 988 |
+
"loss": 0.0166,
|
| 989 |
+
"step": 3975
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 0.8,
|
| 993 |
+
"learning_rate": 2.1094736842105264e-06,
|
| 994 |
+
"loss": 0.0168,
|
| 995 |
+
"step": 4000
|
| 996 |
+
},
|
| 997 |
+
{
|
| 998 |
+
"epoch": 0.8,
|
| 999 |
+
"eval_loss": 0.24368737637996674,
|
| 1000 |
+
"eval_runtime": 2933.3968,
|
| 1001 |
+
"eval_samples_per_second": 1.728,
|
| 1002 |
+
"eval_steps_per_second": 0.108,
|
| 1003 |
+
"eval_wer": 10.471992258896892,
|
| 1004 |
+
"step": 4000
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 0.81,
|
| 1008 |
+
"learning_rate": 2.056842105263158e-06,
|
| 1009 |
+
"loss": 0.0111,
|
| 1010 |
+
"step": 4025
|
| 1011 |
+
},
|
| 1012 |
+
{
|
| 1013 |
+
"epoch": 0.81,
|
| 1014 |
+
"learning_rate": 2.0042105263157896e-06,
|
| 1015 |
+
"loss": 0.0105,
|
| 1016 |
+
"step": 4050
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 0.81,
|
| 1020 |
+
"learning_rate": 1.9515789473684213e-06,
|
| 1021 |
+
"loss": 0.0109,
|
| 1022 |
+
"step": 4075
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 0.82,
|
| 1026 |
+
"learning_rate": 1.8989473684210527e-06,
|
| 1027 |
+
"loss": 0.0167,
|
| 1028 |
+
"step": 4100
|
| 1029 |
+
},
|
| 1030 |
+
{
|
| 1031 |
+
"epoch": 0.82,
|
| 1032 |
+
"learning_rate": 1.8463157894736843e-06,
|
| 1033 |
+
"loss": 0.0102,
|
| 1034 |
+
"step": 4125
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 0.83,
|
| 1038 |
+
"learning_rate": 1.7936842105263158e-06,
|
| 1039 |
+
"loss": 0.0099,
|
| 1040 |
+
"step": 4150
|
| 1041 |
+
},
|
| 1042 |
+
{
|
| 1043 |
+
"epoch": 0.83,
|
| 1044 |
+
"learning_rate": 1.7410526315789474e-06,
|
| 1045 |
+
"loss": 0.0157,
|
| 1046 |
+
"step": 4175
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 0.84,
|
| 1050 |
+
"learning_rate": 1.6884210526315792e-06,
|
| 1051 |
+
"loss": 0.0203,
|
| 1052 |
+
"step": 4200
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 0.84,
|
| 1056 |
+
"learning_rate": 1.6357894736842107e-06,
|
| 1057 |
+
"loss": 0.0222,
|
| 1058 |
+
"step": 4225
|
| 1059 |
+
},
|
| 1060 |
+
{
|
| 1061 |
+
"epoch": 0.85,
|
| 1062 |
+
"learning_rate": 1.5831578947368423e-06,
|
| 1063 |
+
"loss": 0.0159,
|
| 1064 |
+
"step": 4250
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 0.85,
|
| 1068 |
+
"learning_rate": 1.5305263157894737e-06,
|
| 1069 |
+
"loss": 0.0172,
|
| 1070 |
+
"step": 4275
|
| 1071 |
+
},
|
| 1072 |
+
{
|
| 1073 |
+
"epoch": 0.86,
|
| 1074 |
+
"learning_rate": 1.4778947368421054e-06,
|
| 1075 |
+
"loss": 0.0106,
|
| 1076 |
+
"step": 4300
|
| 1077 |
+
},
|
| 1078 |
+
{
|
| 1079 |
+
"epoch": 0.86,
|
| 1080 |
+
"learning_rate": 1.425263157894737e-06,
|
| 1081 |
+
"loss": 0.0077,
|
| 1082 |
+
"step": 4325
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"epoch": 0.87,
|
| 1086 |
+
"learning_rate": 1.3726315789473684e-06,
|
| 1087 |
+
"loss": 0.0095,
|
| 1088 |
+
"step": 4350
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"epoch": 0.88,
|
| 1092 |
+
"learning_rate": 1.32e-06,
|
| 1093 |
+
"loss": 0.0073,
|
| 1094 |
+
"step": 4375
|
| 1095 |
+
},
|
| 1096 |
+
{
|
| 1097 |
+
"epoch": 0.88,
|
| 1098 |
+
"learning_rate": 1.2673684210526315e-06,
|
| 1099 |
+
"loss": 0.0064,
|
| 1100 |
+
"step": 4400
|
| 1101 |
+
},
|
| 1102 |
+
{
|
| 1103 |
+
"epoch": 0.89,
|
| 1104 |
+
"learning_rate": 1.2147368421052633e-06,
|
| 1105 |
+
"loss": 0.0088,
|
| 1106 |
+
"step": 4425
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 0.89,
|
| 1110 |
+
"learning_rate": 1.1621052631578948e-06,
|
| 1111 |
+
"loss": 0.0085,
|
| 1112 |
+
"step": 4450
|
| 1113 |
+
},
|
| 1114 |
+
{
|
| 1115 |
+
"epoch": 0.9,
|
| 1116 |
+
"learning_rate": 1.1094736842105264e-06,
|
| 1117 |
+
"loss": 0.0111,
|
| 1118 |
+
"step": 4475
|
| 1119 |
+
},
|
| 1120 |
+
{
|
| 1121 |
+
"epoch": 0.9,
|
| 1122 |
+
"learning_rate": 1.0568421052631578e-06,
|
| 1123 |
+
"loss": 0.0072,
|
| 1124 |
+
"step": 4500
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 0.91,
|
| 1128 |
+
"learning_rate": 1.0042105263157897e-06,
|
| 1129 |
+
"loss": 0.0175,
|
| 1130 |
+
"step": 4525
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"epoch": 0.91,
|
| 1134 |
+
"learning_rate": 9.515789473684212e-07,
|
| 1135 |
+
"loss": 0.0087,
|
| 1136 |
+
"step": 4550
|
| 1137 |
+
},
|
| 1138 |
+
{
|
| 1139 |
+
"epoch": 0.92,
|
| 1140 |
+
"learning_rate": 8.989473684210527e-07,
|
| 1141 |
+
"loss": 0.009,
|
| 1142 |
+
"step": 4575
|
| 1143 |
+
},
|
| 1144 |
+
{
|
| 1145 |
+
"epoch": 0.92,
|
| 1146 |
+
"learning_rate": 8.463157894736843e-07,
|
| 1147 |
+
"loss": 0.0103,
|
| 1148 |
+
"step": 4600
|
| 1149 |
+
},
|
| 1150 |
+
{
|
| 1151 |
+
"epoch": 0.93,
|
| 1152 |
+
"learning_rate": 7.936842105263158e-07,
|
| 1153 |
+
"loss": 0.0091,
|
| 1154 |
+
"step": 4625
|
| 1155 |
+
},
|
| 1156 |
+
{
|
| 1157 |
+
"epoch": 0.93,
|
| 1158 |
+
"learning_rate": 7.410526315789475e-07,
|
| 1159 |
+
"loss": 0.012,
|
| 1160 |
+
"step": 4650
|
| 1161 |
+
},
|
| 1162 |
+
{
|
| 1163 |
+
"epoch": 0.94,
|
| 1164 |
+
"learning_rate": 6.884210526315791e-07,
|
| 1165 |
+
"loss": 0.0113,
|
| 1166 |
+
"step": 4675
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"epoch": 0.94,
|
| 1170 |
+
"learning_rate": 6.357894736842106e-07,
|
| 1171 |
+
"loss": 0.0108,
|
| 1172 |
+
"step": 4700
|
| 1173 |
+
},
|
| 1174 |
+
{
|
| 1175 |
+
"epoch": 0.94,
|
| 1176 |
+
"learning_rate": 5.831578947368421e-07,
|
| 1177 |
+
"loss": 0.0094,
|
| 1178 |
+
"step": 4725
|
| 1179 |
+
},
|
| 1180 |
+
{
|
| 1181 |
+
"epoch": 0.95,
|
| 1182 |
+
"learning_rate": 5.305263157894737e-07,
|
| 1183 |
+
"loss": 0.0072,
|
| 1184 |
+
"step": 4750
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"epoch": 0.95,
|
| 1188 |
+
"learning_rate": 4.778947368421053e-07,
|
| 1189 |
+
"loss": 0.0078,
|
| 1190 |
+
"step": 4775
|
| 1191 |
+
},
|
| 1192 |
+
{
|
| 1193 |
+
"epoch": 0.96,
|
| 1194 |
+
"learning_rate": 4.2526315789473684e-07,
|
| 1195 |
+
"loss": 0.0065,
|
| 1196 |
+
"step": 4800
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 0.96,
|
| 1200 |
+
"learning_rate": 3.726315789473685e-07,
|
| 1201 |
+
"loss": 0.0068,
|
| 1202 |
+
"step": 4825
|
| 1203 |
+
},
|
| 1204 |
+
{
|
| 1205 |
+
"epoch": 0.97,
|
| 1206 |
+
"learning_rate": 3.2e-07,
|
| 1207 |
+
"loss": 0.0105,
|
| 1208 |
+
"step": 4850
|
| 1209 |
+
},
|
| 1210 |
+
{
|
| 1211 |
+
"epoch": 0.97,
|
| 1212 |
+
"learning_rate": 2.6736842105263164e-07,
|
| 1213 |
+
"loss": 0.0054,
|
| 1214 |
+
"step": 4875
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"epoch": 0.98,
|
| 1218 |
+
"learning_rate": 2.1473684210526317e-07,
|
| 1219 |
+
"loss": 0.0077,
|
| 1220 |
+
"step": 4900
|
| 1221 |
+
},
|
| 1222 |
+
{
|
| 1223 |
+
"epoch": 0.98,
|
| 1224 |
+
"learning_rate": 1.6210526315789476e-07,
|
| 1225 |
+
"loss": 0.0085,
|
| 1226 |
+
"step": 4925
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"epoch": 0.99,
|
| 1230 |
+
"learning_rate": 1.0947368421052632e-07,
|
| 1231 |
+
"loss": 0.0092,
|
| 1232 |
+
"step": 4950
|
| 1233 |
+
},
|
| 1234 |
+
{
|
| 1235 |
+
"epoch": 0.99,
|
| 1236 |
+
"learning_rate": 5.68421052631579e-08,
|
| 1237 |
+
"loss": 0.0077,
|
| 1238 |
+
"step": 4975
|
| 1239 |
+
},
|
| 1240 |
+
{
|
| 1241 |
+
"epoch": 1.0,
|
| 1242 |
+
"learning_rate": 4.210526315789474e-09,
|
| 1243 |
+
"loss": 0.0071,
|
| 1244 |
+
"step": 5000
|
| 1245 |
+
},
|
| 1246 |
+
{
|
| 1247 |
+
"epoch": 1.0,
|
| 1248 |
+
"eval_loss": 0.24834245443344116,
|
| 1249 |
+
"eval_runtime": 2917.338,
|
| 1250 |
+
"eval_samples_per_second": 1.738,
|
| 1251 |
+
"eval_steps_per_second": 0.109,
|
| 1252 |
+
"eval_wer": 9.891409525857435,
|
| 1253 |
+
"step": 5000
|
| 1254 |
+
},
|
| 1255 |
+
{
|
| 1256 |
+
"epoch": 1.0,
|
| 1257 |
+
"step": 5000,
|
| 1258 |
+
"total_flos": 1.632967852032e+20,
|
| 1259 |
+
"train_loss": 0.025400285175442697,
|
| 1260 |
+
"train_runtime": 51804.3597,
|
| 1261 |
+
"train_samples_per_second": 3.089,
|
| 1262 |
+
"train_steps_per_second": 0.097
|
| 1263 |
}
|
| 1264 |
],
|
| 1265 |
+
"max_steps": 5000,
|
| 1266 |
"num_train_epochs": 9223372036854775807,
|
| 1267 |
+
"total_flos": 1.632967852032e+20,
|
| 1268 |
"trial_name": null,
|
| 1269 |
"trial_params": null
|
| 1270 |
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 3579
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e8a92c5600d30cca742b5eb96009172353c46564d5f8ae2f5babd1d6db76efb3
|
| 3 |
size 3579
|