Upload full trainer checkpoint step=1521 (profile=llama3_2_1b_base_grpo, run_type=grpo)
Browse files- meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/adapter_config.json +4 -4
- meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/adapter_model.safetensors +1 -1
- meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/optimizer.pt +1 -1
- meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/rng_state.pth +1 -1
- meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/scheduler.pt +1 -1
- meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/trainer_state.json +594 -594
- meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/training_args.bin +1 -1
- meta-llama__llama-3.2-1b/grpo/run_manifest.json +11 -5
meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/adapter_config.json
CHANGED
|
@@ -33,13 +33,13 @@
|
|
| 33 |
"rank_pattern": {},
|
| 34 |
"revision": null,
|
| 35 |
"target_modules": [
|
|
|
|
|
|
|
|
|
|
| 36 |
"down_proj",
|
| 37 |
"o_proj",
|
| 38 |
-
"q_proj",
|
| 39 |
"gate_proj",
|
| 40 |
-
"
|
| 41 |
-
"up_proj",
|
| 42 |
-
"v_proj"
|
| 43 |
],
|
| 44 |
"target_parameters": null,
|
| 45 |
"task_type": "CAUSAL_LM",
|
|
|
|
| 33 |
"rank_pattern": {},
|
| 34 |
"revision": null,
|
| 35 |
"target_modules": [
|
| 36 |
+
"up_proj",
|
| 37 |
+
"k_proj",
|
| 38 |
+
"v_proj",
|
| 39 |
"down_proj",
|
| 40 |
"o_proj",
|
|
|
|
| 41 |
"gate_proj",
|
| 42 |
+
"q_proj"
|
|
|
|
|
|
|
| 43 |
],
|
| 44 |
"target_parameters": null,
|
| 45 |
"task_type": "CAUSAL_LM",
|
meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/adapter_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 45118424
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:525397e0ef3aa15a7fcebf66df0a5900bdbf57328545211f70ccbe51da3755fe
|
| 3 |
size 45118424
|
meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/optimizer.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 23162187
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a7bebb0a43f2b28a2d033fc7b2e0996db510e45e86b7c962f42f3273e1b516e
|
| 3 |
size 23162187
|
meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/rng_state.pth
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14645
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:00f874dd9d7ae4223b7072142300bf2ea2e985a109e5fc02ebebfca0a03f44ac
|
| 3 |
size 14645
|
meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/scheduler.pt
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 1465
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6fe0516817ef50b8c17ec21995e3135256f60beca23aec092a3ecc2038dc33d6
|
| 3 |
size 1465
|
meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/trainer_state.json
CHANGED
|
@@ -2,7 +2,7 @@
|
|
| 2 |
"best_global_step": null,
|
| 3 |
"best_metric": null,
|
| 4 |
"best_model_checkpoint": null,
|
| 5 |
-
"epoch": 0.
|
| 6 |
"eval_steps": 500,
|
| 7 |
"global_step": 1521,
|
| 8 |
"is_hyper_param_search": false,
|
|
@@ -15,25 +15,25 @@
|
|
| 15 |
"clip_ratio/low_mean": 0.0,
|
| 16 |
"clip_ratio/low_min": 0.0,
|
| 17 |
"clip_ratio/region_mean": 0.0,
|
| 18 |
-
"completion_length": 65.
|
| 19 |
"completions/clipped_ratio": 0.0,
|
| 20 |
-
"completions/max_length":
|
| 21 |
-
"completions/max_terminated_length":
|
| 22 |
-
"completions/mean_length":
|
| 23 |
-
"completions/mean_terminated_length":
|
| 24 |
-
"completions/min_length":
|
| 25 |
-
"completions/min_terminated_length":
|
| 26 |
-
"epoch": 8.
|
| 27 |
-
"frac_reward_zero_std": 0.
|
| 28 |
-
"grad_norm": 0.
|
| 29 |
"kl": 0.0,
|
| 30 |
-
"learning_rate":
|
| 31 |
-
"loss": -0.
|
| 32 |
-
"num_tokens":
|
| 33 |
-
"reward":
|
| 34 |
-
"reward_std": 0.
|
| 35 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 36 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 37 |
"step": 50
|
| 38 |
},
|
| 39 |
{
|
|
@@ -42,52 +42,52 @@
|
|
| 42 |
"clip_ratio/low_mean": 0.0,
|
| 43 |
"clip_ratio/low_min": 0.0,
|
| 44 |
"clip_ratio/region_mean": 0.0,
|
| 45 |
-
"completion_length":
|
| 46 |
"completions/clipped_ratio": 0.0,
|
| 47 |
-
"completions/max_length":
|
| 48 |
-
"completions/max_terminated_length":
|
| 49 |
-
"completions/mean_length":
|
| 50 |
-
"completions/mean_terminated_length":
|
| 51 |
-
"completions/min_length":
|
| 52 |
-
"completions/min_terminated_length":
|
| 53 |
-
"epoch": 0.
|
| 54 |
-
"frac_reward_zero_std": 0.
|
| 55 |
"grad_norm": 0.0,
|
| 56 |
"kl": 0.0,
|
| 57 |
-
"learning_rate":
|
| 58 |
-
"loss": 0.
|
| 59 |
-
"num_tokens":
|
| 60 |
-
"reward":
|
| 61 |
-
"reward_std": 0.
|
| 62 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 63 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 64 |
"step": 100
|
| 65 |
},
|
| 66 |
{
|
| 67 |
"clip_ratio/high_max": 0.0,
|
| 68 |
"clip_ratio/high_mean": 0.0,
|
| 69 |
-
"clip_ratio/low_mean": 0.
|
| 70 |
-
"clip_ratio/low_min": 0.
|
| 71 |
-
"clip_ratio/region_mean": 0.
|
| 72 |
-
"completion_length":
|
| 73 |
"completions/clipped_ratio": 0.0,
|
| 74 |
-
"completions/max_length":
|
| 75 |
-
"completions/max_terminated_length":
|
| 76 |
-
"completions/mean_length":
|
| 77 |
-
"completions/mean_terminated_length":
|
| 78 |
-
"completions/min_length":
|
| 79 |
-
"completions/min_terminated_length":
|
| 80 |
-
"epoch": 0.
|
| 81 |
-
"frac_reward_zero_std": 0.
|
| 82 |
-
"grad_norm": 0.
|
| 83 |
"kl": 0.0,
|
| 84 |
-
"learning_rate":
|
| 85 |
-
"loss": 0.
|
| 86 |
-
"num_tokens":
|
| 87 |
-
"reward":
|
| 88 |
-
"reward_std": 0.
|
| 89 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 90 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 91 |
"step": 150
|
| 92 |
},
|
| 93 |
{
|
|
@@ -96,52 +96,52 @@
|
|
| 96 |
"clip_ratio/low_mean": 0.0,
|
| 97 |
"clip_ratio/low_min": 0.0,
|
| 98 |
"clip_ratio/region_mean": 0.0,
|
| 99 |
-
"completion_length":
|
| 100 |
"completions/clipped_ratio": 0.0,
|
| 101 |
-
"completions/max_length":
|
| 102 |
-
"completions/max_terminated_length":
|
| 103 |
-
"completions/mean_length":
|
| 104 |
-
"completions/mean_terminated_length":
|
| 105 |
-
"completions/min_length":
|
| 106 |
-
"completions/min_terminated_length":
|
| 107 |
-
"epoch": 0.
|
| 108 |
-
"frac_reward_zero_std": 0.
|
| 109 |
-
"grad_norm": 0.
|
| 110 |
"kl": 0.0,
|
| 111 |
-
"learning_rate":
|
| 112 |
-
"loss": -0.
|
| 113 |
-
"num_tokens":
|
| 114 |
-
"reward":
|
| 115 |
-
"reward_std": 0.
|
| 116 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 117 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 118 |
"step": 200
|
| 119 |
},
|
| 120 |
{
|
| 121 |
"clip_ratio/high_max": 0.0,
|
| 122 |
"clip_ratio/high_mean": 0.0,
|
| 123 |
-
"clip_ratio/low_mean": 0.
|
| 124 |
-
"clip_ratio/low_min": 0.
|
| 125 |
-
"clip_ratio/region_mean": 0.
|
| 126 |
-
"completion_length":
|
| 127 |
"completions/clipped_ratio": 0.0,
|
| 128 |
-
"completions/max_length":
|
| 129 |
-
"completions/max_terminated_length":
|
| 130 |
-
"completions/mean_length":
|
| 131 |
-
"completions/mean_terminated_length":
|
| 132 |
-
"completions/min_length":
|
| 133 |
-
"completions/min_terminated_length":
|
| 134 |
-
"epoch": 0.
|
| 135 |
-
"frac_reward_zero_std": 0.
|
| 136 |
-
"grad_norm": 0.
|
| 137 |
"kl": 0.0,
|
| 138 |
-
"learning_rate":
|
| 139 |
-
"loss": 0.
|
| 140 |
-
"num_tokens":
|
| 141 |
-
"reward":
|
| 142 |
-
"reward_std": 0.
|
| 143 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 144 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 145 |
"step": 250
|
| 146 |
},
|
| 147 |
{
|
|
@@ -150,106 +150,106 @@
|
|
| 150 |
"clip_ratio/low_mean": 0.0,
|
| 151 |
"clip_ratio/low_min": 0.0,
|
| 152 |
"clip_ratio/region_mean": 0.0,
|
| 153 |
-
"completion_length":
|
| 154 |
"completions/clipped_ratio": 0.0,
|
| 155 |
-
"completions/max_length":
|
| 156 |
-
"completions/max_terminated_length":
|
| 157 |
-
"completions/mean_length":
|
| 158 |
-
"completions/mean_terminated_length":
|
| 159 |
-
"completions/min_length":
|
| 160 |
-
"completions/min_terminated_length":
|
| 161 |
-
"epoch": 0.
|
| 162 |
-
"frac_reward_zero_std": 0.
|
| 163 |
-
"grad_norm": 0.
|
| 164 |
"kl": 0.0,
|
| 165 |
-
"learning_rate":
|
| 166 |
-
"loss":
|
| 167 |
-
"num_tokens":
|
| 168 |
-
"reward":
|
| 169 |
-
"reward_std": 0.
|
| 170 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 171 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 172 |
"step": 300
|
| 173 |
},
|
| 174 |
{
|
| 175 |
"clip_ratio/high_max": 0.0,
|
| 176 |
"clip_ratio/high_mean": 0.0,
|
| 177 |
-
"clip_ratio/low_mean":
|
| 178 |
-
"clip_ratio/low_min":
|
| 179 |
-
"clip_ratio/region_mean":
|
| 180 |
-
"completion_length":
|
| 181 |
"completions/clipped_ratio": 0.0,
|
| 182 |
-
"completions/max_length":
|
| 183 |
-
"completions/max_terminated_length":
|
| 184 |
-
"completions/mean_length": 61.
|
| 185 |
-
"completions/mean_terminated_length": 61.
|
| 186 |
-
"completions/min_length":
|
| 187 |
-
"completions/min_terminated_length":
|
| 188 |
-
"epoch": 0.
|
| 189 |
-
"frac_reward_zero_std": 0.
|
| 190 |
-
"grad_norm": 0.
|
| 191 |
"kl": 0.0,
|
| 192 |
-
"learning_rate":
|
| 193 |
-
"loss": -
|
| 194 |
-
"num_tokens":
|
| 195 |
-
"reward":
|
| 196 |
-
"reward_std": 0.
|
| 197 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 198 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 199 |
"step": 350
|
| 200 |
},
|
| 201 |
{
|
| 202 |
"clip_ratio/high_max": 0.0,
|
| 203 |
"clip_ratio/high_mean": 0.0,
|
| 204 |
-
"clip_ratio/low_mean":
|
| 205 |
-
"clip_ratio/low_min":
|
| 206 |
-
"clip_ratio/region_mean":
|
| 207 |
-
"completion_length":
|
| 208 |
"completions/clipped_ratio": 0.0,
|
| 209 |
-
"completions/max_length":
|
| 210 |
-
"completions/max_terminated_length":
|
| 211 |
-
"completions/mean_length":
|
| 212 |
-
"completions/mean_terminated_length":
|
| 213 |
-
"completions/min_length":
|
| 214 |
-
"completions/min_terminated_length":
|
| 215 |
-
"epoch": 0.
|
| 216 |
-
"frac_reward_zero_std": 0.
|
| 217 |
-
"grad_norm": 0.
|
| 218 |
"kl": 0.0,
|
| 219 |
-
"learning_rate":
|
| 220 |
-
"loss": 0.
|
| 221 |
-
"num_tokens":
|
| 222 |
-
"reward":
|
| 223 |
-
"reward_std": 0.
|
| 224 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 225 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 226 |
"step": 400
|
| 227 |
},
|
| 228 |
{
|
| 229 |
"clip_ratio/high_max": 0.0,
|
| 230 |
"clip_ratio/high_mean": 0.0,
|
| 231 |
-
"clip_ratio/low_mean":
|
| 232 |
-
"clip_ratio/low_min":
|
| 233 |
-
"clip_ratio/region_mean":
|
| 234 |
-
"completion_length":
|
| 235 |
"completions/clipped_ratio": 0.0,
|
| 236 |
-
"completions/max_length":
|
| 237 |
-
"completions/max_terminated_length":
|
| 238 |
-
"completions/mean_length":
|
| 239 |
-
"completions/mean_terminated_length":
|
| 240 |
-
"completions/min_length":
|
| 241 |
-
"completions/min_terminated_length":
|
| 242 |
-
"epoch": 0.
|
| 243 |
-
"frac_reward_zero_std": 0.
|
| 244 |
-
"grad_norm": 0.
|
| 245 |
"kl": 0.0,
|
| 246 |
-
"learning_rate":
|
| 247 |
-
"loss": -0.
|
| 248 |
-
"num_tokens":
|
| 249 |
-
"reward": 1.
|
| 250 |
-
"reward_std": 0.
|
| 251 |
-
"rewards/TranscriptCorrectionGrpoReward/mean": 1.
|
| 252 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 253 |
"step": 450
|
| 254 |
},
|
| 255 |
{
|
|
@@ -258,115 +258,115 @@
|
|
| 258 |
"clip_ratio/low_mean": 0.0,
|
| 259 |
"clip_ratio/low_min": 0.0,
|
| 260 |
"clip_ratio/region_mean": 0.0,
|
| 261 |
-
"completion_length":
|
| 262 |
"completions/clipped_ratio": 0.0,
|
| 263 |
-
"completions/max_length":
|
| 264 |
-
"completions/max_terminated_length":
|
| 265 |
-
"completions/mean_length":
|
| 266 |
-
"completions/mean_terminated_length":
|
| 267 |
-
"completions/min_length":
|
| 268 |
-
"completions/min_terminated_length":
|
| 269 |
-
"epoch": 0.
|
| 270 |
-
"frac_reward_zero_std": 0.
|
| 271 |
-
"grad_norm": 0.
|
| 272 |
"kl": 0.0,
|
| 273 |
-
"learning_rate":
|
| 274 |
-
"loss": 0.
|
| 275 |
-
"num_tokens":
|
| 276 |
-
"reward":
|
| 277 |
-
"reward_std": 0.
|
| 278 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 279 |
-
"rewards/TranscriptCorrectionGrpoReward/std": 0.
|
| 280 |
"step": 500
|
| 281 |
},
|
| 282 |
{
|
| 283 |
-
"eval_cer_subset": 0.
|
| 284 |
-
"eval_cer_subset_edit_distance":
|
| 285 |
"eval_cer_subset_groups": 250,
|
| 286 |
"eval_cer_subset_items": 250,
|
| 287 |
"eval_cer_subset_ref_chars": 61407,
|
| 288 |
-
"eval_loss": 0.
|
| 289 |
"step": 507
|
| 290 |
},
|
| 291 |
{
|
| 292 |
-
"clip_ratio/high_max":
|
| 293 |
-
"clip_ratio/high_mean":
|
| 294 |
-
"clip_ratio/low_mean":
|
| 295 |
-
"clip_ratio/low_min":
|
| 296 |
-
"clip_ratio/region_mean": 0.
|
| 297 |
-
"completion_length":
|
| 298 |
"completions/clipped_ratio": 0.0,
|
| 299 |
-
"completions/max_length":
|
| 300 |
-
"completions/max_terminated_length":
|
| 301 |
-
"completions/mean_length":
|
| 302 |
-
"completions/mean_terminated_length":
|
| 303 |
-
"completions/min_length":
|
| 304 |
-
"completions/min_terminated_length":
|
| 305 |
-
"epoch": 0.
|
| 306 |
-
"frac_reward_zero_std": 0.
|
| 307 |
-
"grad_norm": 0.
|
| 308 |
"kl": 0.0,
|
| 309 |
-
"learning_rate":
|
| 310 |
-
"loss":
|
| 311 |
-
"num_tokens":
|
| 312 |
-
"reward":
|
| 313 |
-
"reward_std": 0.
|
| 314 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 315 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 316 |
"step": 550
|
| 317 |
},
|
| 318 |
{
|
| 319 |
"clip_ratio/high_max": 0.0,
|
| 320 |
"clip_ratio/high_mean": 0.0,
|
| 321 |
-
"clip_ratio/low_mean":
|
| 322 |
-
"clip_ratio/low_min":
|
| 323 |
-
"clip_ratio/region_mean":
|
| 324 |
-
"completion_length":
|
| 325 |
"completions/clipped_ratio": 0.0,
|
| 326 |
-
"completions/max_length":
|
| 327 |
-
"completions/max_terminated_length":
|
| 328 |
-
"completions/mean_length":
|
| 329 |
-
"completions/mean_terminated_length":
|
| 330 |
-
"completions/min_length":
|
| 331 |
-
"completions/min_terminated_length":
|
| 332 |
-
"epoch": 0.
|
| 333 |
-
"frac_reward_zero_std": 0.
|
| 334 |
-
"grad_norm": 0.
|
| 335 |
"kl": 0.0,
|
| 336 |
-
"learning_rate":
|
| 337 |
-
"loss": 0.
|
| 338 |
-
"num_tokens":
|
| 339 |
-
"reward":
|
| 340 |
-
"reward_std": 0.
|
| 341 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 342 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 343 |
"step": 600
|
| 344 |
},
|
| 345 |
{
|
| 346 |
"clip_ratio/high_max": 0.0,
|
| 347 |
"clip_ratio/high_mean": 0.0,
|
| 348 |
-
"clip_ratio/low_mean":
|
| 349 |
-
"clip_ratio/low_min":
|
| 350 |
-
"clip_ratio/region_mean":
|
| 351 |
-
"completion_length": 71.
|
| 352 |
"completions/clipped_ratio": 0.0,
|
| 353 |
-
"completions/max_length":
|
| 354 |
-
"completions/max_terminated_length":
|
| 355 |
-
"completions/mean_length": 71.
|
| 356 |
-
"completions/mean_terminated_length": 71.
|
| 357 |
-
"completions/min_length":
|
| 358 |
-
"completions/min_terminated_length":
|
| 359 |
-
"epoch": 0.
|
| 360 |
-
"frac_reward_zero_std": 0.
|
| 361 |
-
"grad_norm": 0.
|
| 362 |
"kl": 0.0,
|
| 363 |
-
"learning_rate":
|
| 364 |
-
"loss": -0.
|
| 365 |
-
"num_tokens":
|
| 366 |
-
"reward": -
|
| 367 |
-
"reward_std": 0.
|
| 368 |
-
"rewards/TranscriptCorrectionGrpoReward/mean": -
|
| 369 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 370 |
"step": 650
|
| 371 |
},
|
| 372 |
{
|
|
@@ -375,304 +375,304 @@
|
|
| 375 |
"clip_ratio/low_mean": 0.0,
|
| 376 |
"clip_ratio/low_min": 0.0,
|
| 377 |
"clip_ratio/region_mean": 0.0,
|
| 378 |
-
"completion_length":
|
| 379 |
"completions/clipped_ratio": 0.0,
|
| 380 |
-
"completions/max_length":
|
| 381 |
-
"completions/max_terminated_length":
|
| 382 |
-
"completions/mean_length":
|
| 383 |
-
"completions/mean_terminated_length":
|
| 384 |
-
"completions/min_length":
|
| 385 |
-
"completions/min_terminated_length":
|
| 386 |
-
"epoch": 0.
|
| 387 |
-
"frac_reward_zero_std": 0.
|
| 388 |
-
"grad_norm": 0.
|
| 389 |
"kl": 0.0,
|
| 390 |
-
"learning_rate":
|
| 391 |
-
"loss": -0.
|
| 392 |
-
"num_tokens":
|
| 393 |
-
"reward":
|
| 394 |
-
"reward_std": 0.
|
| 395 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 396 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 397 |
"step": 700
|
| 398 |
},
|
| 399 |
{
|
| 400 |
-
"clip_ratio/high_max":
|
| 401 |
-
"clip_ratio/high_mean":
|
| 402 |
"clip_ratio/low_mean": 0.0,
|
| 403 |
"clip_ratio/low_min": 0.0,
|
| 404 |
-
"clip_ratio/region_mean":
|
| 405 |
-
"completion_length":
|
| 406 |
"completions/clipped_ratio": 0.0,
|
| 407 |
-
"completions/max_length":
|
| 408 |
-
"completions/max_terminated_length":
|
| 409 |
-
"completions/mean_length":
|
| 410 |
-
"completions/mean_terminated_length":
|
| 411 |
-
"completions/min_length":
|
| 412 |
-
"completions/min_terminated_length":
|
| 413 |
-
"epoch": 0.
|
| 414 |
-
"frac_reward_zero_std": 0.
|
| 415 |
-
"grad_norm": 0.
|
| 416 |
"kl": 0.0,
|
| 417 |
-
"learning_rate":
|
| 418 |
-
"loss": 0.
|
| 419 |
-
"num_tokens":
|
| 420 |
-
"reward": 2.
|
| 421 |
-
"reward_std": 0.
|
| 422 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 423 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 424 |
"step": 750
|
| 425 |
},
|
| 426 |
{
|
| 427 |
"clip_ratio/high_max": 0.0,
|
| 428 |
"clip_ratio/high_mean": 0.0,
|
| 429 |
-
"clip_ratio/low_mean": 0.
|
| 430 |
-
"clip_ratio/low_min": 0.
|
| 431 |
-
"clip_ratio/region_mean": 0.
|
| 432 |
-
"completion_length": 67.
|
| 433 |
"completions/clipped_ratio": 0.0,
|
| 434 |
-
"completions/max_length":
|
| 435 |
-
"completions/max_terminated_length":
|
| 436 |
-
"completions/mean_length": 67.
|
| 437 |
-
"completions/mean_terminated_length": 67.
|
| 438 |
-
"completions/min_length":
|
| 439 |
-
"completions/min_terminated_length":
|
| 440 |
-
"epoch": 0.
|
| 441 |
-
"frac_reward_zero_std": 0.
|
| 442 |
-
"grad_norm": 0.
|
| 443 |
"kl": 0.0,
|
| 444 |
-
"learning_rate":
|
| 445 |
-
"loss": -
|
| 446 |
-
"num_tokens":
|
| 447 |
-
"reward":
|
| 448 |
-
"reward_std": 0.
|
| 449 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 450 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 451 |
"step": 800
|
| 452 |
},
|
| 453 |
{
|
| 454 |
"clip_ratio/high_max": 0.0,
|
| 455 |
"clip_ratio/high_mean": 0.0,
|
| 456 |
-
"clip_ratio/low_mean":
|
| 457 |
-
"clip_ratio/low_min":
|
| 458 |
-
"clip_ratio/region_mean":
|
| 459 |
-
"completion_length":
|
| 460 |
"completions/clipped_ratio": 0.0,
|
| 461 |
-
"completions/max_length":
|
| 462 |
-
"completions/max_terminated_length":
|
| 463 |
-
"completions/mean_length": 75.
|
| 464 |
-
"completions/mean_terminated_length": 75.
|
| 465 |
-
"completions/min_length":
|
| 466 |
-
"completions/min_terminated_length":
|
| 467 |
-
"epoch": 0.
|
| 468 |
-
"frac_reward_zero_std": 0.
|
| 469 |
-
"grad_norm": 0.
|
| 470 |
"kl": 0.0,
|
| 471 |
-
"learning_rate":
|
| 472 |
-
"loss": -0.
|
| 473 |
-
"num_tokens":
|
| 474 |
-
"reward":
|
| 475 |
-
"reward_std": 0.
|
| 476 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 477 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 478 |
"step": 850
|
| 479 |
},
|
| 480 |
{
|
| 481 |
-
"clip_ratio/high_max":
|
| 482 |
-
"clip_ratio/high_mean":
|
| 483 |
-
"clip_ratio/low_mean":
|
| 484 |
-
"clip_ratio/low_min":
|
| 485 |
-
"clip_ratio/region_mean":
|
| 486 |
-
"completion_length":
|
| 487 |
"completions/clipped_ratio": 0.0,
|
| 488 |
-
"completions/max_length":
|
| 489 |
-
"completions/max_terminated_length":
|
| 490 |
-
"completions/mean_length":
|
| 491 |
-
"completions/mean_terminated_length":
|
| 492 |
-
"completions/min_length":
|
| 493 |
-
"completions/min_terminated_length":
|
| 494 |
-
"epoch": 0.
|
| 495 |
-
"frac_reward_zero_std": 0.
|
| 496 |
-
"grad_norm": 0.
|
| 497 |
"kl": 0.0,
|
| 498 |
-
"learning_rate":
|
| 499 |
-
"loss":
|
| 500 |
-
"num_tokens":
|
| 501 |
-
"reward":
|
| 502 |
-
"reward_std": 0.
|
| 503 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 504 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 505 |
"step": 900
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"clip_ratio/high_max": 0.0,
|
| 509 |
"clip_ratio/high_mean": 0.0,
|
| 510 |
-
"clip_ratio/low_mean":
|
| 511 |
-
"clip_ratio/low_min":
|
| 512 |
-
"clip_ratio/region_mean":
|
| 513 |
-
"completion_length": 69.
|
| 514 |
"completions/clipped_ratio": 0.0,
|
| 515 |
-
"completions/max_length":
|
| 516 |
-
"completions/max_terminated_length":
|
| 517 |
-
"completions/mean_length":
|
| 518 |
-
"completions/mean_terminated_length":
|
| 519 |
-
"completions/min_length":
|
| 520 |
-
"completions/min_terminated_length":
|
| 521 |
-
"epoch": 0.
|
| 522 |
-
"frac_reward_zero_std": 0.
|
| 523 |
-
"grad_norm": 0.
|
| 524 |
"kl": 0.0,
|
| 525 |
-
"learning_rate":
|
| 526 |
-
"loss":
|
| 527 |
-
"num_tokens":
|
| 528 |
-
"reward":
|
| 529 |
-
"reward_std": 0.
|
| 530 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 531 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 532 |
"step": 950
|
| 533 |
},
|
| 534 |
{
|
| 535 |
"clip_ratio/high_max": 0.0,
|
| 536 |
"clip_ratio/high_mean": 0.0,
|
| 537 |
-
"clip_ratio/low_mean": 0.
|
| 538 |
-
"clip_ratio/low_min": 0.
|
| 539 |
-
"clip_ratio/region_mean": 0.
|
| 540 |
-
"completion_length":
|
| 541 |
"completions/clipped_ratio": 0.0,
|
| 542 |
-
"completions/max_length":
|
| 543 |
-
"completions/max_terminated_length":
|
| 544 |
-
"completions/mean_length":
|
| 545 |
-
"completions/mean_terminated_length":
|
| 546 |
-
"completions/min_length":
|
| 547 |
-
"completions/min_terminated_length":
|
| 548 |
-
"epoch": 0.
|
| 549 |
-
"frac_reward_zero_std": 0.
|
| 550 |
-
"grad_norm":
|
| 551 |
"kl": 0.0,
|
| 552 |
-
"learning_rate":
|
| 553 |
-
"loss": 0.
|
| 554 |
-
"num_tokens":
|
| 555 |
-
"reward":
|
| 556 |
-
"reward_std": 0.
|
| 557 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 558 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 559 |
"step": 1000
|
| 560 |
},
|
| 561 |
{
|
| 562 |
-
"eval_cer_subset": 0.
|
| 563 |
-
"eval_cer_subset_edit_distance":
|
| 564 |
"eval_cer_subset_groups": 250,
|
| 565 |
"eval_cer_subset_items": 250,
|
| 566 |
"eval_cer_subset_ref_chars": 61407,
|
| 567 |
-
"eval_loss": 0.
|
| 568 |
"step": 1014
|
| 569 |
},
|
| 570 |
{
|
| 571 |
"clip_ratio/high_max": 0.0,
|
| 572 |
"clip_ratio/high_mean": 0.0,
|
| 573 |
-
"clip_ratio/low_mean":
|
| 574 |
-
"clip_ratio/low_min":
|
| 575 |
-
"clip_ratio/region_mean":
|
| 576 |
-
"completion_length":
|
| 577 |
"completions/clipped_ratio": 0.0,
|
| 578 |
-
"completions/max_length":
|
| 579 |
-
"completions/max_terminated_length":
|
| 580 |
-
"completions/mean_length":
|
| 581 |
-
"completions/mean_terminated_length":
|
| 582 |
-
"completions/min_length":
|
| 583 |
-
"completions/min_terminated_length":
|
| 584 |
-
"epoch": 0.
|
| 585 |
-
"frac_reward_zero_std": 0.
|
| 586 |
-
"grad_norm": 0.
|
| 587 |
"kl": 0.0,
|
| 588 |
-
"learning_rate":
|
| 589 |
-
"loss":
|
| 590 |
-
"num_tokens":
|
| 591 |
-
"reward":
|
| 592 |
-
"reward_std": 0.
|
| 593 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 594 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 595 |
"step": 1050
|
| 596 |
},
|
| 597 |
{
|
| 598 |
"clip_ratio/high_max": 0.0,
|
| 599 |
"clip_ratio/high_mean": 0.0,
|
| 600 |
-
"clip_ratio/low_mean":
|
| 601 |
-
"clip_ratio/low_min":
|
| 602 |
-
"clip_ratio/region_mean":
|
| 603 |
-
"completion_length": 62.
|
| 604 |
"completions/clipped_ratio": 0.0,
|
| 605 |
-
"completions/max_length":
|
| 606 |
-
"completions/max_terminated_length":
|
| 607 |
-
"completions/mean_length": 62.
|
| 608 |
-
"completions/mean_terminated_length": 62.
|
| 609 |
-
"completions/min_length":
|
| 610 |
-
"completions/min_terminated_length":
|
| 611 |
-
"epoch": 0.
|
| 612 |
-
"frac_reward_zero_std": 0.
|
| 613 |
-
"grad_norm": 0.
|
| 614 |
"kl": 0.0,
|
| 615 |
-
"learning_rate":
|
| 616 |
-
"loss": 0.
|
| 617 |
-
"num_tokens":
|
| 618 |
-
"reward":
|
| 619 |
-
"reward_std": 0.
|
| 620 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 621 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 622 |
"step": 1100
|
| 623 |
},
|
| 624 |
{
|
| 625 |
"clip_ratio/high_max": 0.0,
|
| 626 |
"clip_ratio/high_mean": 0.0,
|
| 627 |
-
"clip_ratio/low_mean":
|
| 628 |
-
"clip_ratio/low_min":
|
| 629 |
-
"clip_ratio/region_mean":
|
| 630 |
-
"completion_length": 67.
|
| 631 |
"completions/clipped_ratio": 0.0,
|
| 632 |
-
"completions/max_length":
|
| 633 |
-
"completions/max_terminated_length":
|
| 634 |
-
"completions/mean_length": 67.
|
| 635 |
-
"completions/mean_terminated_length": 67.
|
| 636 |
-
"completions/min_length":
|
| 637 |
-
"completions/min_terminated_length":
|
| 638 |
-
"epoch": 0.
|
| 639 |
-
"frac_reward_zero_std": 0.
|
| 640 |
-
"grad_norm": 0.
|
| 641 |
"kl": 0.0,
|
| 642 |
-
"learning_rate":
|
| 643 |
-
"loss": -0.
|
| 644 |
-
"num_tokens":
|
| 645 |
-
"reward":
|
| 646 |
-
"reward_std": 0.
|
| 647 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 648 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 649 |
"step": 1150
|
| 650 |
},
|
| 651 |
{
|
| 652 |
-
"clip_ratio/high_max":
|
| 653 |
-
"clip_ratio/high_mean":
|
| 654 |
-
"clip_ratio/low_mean":
|
| 655 |
-
"clip_ratio/low_min":
|
| 656 |
-
"clip_ratio/region_mean":
|
| 657 |
-
"completion_length":
|
| 658 |
"completions/clipped_ratio": 0.0,
|
| 659 |
-
"completions/max_length":
|
| 660 |
-
"completions/max_terminated_length":
|
| 661 |
-
"completions/mean_length": 67.
|
| 662 |
-
"completions/mean_terminated_length": 67.
|
| 663 |
-
"completions/min_length":
|
| 664 |
-
"completions/min_terminated_length":
|
| 665 |
-
"epoch": 0.
|
| 666 |
-
"frac_reward_zero_std": 0.
|
| 667 |
-
"grad_norm": 0.
|
| 668 |
"kl": 0.0,
|
| 669 |
-
"learning_rate":
|
| 670 |
-
"loss":
|
| 671 |
-
"num_tokens":
|
| 672 |
-
"reward":
|
| 673 |
-
"reward_std": 0.
|
| 674 |
-
"rewards/TranscriptCorrectionGrpoReward/mean": 1.
|
| 675 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 676 |
"step": 1200
|
| 677 |
},
|
| 678 |
{
|
|
@@ -681,175 +681,175 @@
|
|
| 681 |
"clip_ratio/low_mean": 0.0,
|
| 682 |
"clip_ratio/low_min": 0.0,
|
| 683 |
"clip_ratio/region_mean": 0.0,
|
| 684 |
-
"completion_length":
|
| 685 |
"completions/clipped_ratio": 0.0,
|
| 686 |
-
"completions/max_length":
|
| 687 |
-
"completions/max_terminated_length":
|
| 688 |
-
"completions/mean_length":
|
| 689 |
-
"completions/mean_terminated_length":
|
| 690 |
-
"completions/min_length":
|
| 691 |
-
"completions/min_terminated_length":
|
| 692 |
-
"epoch": 0.
|
| 693 |
-
"frac_reward_zero_std": 0.
|
| 694 |
-
"grad_norm": 0.
|
| 695 |
"kl": 0.0,
|
| 696 |
-
"learning_rate":
|
| 697 |
-
"loss": -0.
|
| 698 |
-
"num_tokens":
|
| 699 |
-
"reward":
|
| 700 |
-
"reward_std": 0.
|
| 701 |
-
"rewards/TranscriptCorrectionGrpoReward/mean": 2.
|
| 702 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 703 |
"step": 1250
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"clip_ratio/high_max": 0.0,
|
| 707 |
"clip_ratio/high_mean": 0.0,
|
| 708 |
-
"clip_ratio/low_mean": 0.
|
| 709 |
-
"clip_ratio/low_min": 0.
|
| 710 |
-
"clip_ratio/region_mean": 0.
|
| 711 |
-
"completion_length":
|
| 712 |
"completions/clipped_ratio": 0.0,
|
| 713 |
-
"completions/max_length":
|
| 714 |
-
"completions/max_terminated_length":
|
| 715 |
-
"completions/mean_length":
|
| 716 |
-
"completions/mean_terminated_length":
|
| 717 |
-
"completions/min_length":
|
| 718 |
-
"completions/min_terminated_length":
|
| 719 |
-
"epoch": 0.
|
| 720 |
-
"frac_reward_zero_std": 0.
|
| 721 |
-
"grad_norm": 0.
|
| 722 |
"kl": 0.0,
|
| 723 |
-
"learning_rate":
|
| 724 |
-
"loss": -0.
|
| 725 |
-
"num_tokens":
|
| 726 |
"reward": 0.0,
|
| 727 |
-
"reward_std": 0.
|
| 728 |
"rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
|
| 729 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 730 |
"step": 1300
|
| 731 |
},
|
| 732 |
{
|
| 733 |
"clip_ratio/high_max": 0.0,
|
| 734 |
"clip_ratio/high_mean": 0.0,
|
| 735 |
-
"clip_ratio/low_mean":
|
| 736 |
-
"clip_ratio/low_min":
|
| 737 |
-
"clip_ratio/region_mean":
|
| 738 |
-
"completion_length":
|
| 739 |
"completions/clipped_ratio": 0.0,
|
| 740 |
-
"completions/max_length":
|
| 741 |
-
"completions/max_terminated_length":
|
| 742 |
-
"completions/mean_length":
|
| 743 |
-
"completions/mean_terminated_length":
|
| 744 |
-
"completions/min_length":
|
| 745 |
-
"completions/min_terminated_length":
|
| 746 |
-
"epoch": 0.
|
| 747 |
-
"frac_reward_zero_std": 0.
|
| 748 |
-
"grad_norm": 0.
|
| 749 |
"kl": 0.0,
|
| 750 |
-
"learning_rate":
|
| 751 |
-
"loss": 0.
|
| 752 |
-
"num_tokens":
|
| 753 |
-
"reward":
|
| 754 |
-
"reward_std": 0.
|
| 755 |
-
"rewards/TranscriptCorrectionGrpoReward/mean": 1.
|
| 756 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 757 |
"step": 1350
|
| 758 |
},
|
| 759 |
{
|
| 760 |
"clip_ratio/high_max": 0.0,
|
| 761 |
"clip_ratio/high_mean": 0.0,
|
| 762 |
-
"clip_ratio/low_mean": 0.
|
| 763 |
-
"clip_ratio/low_min": 0.
|
| 764 |
-
"clip_ratio/region_mean": 0.
|
| 765 |
-
"completion_length":
|
| 766 |
"completions/clipped_ratio": 0.0,
|
| 767 |
-
"completions/max_length":
|
| 768 |
-
"completions/max_terminated_length":
|
| 769 |
-
"completions/mean_length":
|
| 770 |
-
"completions/mean_terminated_length":
|
| 771 |
-
"completions/min_length":
|
| 772 |
-
"completions/min_terminated_length":
|
| 773 |
-
"epoch": 0.
|
| 774 |
-
"frac_reward_zero_std": 0.
|
| 775 |
-
"grad_norm": 0.
|
| 776 |
"kl": 0.0,
|
| 777 |
-
"learning_rate":
|
| 778 |
-
"loss":
|
| 779 |
-
"num_tokens":
|
| 780 |
-
"reward": 2.
|
| 781 |
-
"reward_std": 0.
|
| 782 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 783 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 784 |
"step": 1400
|
| 785 |
},
|
| 786 |
{
|
| 787 |
"clip_ratio/high_max": 0.0,
|
| 788 |
"clip_ratio/high_mean": 0.0,
|
| 789 |
-
"clip_ratio/low_mean":
|
| 790 |
-
"clip_ratio/low_min":
|
| 791 |
-
"clip_ratio/region_mean":
|
| 792 |
-
"completion_length":
|
| 793 |
"completions/clipped_ratio": 0.0,
|
| 794 |
-
"completions/max_length":
|
| 795 |
-
"completions/max_terminated_length":
|
| 796 |
-
"completions/mean_length":
|
| 797 |
-
"completions/mean_terminated_length":
|
| 798 |
-
"completions/min_length":
|
| 799 |
-
"completions/min_terminated_length":
|
| 800 |
-
"epoch": 0.
|
| 801 |
-
"frac_reward_zero_std": 0.
|
| 802 |
"grad_norm": 0.0,
|
| 803 |
"kl": 0.0,
|
| 804 |
-
"learning_rate":
|
| 805 |
-
"loss":
|
| 806 |
-
"num_tokens":
|
| 807 |
-
"reward":
|
| 808 |
-
"reward_std": 0.
|
| 809 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 810 |
-
"rewards/TranscriptCorrectionGrpoReward/std": 0.
|
| 811 |
"step": 1450
|
| 812 |
},
|
| 813 |
{
|
| 814 |
"clip_ratio/high_max": 0.0,
|
| 815 |
"clip_ratio/high_mean": 0.0,
|
| 816 |
-
"clip_ratio/low_mean": 0.
|
| 817 |
-
"clip_ratio/low_min": 0.
|
| 818 |
-
"clip_ratio/region_mean": 0.
|
| 819 |
-
"completion_length":
|
| 820 |
"completions/clipped_ratio": 0.0,
|
| 821 |
-
"completions/max_length":
|
| 822 |
-
"completions/max_terminated_length":
|
| 823 |
-
"completions/mean_length":
|
| 824 |
-
"completions/mean_terminated_length":
|
| 825 |
-
"completions/min_length":
|
| 826 |
-
"completions/min_terminated_length":
|
| 827 |
-
"epoch": 0.
|
| 828 |
-
"frac_reward_zero_std": 0.
|
| 829 |
-
"grad_norm": 0.
|
| 830 |
"kl": 0.0,
|
| 831 |
-
"learning_rate":
|
| 832 |
-
"loss": -0.
|
| 833 |
-
"num_tokens":
|
| 834 |
-
"reward":
|
| 835 |
-
"reward_std": 0.
|
| 836 |
-
"rewards/TranscriptCorrectionGrpoReward/mean":
|
| 837 |
-
"rewards/TranscriptCorrectionGrpoReward/std":
|
| 838 |
"step": 1500
|
| 839 |
},
|
| 840 |
{
|
| 841 |
-
"eval_cer_subset": 0.
|
| 842 |
-
"eval_cer_subset_edit_distance":
|
| 843 |
"eval_cer_subset_groups": 250,
|
| 844 |
"eval_cer_subset_items": 250,
|
| 845 |
"eval_cer_subset_ref_chars": 61407,
|
| 846 |
-
"eval_loss": 0.
|
| 847 |
"step": 1521
|
| 848 |
}
|
| 849 |
],
|
| 850 |
"logging_steps": 50,
|
| 851 |
-
"max_steps":
|
| 852 |
-
"num_input_tokens_seen":
|
| 853 |
"num_train_epochs": 4,
|
| 854 |
"save_steps": 507,
|
| 855 |
"stateful_callbacks": {
|
|
@@ -865,7 +865,7 @@
|
|
| 865 |
}
|
| 866 |
},
|
| 867 |
"total_flos": 0.0,
|
| 868 |
-
"train_batch_size":
|
| 869 |
"trial_name": null,
|
| 870 |
"trial_params": null
|
| 871 |
}
|
|
|
|
| 2 |
"best_global_step": null,
|
| 3 |
"best_metric": null,
|
| 4 |
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.0027015505941279908,
|
| 6 |
"eval_steps": 500,
|
| 7 |
"global_step": 1521,
|
| 8 |
"is_hyper_param_search": false,
|
|
|
|
| 15 |
"clip_ratio/low_mean": 0.0,
|
| 16 |
"clip_ratio/low_min": 0.0,
|
| 17 |
"clip_ratio/region_mean": 0.0,
|
| 18 |
+
"completion_length": 65.68,
|
| 19 |
"completions/clipped_ratio": 0.0,
|
| 20 |
+
"completions/max_length": 65.88,
|
| 21 |
+
"completions/max_terminated_length": 65.88,
|
| 22 |
+
"completions/mean_length": 65.68,
|
| 23 |
+
"completions/mean_terminated_length": 65.68,
|
| 24 |
+
"completions/min_length": 65.48,
|
| 25 |
+
"completions/min_terminated_length": 65.48,
|
| 26 |
+
"epoch": 8.88083693007229e-05,
|
| 27 |
+
"frac_reward_zero_std": 0.56,
|
| 28 |
+
"grad_norm": 0.5719226598739624,
|
| 29 |
"kl": 0.0,
|
| 30 |
+
"learning_rate": 5.439488466064254e-09,
|
| 31 |
+
"loss": -0.000577671229839325,
|
| 32 |
+
"num_tokens": 46552.0,
|
| 33 |
+
"reward": 2.9802322387695314e-09,
|
| 34 |
+
"reward_std": 0.47037978172302247,
|
| 35 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 2.9802322387695314e-09,
|
| 36 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.4703797769546509,
|
| 37 |
"step": 50
|
| 38 |
},
|
| 39 |
{
|
|
|
|
| 42 |
"clip_ratio/low_mean": 0.0,
|
| 43 |
"clip_ratio/low_min": 0.0,
|
| 44 |
"clip_ratio/region_mean": 0.0,
|
| 45 |
+
"completion_length": 53.46,
|
| 46 |
"completions/clipped_ratio": 0.0,
|
| 47 |
+
"completions/max_length": 53.6,
|
| 48 |
+
"completions/max_terminated_length": 53.6,
|
| 49 |
+
"completions/mean_length": 53.46,
|
| 50 |
+
"completions/mean_terminated_length": 53.46,
|
| 51 |
+
"completions/min_length": 53.24,
|
| 52 |
+
"completions/min_terminated_length": 53.24,
|
| 53 |
+
"epoch": 0.0001776167386014458,
|
| 54 |
+
"frac_reward_zero_std": 0.64,
|
| 55 |
"grad_norm": 0.0,
|
| 56 |
"kl": 0.0,
|
| 57 |
+
"learning_rate": 1.0989986900823695e-08,
|
| 58 |
+
"loss": -0.0009160846471786499,
|
| 59 |
+
"num_tokens": 85556.0,
|
| 60 |
+
"reward": 3.874301910400391e-09,
|
| 61 |
+
"reward_std": 0.38485618114471437,
|
| 62 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 3.874301910400391e-09,
|
| 63 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.38485618114471437,
|
| 64 |
"step": 100
|
| 65 |
},
|
| 66 |
{
|
| 67 |
"clip_ratio/high_max": 0.0,
|
| 68 |
"clip_ratio/high_mean": 0.0,
|
| 69 |
+
"clip_ratio/low_mean": 0.00011904762126505375,
|
| 70 |
+
"clip_ratio/low_min": 0.00011904762126505375,
|
| 71 |
+
"clip_ratio/region_mean": 0.00011904762126505375,
|
| 72 |
+
"completion_length": 62.48,
|
| 73 |
"completions/clipped_ratio": 0.0,
|
| 74 |
+
"completions/max_length": 62.6,
|
| 75 |
+
"completions/max_terminated_length": 62.6,
|
| 76 |
+
"completions/mean_length": 62.48,
|
| 77 |
+
"completions/mean_terminated_length": 62.48,
|
| 78 |
+
"completions/min_length": 62.24,
|
| 79 |
+
"completions/min_terminated_length": 62.24,
|
| 80 |
+
"epoch": 0.0002664251079021687,
|
| 81 |
+
"frac_reward_zero_std": 0.56,
|
| 82 |
+
"grad_norm": 0.2554231286048889,
|
| 83 |
"kl": 0.0,
|
| 84 |
+
"learning_rate": 1.6540485335583138e-08,
|
| 85 |
+
"loss": -0.0009111672639846802,
|
| 86 |
+
"num_tokens": 130380.0,
|
| 87 |
+
"reward": -1.1920928955078125e-09,
|
| 88 |
+
"reward_std": 0.4703797769546509,
|
| 89 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": -1.1920928955078125e-09,
|
| 90 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.47037978172302247,
|
| 91 |
"step": 150
|
| 92 |
},
|
| 93 |
{
|
|
|
|
| 96 |
"clip_ratio/low_mean": 0.0,
|
| 97 |
"clip_ratio/low_min": 0.0,
|
| 98 |
"clip_ratio/region_mean": 0.0,
|
| 99 |
+
"completion_length": 62.175,
|
| 100 |
"completions/clipped_ratio": 0.0,
|
| 101 |
+
"completions/max_length": 62.36,
|
| 102 |
+
"completions/max_terminated_length": 62.36,
|
| 103 |
+
"completions/mean_length": 62.175,
|
| 104 |
+
"completions/mean_terminated_length": 62.175,
|
| 105 |
+
"completions/min_length": 61.88,
|
| 106 |
+
"completions/min_terminated_length": 61.88,
|
| 107 |
+
"epoch": 0.0003552334772028916,
|
| 108 |
+
"frac_reward_zero_std": 0.56,
|
| 109 |
+
"grad_norm": 0.627755343914032,
|
| 110 |
"kl": 0.0,
|
| 111 |
+
"learning_rate": 2.2090983770342578e-08,
|
| 112 |
+
"loss": -0.0005262196063995361,
|
| 113 |
+
"num_tokens": 174959.0,
|
| 114 |
+
"reward": 1.1920928955078125e-09,
|
| 115 |
+
"reward_std": 0.47037978172302247,
|
| 116 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 1.1920928955078125e-09,
|
| 117 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.4703797769546509,
|
| 118 |
"step": 200
|
| 119 |
},
|
| 120 |
{
|
| 121 |
"clip_ratio/high_max": 0.0,
|
| 122 |
"clip_ratio/high_mean": 0.0,
|
| 123 |
+
"clip_ratio/low_mean": 0.00016129031777381897,
|
| 124 |
+
"clip_ratio/low_min": 0.00016129031777381897,
|
| 125 |
+
"clip_ratio/region_mean": 0.00016129031777381897,
|
| 126 |
+
"completion_length": 65.835,
|
| 127 |
"completions/clipped_ratio": 0.0,
|
| 128 |
+
"completions/max_length": 66.08,
|
| 129 |
+
"completions/max_terminated_length": 66.08,
|
| 130 |
+
"completions/mean_length": 65.835,
|
| 131 |
+
"completions/mean_terminated_length": 65.835,
|
| 132 |
+
"completions/min_length": 65.6,
|
| 133 |
+
"completions/min_terminated_length": 65.6,
|
| 134 |
+
"epoch": 0.0004440418465036145,
|
| 135 |
+
"frac_reward_zero_std": 0.52,
|
| 136 |
+
"grad_norm": 0.0,
|
| 137 |
"kl": 0.0,
|
| 138 |
+
"learning_rate": 2.764148220510202e-08,
|
| 139 |
+
"loss": -0.00036669328808784483,
|
| 140 |
+
"num_tokens": 222270.0,
|
| 141 |
+
"reward": 3.874301910400391e-09,
|
| 142 |
+
"reward_std": 0.5131415796279907,
|
| 143 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 3.874301910400391e-09,
|
| 144 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.5131415748596191,
|
| 145 |
"step": 250
|
| 146 |
},
|
| 147 |
{
|
|
|
|
| 150 |
"clip_ratio/low_mean": 0.0,
|
| 151 |
"clip_ratio/low_min": 0.0,
|
| 152 |
"clip_ratio/region_mean": 0.0,
|
| 153 |
+
"completion_length": 68.75,
|
| 154 |
"completions/clipped_ratio": 0.0,
|
| 155 |
+
"completions/max_length": 69.04,
|
| 156 |
+
"completions/max_terminated_length": 69.04,
|
| 157 |
+
"completions/mean_length": 68.75,
|
| 158 |
+
"completions/mean_terminated_length": 68.75,
|
| 159 |
+
"completions/min_length": 68.56,
|
| 160 |
+
"completions/min_terminated_length": 68.56,
|
| 161 |
+
"epoch": 0.0005328502158043374,
|
| 162 |
+
"frac_reward_zero_std": 0.6,
|
| 163 |
+
"grad_norm": 0.0,
|
| 164 |
"kl": 0.0,
|
| 165 |
+
"learning_rate": 3.3191980639861464e-08,
|
| 166 |
+
"loss": 0.0001394149661064148,
|
| 167 |
+
"num_tokens": 271740.0,
|
| 168 |
+
"reward": 6.258487701416015e-09,
|
| 169 |
+
"reward_std": 0.4276179838180542,
|
| 170 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 6.258487701416015e-09,
|
| 171 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.42761797904968263,
|
| 172 |
"step": 300
|
| 173 |
},
|
| 174 |
{
|
| 175 |
"clip_ratio/high_max": 0.0,
|
| 176 |
"clip_ratio/high_mean": 0.0,
|
| 177 |
+
"clip_ratio/low_mean": 0.0,
|
| 178 |
+
"clip_ratio/low_min": 0.0,
|
| 179 |
+
"clip_ratio/region_mean": 0.0,
|
| 180 |
+
"completion_length": 61.07,
|
| 181 |
"completions/clipped_ratio": 0.0,
|
| 182 |
+
"completions/max_length": 61.2,
|
| 183 |
+
"completions/max_terminated_length": 61.2,
|
| 184 |
+
"completions/mean_length": 61.07,
|
| 185 |
+
"completions/mean_terminated_length": 61.07,
|
| 186 |
+
"completions/min_length": 60.96,
|
| 187 |
+
"completions/min_terminated_length": 60.96,
|
| 188 |
+
"epoch": 0.0006216585851050603,
|
| 189 |
+
"frac_reward_zero_std": 0.72,
|
| 190 |
+
"grad_norm": 0.0,
|
| 191 |
"kl": 0.0,
|
| 192 |
+
"learning_rate": 3.8742479074620904e-08,
|
| 193 |
+
"loss": -5.508854985237121e-05,
|
| 194 |
+
"num_tokens": 316354.0,
|
| 195 |
+
"reward": 2.384185791015625e-09,
|
| 196 |
+
"reward_std": 0.29933258533477786,
|
| 197 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 2.384185791015625e-09,
|
| 198 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.2993325901031494,
|
| 199 |
"step": 350
|
| 200 |
},
|
| 201 |
{
|
| 202 |
"clip_ratio/high_max": 0.0,
|
| 203 |
"clip_ratio/high_mean": 0.0,
|
| 204 |
+
"clip_ratio/low_mean": 0.0,
|
| 205 |
+
"clip_ratio/low_min": 0.0,
|
| 206 |
+
"clip_ratio/region_mean": 0.0,
|
| 207 |
+
"completion_length": 64.22,
|
| 208 |
"completions/clipped_ratio": 0.0,
|
| 209 |
+
"completions/max_length": 64.6,
|
| 210 |
+
"completions/max_terminated_length": 64.6,
|
| 211 |
+
"completions/mean_length": 64.22,
|
| 212 |
+
"completions/mean_terminated_length": 64.22,
|
| 213 |
+
"completions/min_length": 63.92,
|
| 214 |
+
"completions/min_terminated_length": 63.92,
|
| 215 |
+
"epoch": 0.0007104669544057832,
|
| 216 |
+
"frac_reward_zero_std": 0.56,
|
| 217 |
+
"grad_norm": 0.0,
|
| 218 |
"kl": 0.0,
|
| 219 |
+
"learning_rate": 4.429297750938035e-08,
|
| 220 |
+
"loss": -0.0010024748742580413,
|
| 221 |
+
"num_tokens": 362550.0,
|
| 222 |
+
"reward": 5.960464477539063e-10,
|
| 223 |
+
"reward_std": 0.4703797769546509,
|
| 224 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 5.960464477539063e-10,
|
| 225 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.4703797769546509,
|
| 226 |
"step": 400
|
| 227 |
},
|
| 228 |
{
|
| 229 |
"clip_ratio/high_max": 0.0,
|
| 230 |
"clip_ratio/high_mean": 0.0,
|
| 231 |
+
"clip_ratio/low_mean": 0.0,
|
| 232 |
+
"clip_ratio/low_min": 0.0,
|
| 233 |
+
"clip_ratio/region_mean": 0.0,
|
| 234 |
+
"completion_length": 65.25,
|
| 235 |
"completions/clipped_ratio": 0.0,
|
| 236 |
+
"completions/max_length": 65.36,
|
| 237 |
+
"completions/max_terminated_length": 65.36,
|
| 238 |
+
"completions/mean_length": 65.25,
|
| 239 |
+
"completions/mean_terminated_length": 65.25,
|
| 240 |
+
"completions/min_length": 65.04,
|
| 241 |
+
"completions/min_terminated_length": 65.04,
|
| 242 |
+
"epoch": 0.000799275323706506,
|
| 243 |
+
"frac_reward_zero_std": 0.6,
|
| 244 |
+
"grad_norm": 0.0,
|
| 245 |
"kl": 0.0,
|
| 246 |
+
"learning_rate": 4.984347594413978e-08,
|
| 247 |
+
"loss": -0.0002909022569656372,
|
| 248 |
+
"num_tokens": 409416.0,
|
| 249 |
+
"reward": 1.7881393432617187e-09,
|
| 250 |
+
"reward_std": 0.4276179838180542,
|
| 251 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 1.7881393432617187e-09,
|
| 252 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.4276179838180542,
|
| 253 |
"step": 450
|
| 254 |
},
|
| 255 |
{
|
|
|
|
| 258 |
"clip_ratio/low_mean": 0.0,
|
| 259 |
"clip_ratio/low_min": 0.0,
|
| 260 |
"clip_ratio/region_mean": 0.0,
|
| 261 |
+
"completion_length": 65.76,
|
| 262 |
"completions/clipped_ratio": 0.0,
|
| 263 |
+
"completions/max_length": 65.92,
|
| 264 |
+
"completions/max_terminated_length": 65.92,
|
| 265 |
+
"completions/mean_length": 65.76,
|
| 266 |
+
"completions/mean_terminated_length": 65.76,
|
| 267 |
+
"completions/min_length": 65.68,
|
| 268 |
+
"completions/min_terminated_length": 65.68,
|
| 269 |
+
"epoch": 0.000888083693007229,
|
| 270 |
+
"frac_reward_zero_std": 0.84,
|
| 271 |
+
"grad_norm": 0.035873379558324814,
|
| 272 |
"kl": 0.0,
|
| 273 |
+
"learning_rate": 5.5393974378899236e-08,
|
| 274 |
+
"loss": -0.000893859937787056,
|
| 275 |
+
"num_tokens": 456024.0,
|
| 276 |
+
"reward": 0.0,
|
| 277 |
+
"reward_std": 0.17104718685150147,
|
| 278 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
|
| 279 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.17104719161987306,
|
| 280 |
"step": 500
|
| 281 |
},
|
| 282 |
{
|
| 283 |
+
"eval_cer_subset": 0.012913837184685785,
|
| 284 |
+
"eval_cer_subset_edit_distance": 793,
|
| 285 |
"eval_cer_subset_groups": 250,
|
| 286 |
"eval_cer_subset_items": 250,
|
| 287 |
"eval_cer_subset_ref_chars": 61407,
|
| 288 |
+
"eval_loss": 0.03959455178967374,
|
| 289 |
"step": 507
|
| 290 |
},
|
| 291 |
{
|
| 292 |
+
"clip_ratio/high_max": 8.333333767950535e-05,
|
| 293 |
+
"clip_ratio/high_mean": 8.333333767950535e-05,
|
| 294 |
+
"clip_ratio/low_mean": 0.00014285714365541935,
|
| 295 |
+
"clip_ratio/low_min": 0.00014285714365541935,
|
| 296 |
+
"clip_ratio/region_mean": 0.0002261904813349247,
|
| 297 |
+
"completion_length": 66.25,
|
| 298 |
"completions/clipped_ratio": 0.0,
|
| 299 |
+
"completions/max_length": 66.64,
|
| 300 |
+
"completions/max_terminated_length": 66.64,
|
| 301 |
+
"completions/mean_length": 66.25,
|
| 302 |
+
"completions/mean_terminated_length": 66.25,
|
| 303 |
+
"completions/min_length": 66.0,
|
| 304 |
+
"completions/min_terminated_length": 66.0,
|
| 305 |
+
"epoch": 0.0009768920623079518,
|
| 306 |
+
"frac_reward_zero_std": 0.64,
|
| 307 |
+
"grad_norm": 0.0,
|
| 308 |
"kl": 0.0,
|
| 309 |
+
"learning_rate": 6.094447281365867e-08,
|
| 310 |
+
"loss": 0.000100860595703125,
|
| 311 |
+
"num_tokens": 503522.0,
|
| 312 |
+
"reward": 5.960464477539063e-10,
|
| 313 |
+
"reward_std": 0.38485618114471437,
|
| 314 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 5.960464477539063e-10,
|
| 315 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.38485618114471437,
|
| 316 |
"step": 550
|
| 317 |
},
|
| 318 |
{
|
| 319 |
"clip_ratio/high_max": 0.0,
|
| 320 |
"clip_ratio/high_mean": 0.0,
|
| 321 |
+
"clip_ratio/low_mean": 0.0,
|
| 322 |
+
"clip_ratio/low_min": 0.0,
|
| 323 |
+
"clip_ratio/region_mean": 0.0,
|
| 324 |
+
"completion_length": 70.03,
|
| 325 |
"completions/clipped_ratio": 0.0,
|
| 326 |
+
"completions/max_length": 70.16,
|
| 327 |
+
"completions/max_terminated_length": 70.16,
|
| 328 |
+
"completions/mean_length": 70.03,
|
| 329 |
+
"completions/mean_terminated_length": 70.03,
|
| 330 |
+
"completions/min_length": 69.8,
|
| 331 |
+
"completions/min_terminated_length": 69.8,
|
| 332 |
+
"epoch": 0.0010657004316086747,
|
| 333 |
+
"frac_reward_zero_std": 0.52,
|
| 334 |
+
"grad_norm": 0.0,
|
| 335 |
"kl": 0.0,
|
| 336 |
+
"learning_rate": 6.649497124841812e-08,
|
| 337 |
+
"loss": -0.00029084362089633944,
|
| 338 |
+
"num_tokens": 553312.0,
|
| 339 |
+
"reward": 2.682209014892578e-09,
|
| 340 |
+
"reward_std": 0.5131415700912476,
|
| 341 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 2.682209014892578e-09,
|
| 342 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.5131415748596191,
|
| 343 |
"step": 600
|
| 344 |
},
|
| 345 |
{
|
| 346 |
"clip_ratio/high_max": 0.0,
|
| 347 |
"clip_ratio/high_mean": 0.0,
|
| 348 |
+
"clip_ratio/low_mean": 0.0,
|
| 349 |
+
"clip_ratio/low_min": 0.0,
|
| 350 |
+
"clip_ratio/region_mean": 0.0,
|
| 351 |
+
"completion_length": 71.26,
|
| 352 |
"completions/clipped_ratio": 0.0,
|
| 353 |
+
"completions/max_length": 71.76,
|
| 354 |
+
"completions/max_terminated_length": 71.76,
|
| 355 |
+
"completions/mean_length": 71.26,
|
| 356 |
+
"completions/mean_terminated_length": 71.26,
|
| 357 |
+
"completions/min_length": 70.52,
|
| 358 |
+
"completions/min_terminated_length": 70.52,
|
| 359 |
+
"epoch": 0.0011545088009093977,
|
| 360 |
+
"frac_reward_zero_std": 0.44,
|
| 361 |
+
"grad_norm": 0.0,
|
| 362 |
"kl": 0.0,
|
| 363 |
+
"learning_rate": 7.204546968317756e-08,
|
| 364 |
+
"loss": -0.0016799652576446533,
|
| 365 |
+
"num_tokens": 603180.0,
|
| 366 |
+
"reward": -1.7881393432617187e-09,
|
| 367 |
+
"reward_std": 0.5986651754379273,
|
| 368 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": -1.7881393432617187e-09,
|
| 369 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.5986651802062988,
|
| 370 |
"step": 650
|
| 371 |
},
|
| 372 |
{
|
|
|
|
| 375 |
"clip_ratio/low_mean": 0.0,
|
| 376 |
"clip_ratio/low_min": 0.0,
|
| 377 |
"clip_ratio/region_mean": 0.0,
|
| 378 |
+
"completion_length": 71.675,
|
| 379 |
"completions/clipped_ratio": 0.0,
|
| 380 |
+
"completions/max_length": 71.88,
|
| 381 |
+
"completions/max_terminated_length": 71.88,
|
| 382 |
+
"completions/mean_length": 71.675,
|
| 383 |
+
"completions/mean_terminated_length": 71.675,
|
| 384 |
+
"completions/min_length": 71.52,
|
| 385 |
+
"completions/min_terminated_length": 71.52,
|
| 386 |
+
"epoch": 0.0012433171702101206,
|
| 387 |
+
"frac_reward_zero_std": 0.6,
|
| 388 |
+
"grad_norm": 0.18659119307994843,
|
| 389 |
"kl": 0.0,
|
| 390 |
+
"learning_rate": 7.7595968117937e-08,
|
| 391 |
+
"loss": -0.0002913355827331543,
|
| 392 |
+
"num_tokens": 654187.0,
|
| 393 |
+
"reward": 4.172325134277344e-09,
|
| 394 |
+
"reward_std": 0.4276179838180542,
|
| 395 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 4.172325134277344e-09,
|
| 396 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.42761797904968263,
|
| 397 |
"step": 700
|
| 398 |
},
|
| 399 |
{
|
| 400 |
+
"clip_ratio/high_max": 3.731343196704984e-05,
|
| 401 |
+
"clip_ratio/high_mean": 3.731343196704984e-05,
|
| 402 |
"clip_ratio/low_mean": 0.0,
|
| 403 |
"clip_ratio/low_min": 0.0,
|
| 404 |
+
"clip_ratio/region_mean": 3.731343196704984e-05,
|
| 405 |
+
"completion_length": 65.635,
|
| 406 |
"completions/clipped_ratio": 0.0,
|
| 407 |
+
"completions/max_length": 65.92,
|
| 408 |
+
"completions/max_terminated_length": 65.92,
|
| 409 |
+
"completions/mean_length": 65.635,
|
| 410 |
+
"completions/mean_terminated_length": 65.635,
|
| 411 |
+
"completions/min_length": 65.4,
|
| 412 |
+
"completions/min_terminated_length": 65.4,
|
| 413 |
+
"epoch": 0.0013321255395108435,
|
| 414 |
+
"frac_reward_zero_std": 0.64,
|
| 415 |
+
"grad_norm": 0.32996097207069397,
|
| 416 |
"kl": 0.0,
|
| 417 |
+
"learning_rate": 8.314646655269643e-08,
|
| 418 |
+
"loss": 0.0007577691972255707,
|
| 419 |
+
"num_tokens": 700882.0,
|
| 420 |
+
"reward": 2.9802322387695314e-09,
|
| 421 |
+
"reward_std": 0.38485618114471437,
|
| 422 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 2.9802322387695314e-09,
|
| 423 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.38485618114471437,
|
| 424 |
"step": 750
|
| 425 |
},
|
| 426 |
{
|
| 427 |
"clip_ratio/high_max": 0.0,
|
| 428 |
"clip_ratio/high_mean": 0.0,
|
| 429 |
+
"clip_ratio/low_mean": 0.0,
|
| 430 |
+
"clip_ratio/low_min": 0.0,
|
| 431 |
+
"clip_ratio/region_mean": 0.0,
|
| 432 |
+
"completion_length": 67.45,
|
| 433 |
"completions/clipped_ratio": 0.0,
|
| 434 |
+
"completions/max_length": 67.64,
|
| 435 |
+
"completions/max_terminated_length": 67.64,
|
| 436 |
+
"completions/mean_length": 67.45,
|
| 437 |
+
"completions/mean_terminated_length": 67.45,
|
| 438 |
+
"completions/min_length": 67.16,
|
| 439 |
+
"completions/min_terminated_length": 67.16,
|
| 440 |
+
"epoch": 0.0014209339088115665,
|
| 441 |
+
"frac_reward_zero_std": 0.36,
|
| 442 |
+
"grad_norm": 0.2260763943195343,
|
| 443 |
"kl": 0.0,
|
| 444 |
+
"learning_rate": 8.869696498745589e-08,
|
| 445 |
+
"loss": -2.6343166828155516e-05,
|
| 446 |
+
"num_tokens": 749036.0,
|
| 447 |
+
"reward": 3.2782554626464844e-09,
|
| 448 |
+
"reward_std": 0.6841887712478638,
|
| 449 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 3.2782554626464844e-09,
|
| 450 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.6841887760162354,
|
| 451 |
"step": 800
|
| 452 |
},
|
| 453 |
{
|
| 454 |
"clip_ratio/high_max": 0.0,
|
| 455 |
"clip_ratio/high_mean": 0.0,
|
| 456 |
+
"clip_ratio/low_mean": 0.0001351351384073496,
|
| 457 |
+
"clip_ratio/low_min": 0.0001351351384073496,
|
| 458 |
+
"clip_ratio/region_mean": 0.0001351351384073496,
|
| 459 |
+
"completion_length": 75.2,
|
| 460 |
"completions/clipped_ratio": 0.0,
|
| 461 |
+
"completions/max_length": 75.36,
|
| 462 |
+
"completions/max_terminated_length": 75.36,
|
| 463 |
+
"completions/mean_length": 75.2,
|
| 464 |
+
"completions/mean_terminated_length": 75.2,
|
| 465 |
+
"completions/min_length": 75.04,
|
| 466 |
+
"completions/min_terminated_length": 75.04,
|
| 467 |
+
"epoch": 0.0015097422781122894,
|
| 468 |
+
"frac_reward_zero_std": 0.52,
|
| 469 |
+
"grad_norm": 0.1280648559331894,
|
| 470 |
"kl": 0.0,
|
| 471 |
+
"learning_rate": 9.424746342221532e-08,
|
| 472 |
+
"loss": -0.00025221601128578186,
|
| 473 |
+
"num_tokens": 802604.0,
|
| 474 |
+
"reward": 0.0,
|
| 475 |
+
"reward_std": 0.5131415748596191,
|
| 476 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
|
| 477 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.5131415796279907,
|
| 478 |
"step": 850
|
| 479 |
},
|
| 480 |
{
|
| 481 |
+
"clip_ratio/high_max": 0.0,
|
| 482 |
+
"clip_ratio/high_mean": 0.0,
|
| 483 |
+
"clip_ratio/low_mean": 0.0,
|
| 484 |
+
"clip_ratio/low_min": 0.0,
|
| 485 |
+
"clip_ratio/region_mean": 0.0,
|
| 486 |
+
"completion_length": 74.185,
|
| 487 |
"completions/clipped_ratio": 0.0,
|
| 488 |
+
"completions/max_length": 74.4,
|
| 489 |
+
"completions/max_terminated_length": 74.4,
|
| 490 |
+
"completions/mean_length": 74.185,
|
| 491 |
+
"completions/mean_terminated_length": 74.185,
|
| 492 |
+
"completions/min_length": 73.92,
|
| 493 |
+
"completions/min_terminated_length": 73.92,
|
| 494 |
+
"epoch": 0.001598550647413012,
|
| 495 |
+
"frac_reward_zero_std": 0.48,
|
| 496 |
+
"grad_norm": 0.4582439661026001,
|
| 497 |
"kl": 0.0,
|
| 498 |
+
"learning_rate": 9.979796185697477e-08,
|
| 499 |
+
"loss": -8.721232414245606e-05,
|
| 500 |
+
"num_tokens": 855385.0,
|
| 501 |
+
"reward": 3.5762786865234374e-09,
|
| 502 |
+
"reward_std": 0.5559033727645875,
|
| 503 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 3.5762786865234374e-09,
|
| 504 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.5559033727645875,
|
| 505 |
"step": 900
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"clip_ratio/high_max": 0.0,
|
| 509 |
"clip_ratio/high_mean": 0.0,
|
| 510 |
+
"clip_ratio/low_mean": 0.0,
|
| 511 |
+
"clip_ratio/low_min": 0.0,
|
| 512 |
+
"clip_ratio/region_mean": 0.0,
|
| 513 |
+
"completion_length": 69.455,
|
| 514 |
"completions/clipped_ratio": 0.0,
|
| 515 |
+
"completions/max_length": 69.68,
|
| 516 |
+
"completions/max_terminated_length": 69.68,
|
| 517 |
+
"completions/mean_length": 69.455,
|
| 518 |
+
"completions/mean_terminated_length": 69.455,
|
| 519 |
+
"completions/min_length": 69.04,
|
| 520 |
+
"completions/min_terminated_length": 69.04,
|
| 521 |
+
"epoch": 0.001687359016713735,
|
| 522 |
+
"frac_reward_zero_std": 0.6,
|
| 523 |
+
"grad_norm": 0.03906433284282684,
|
| 524 |
"kl": 0.0,
|
| 525 |
+
"learning_rate": 1.053484602917342e-07,
|
| 526 |
+
"loss": 0.00033147528767585755,
|
| 527 |
+
"num_tokens": 905012.0,
|
| 528 |
+
"reward": 4.172325134277344e-09,
|
| 529 |
+
"reward_std": 0.42761797428131104,
|
| 530 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 4.172325134277344e-09,
|
| 531 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.42761797904968263,
|
| 532 |
"step": 950
|
| 533 |
},
|
| 534 |
{
|
| 535 |
"clip_ratio/high_max": 0.0,
|
| 536 |
"clip_ratio/high_mean": 0.0,
|
| 537 |
+
"clip_ratio/low_mean": 0.0001351351384073496,
|
| 538 |
+
"clip_ratio/low_min": 0.0001351351384073496,
|
| 539 |
+
"clip_ratio/region_mean": 0.0001351351384073496,
|
| 540 |
+
"completion_length": 61.425,
|
| 541 |
"completions/clipped_ratio": 0.0,
|
| 542 |
+
"completions/max_length": 61.72,
|
| 543 |
+
"completions/max_terminated_length": 61.72,
|
| 544 |
+
"completions/mean_length": 61.425,
|
| 545 |
+
"completions/mean_terminated_length": 61.425,
|
| 546 |
+
"completions/min_length": 61.12,
|
| 547 |
+
"completions/min_terminated_length": 61.12,
|
| 548 |
+
"epoch": 0.001776167386014458,
|
| 549 |
+
"frac_reward_zero_std": 0.56,
|
| 550 |
+
"grad_norm": 1.1754662990570068,
|
| 551 |
"kl": 0.0,
|
| 552 |
+
"learning_rate": 1.1089895872649365e-07,
|
| 553 |
+
"loss": -0.0008977200835943222,
|
| 554 |
+
"num_tokens": 949385.0,
|
| 555 |
+
"reward": -5.960464477539063e-10,
|
| 556 |
+
"reward_std": 0.47037978172302247,
|
| 557 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": -5.960464477539063e-10,
|
| 558 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.4703797769546509,
|
| 559 |
"step": 1000
|
| 560 |
},
|
| 561 |
{
|
| 562 |
+
"eval_cer_subset": 0.013076685068477535,
|
| 563 |
+
"eval_cer_subset_edit_distance": 803,
|
| 564 |
"eval_cer_subset_groups": 250,
|
| 565 |
"eval_cer_subset_items": 250,
|
| 566 |
"eval_cer_subset_ref_chars": 61407,
|
| 567 |
+
"eval_loss": 0.03962622743110548,
|
| 568 |
"step": 1014
|
| 569 |
},
|
| 570 |
{
|
| 571 |
"clip_ratio/high_max": 0.0,
|
| 572 |
"clip_ratio/high_mean": 0.0,
|
| 573 |
+
"clip_ratio/low_mean": 0.0,
|
| 574 |
+
"clip_ratio/low_min": 0.0,
|
| 575 |
+
"clip_ratio/region_mean": 0.0,
|
| 576 |
+
"completion_length": 61.815,
|
| 577 |
"completions/clipped_ratio": 0.0,
|
| 578 |
+
"completions/max_length": 62.04,
|
| 579 |
+
"completions/max_terminated_length": 62.04,
|
| 580 |
+
"completions/mean_length": 61.815,
|
| 581 |
+
"completions/mean_terminated_length": 61.815,
|
| 582 |
+
"completions/min_length": 61.56,
|
| 583 |
+
"completions/min_terminated_length": 61.56,
|
| 584 |
+
"epoch": 0.001864975755315181,
|
| 585 |
+
"frac_reward_zero_std": 0.6,
|
| 586 |
+
"grad_norm": 0.0,
|
| 587 |
"kl": 0.0,
|
| 588 |
+
"learning_rate": 1.1644945716125309e-07,
|
| 589 |
+
"loss": 0.0001289863884449005,
|
| 590 |
+
"num_tokens": 993580.0,
|
| 591 |
+
"reward": 1.4901161193847657e-09,
|
| 592 |
+
"reward_std": 0.4276179885864258,
|
| 593 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 1.4901161193847657e-09,
|
| 594 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.42761797904968263,
|
| 595 |
"step": 1050
|
| 596 |
},
|
| 597 |
{
|
| 598 |
"clip_ratio/high_max": 0.0,
|
| 599 |
"clip_ratio/high_mean": 0.0,
|
| 600 |
+
"clip_ratio/low_mean": 0.00011363636702299118,
|
| 601 |
+
"clip_ratio/low_min": 0.00011363636702299118,
|
| 602 |
+
"clip_ratio/region_mean": 0.00011363636702299118,
|
| 603 |
+
"completion_length": 62.35,
|
| 604 |
"completions/clipped_ratio": 0.0,
|
| 605 |
+
"completions/max_length": 62.56,
|
| 606 |
+
"completions/max_terminated_length": 62.56,
|
| 607 |
+
"completions/mean_length": 62.35,
|
| 608 |
+
"completions/mean_terminated_length": 62.35,
|
| 609 |
+
"completions/min_length": 62.24,
|
| 610 |
+
"completions/min_terminated_length": 62.24,
|
| 611 |
+
"epoch": 0.0019537841246159036,
|
| 612 |
+
"frac_reward_zero_std": 0.64,
|
| 613 |
+
"grad_norm": 0.14012472331523895,
|
| 614 |
"kl": 0.0,
|
| 615 |
+
"learning_rate": 1.2199995559601254e-07,
|
| 616 |
+
"loss": 0.0005487602949142456,
|
| 617 |
+
"num_tokens": 1038234.0,
|
| 618 |
+
"reward": 3.5762786865234374e-09,
|
| 619 |
+
"reward_std": 0.38485618591308596,
|
| 620 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 3.5762786865234374e-09,
|
| 621 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.38485618114471437,
|
| 622 |
"step": 1100
|
| 623 |
},
|
| 624 |
{
|
| 625 |
"clip_ratio/high_max": 0.0,
|
| 626 |
"clip_ratio/high_mean": 0.0,
|
| 627 |
+
"clip_ratio/low_mean": 0.0,
|
| 628 |
+
"clip_ratio/low_min": 0.0,
|
| 629 |
+
"clip_ratio/region_mean": 0.0,
|
| 630 |
+
"completion_length": 67.58,
|
| 631 |
"completions/clipped_ratio": 0.0,
|
| 632 |
+
"completions/max_length": 67.76,
|
| 633 |
+
"completions/max_terminated_length": 67.76,
|
| 634 |
+
"completions/mean_length": 67.58,
|
| 635 |
+
"completions/mean_terminated_length": 67.58,
|
| 636 |
+
"completions/min_length": 67.32,
|
| 637 |
+
"completions/min_terminated_length": 67.32,
|
| 638 |
+
"epoch": 0.0020425924939166265,
|
| 639 |
+
"frac_reward_zero_std": 0.48,
|
| 640 |
+
"grad_norm": 0.0,
|
| 641 |
"kl": 0.0,
|
| 642 |
+
"learning_rate": 1.2755045403077197e-07,
|
| 643 |
+
"loss": -0.00019566014409065248,
|
| 644 |
+
"num_tokens": 1086214.0,
|
| 645 |
+
"reward": 2.682209014892578e-09,
|
| 646 |
+
"reward_std": 0.5559033727645875,
|
| 647 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 2.682209014892578e-09,
|
| 648 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.5559033727645875,
|
| 649 |
"step": 1150
|
| 650 |
},
|
| 651 |
{
|
| 652 |
+
"clip_ratio/high_max": 0.0,
|
| 653 |
+
"clip_ratio/high_mean": 0.0,
|
| 654 |
+
"clip_ratio/low_mean": 6.849315017461777e-05,
|
| 655 |
+
"clip_ratio/low_min": 6.849315017461777e-05,
|
| 656 |
+
"clip_ratio/region_mean": 6.849315017461777e-05,
|
| 657 |
+
"completion_length": 67.835,
|
| 658 |
"completions/clipped_ratio": 0.0,
|
| 659 |
+
"completions/max_length": 68.12,
|
| 660 |
+
"completions/max_terminated_length": 68.12,
|
| 661 |
+
"completions/mean_length": 67.835,
|
| 662 |
+
"completions/mean_terminated_length": 67.835,
|
| 663 |
+
"completions/min_length": 67.64,
|
| 664 |
+
"completions/min_terminated_length": 67.64,
|
| 665 |
+
"epoch": 0.0021314008632173495,
|
| 666 |
+
"frac_reward_zero_std": 0.64,
|
| 667 |
+
"grad_norm": 0.0,
|
| 668 |
"kl": 0.0,
|
| 669 |
+
"learning_rate": 1.331009524655314e-07,
|
| 670 |
+
"loss": 0.0007767707109451294,
|
| 671 |
+
"num_tokens": 1134493.0,
|
| 672 |
+
"reward": 1.1920928955078125e-09,
|
| 673 |
+
"reward_std": 0.3848561763763428,
|
| 674 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 1.1920928955078125e-09,
|
| 675 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.38485618114471437,
|
| 676 |
"step": 1200
|
| 677 |
},
|
| 678 |
{
|
|
|
|
| 681 |
"clip_ratio/low_mean": 0.0,
|
| 682 |
"clip_ratio/low_min": 0.0,
|
| 683 |
"clip_ratio/region_mean": 0.0,
|
| 684 |
+
"completion_length": 68.39,
|
| 685 |
"completions/clipped_ratio": 0.0,
|
| 686 |
+
"completions/max_length": 68.68,
|
| 687 |
+
"completions/max_terminated_length": 68.68,
|
| 688 |
+
"completions/mean_length": 68.39,
|
| 689 |
+
"completions/mean_terminated_length": 68.39,
|
| 690 |
+
"completions/min_length": 68.2,
|
| 691 |
+
"completions/min_terminated_length": 68.2,
|
| 692 |
+
"epoch": 0.0022202092325180724,
|
| 693 |
+
"frac_reward_zero_std": 0.6,
|
| 694 |
+
"grad_norm": 0.18301734328269958,
|
| 695 |
"kl": 0.0,
|
| 696 |
+
"learning_rate": 1.3865145090029086e-07,
|
| 697 |
+
"loss": -0.00031893253326416017,
|
| 698 |
+
"num_tokens": 1183027.0,
|
| 699 |
+
"reward": 2.384185791015625e-09,
|
| 700 |
+
"reward_std": 0.42761797904968263,
|
| 701 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 2.384185791015625e-09,
|
| 702 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.42761797904968263,
|
| 703 |
"step": 1250
|
| 704 |
},
|
| 705 |
{
|
| 706 |
"clip_ratio/high_max": 0.0,
|
| 707 |
"clip_ratio/high_mean": 0.0,
|
| 708 |
+
"clip_ratio/low_mean": 0.0,
|
| 709 |
+
"clip_ratio/low_min": 0.0,
|
| 710 |
+
"clip_ratio/region_mean": 0.0,
|
| 711 |
+
"completion_length": 64.62,
|
| 712 |
"completions/clipped_ratio": 0.0,
|
| 713 |
+
"completions/max_length": 64.72,
|
| 714 |
+
"completions/max_terminated_length": 64.72,
|
| 715 |
+
"completions/mean_length": 64.62,
|
| 716 |
+
"completions/mean_terminated_length": 64.62,
|
| 717 |
+
"completions/min_length": 64.48,
|
| 718 |
+
"completions/min_terminated_length": 64.48,
|
| 719 |
+
"epoch": 0.0023090176018187953,
|
| 720 |
+
"frac_reward_zero_std": 0.8,
|
| 721 |
+
"grad_norm": 0.0,
|
| 722 |
"kl": 0.0,
|
| 723 |
+
"learning_rate": 1.442019493350503e-07,
|
| 724 |
+
"loss": -0.0005694808065891266,
|
| 725 |
+
"num_tokens": 1229847.0,
|
| 726 |
"reward": 0.0,
|
| 727 |
+
"reward_std": 0.21380898952484131,
|
| 728 |
"rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
|
| 729 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.21380898952484131,
|
| 730 |
"step": 1300
|
| 731 |
},
|
| 732 |
{
|
| 733 |
"clip_ratio/high_max": 0.0,
|
| 734 |
"clip_ratio/high_mean": 0.0,
|
| 735 |
+
"clip_ratio/low_mean": 0.00011627906933426857,
|
| 736 |
+
"clip_ratio/low_min": 0.00011627906933426857,
|
| 737 |
+
"clip_ratio/region_mean": 0.00011627906933426857,
|
| 738 |
+
"completion_length": 62.82,
|
| 739 |
"completions/clipped_ratio": 0.0,
|
| 740 |
+
"completions/max_length": 63.04,
|
| 741 |
+
"completions/max_terminated_length": 63.04,
|
| 742 |
+
"completions/mean_length": 62.82,
|
| 743 |
+
"completions/mean_terminated_length": 62.82,
|
| 744 |
+
"completions/min_length": 62.6,
|
| 745 |
+
"completions/min_terminated_length": 62.6,
|
| 746 |
+
"epoch": 0.0023978259711195183,
|
| 747 |
+
"frac_reward_zero_std": 0.52,
|
| 748 |
+
"grad_norm": 0.0,
|
| 749 |
"kl": 0.0,
|
| 750 |
+
"learning_rate": 1.4975244776980973e-07,
|
| 751 |
+
"loss": -0.0006587636470794677,
|
| 752 |
+
"num_tokens": 1274827.0,
|
| 753 |
+
"reward": -1.9371509552001954e-09,
|
| 754 |
+
"reward_std": 0.5131415700912476,
|
| 755 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": -1.9371509552001954e-09,
|
| 756 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.5131415796279907,
|
| 757 |
"step": 1350
|
| 758 |
},
|
| 759 |
{
|
| 760 |
"clip_ratio/high_max": 0.0,
|
| 761 |
"clip_ratio/high_mean": 0.0,
|
| 762 |
+
"clip_ratio/low_mean": 0.0,
|
| 763 |
+
"clip_ratio/low_min": 0.0,
|
| 764 |
+
"clip_ratio/region_mean": 0.0,
|
| 765 |
+
"completion_length": 62.54,
|
| 766 |
"completions/clipped_ratio": 0.0,
|
| 767 |
+
"completions/max_length": 62.68,
|
| 768 |
+
"completions/max_terminated_length": 62.68,
|
| 769 |
+
"completions/mean_length": 62.54,
|
| 770 |
+
"completions/mean_terminated_length": 62.54,
|
| 771 |
+
"completions/min_length": 62.36,
|
| 772 |
+
"completions/min_terminated_length": 62.36,
|
| 773 |
+
"epoch": 0.002486634340420241,
|
| 774 |
+
"frac_reward_zero_std": 0.56,
|
| 775 |
+
"grad_norm": 0.0,
|
| 776 |
"kl": 0.0,
|
| 777 |
+
"learning_rate": 1.553029462045692e-07,
|
| 778 |
+
"loss": 0.00019890040159225463,
|
| 779 |
+
"num_tokens": 1320071.0,
|
| 780 |
+
"reward": 2.2351741790771484e-09,
|
| 781 |
+
"reward_std": 0.47037978172302247,
|
| 782 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 2.2351741790771484e-09,
|
| 783 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.47037978172302247,
|
| 784 |
"step": 1400
|
| 785 |
},
|
| 786 |
{
|
| 787 |
"clip_ratio/high_max": 0.0,
|
| 788 |
"clip_ratio/high_mean": 0.0,
|
| 789 |
+
"clip_ratio/low_mean": 0.0,
|
| 790 |
+
"clip_ratio/low_min": 0.0,
|
| 791 |
+
"clip_ratio/region_mean": 0.0,
|
| 792 |
+
"completion_length": 63.6,
|
| 793 |
"completions/clipped_ratio": 0.0,
|
| 794 |
+
"completions/max_length": 63.68,
|
| 795 |
+
"completions/max_terminated_length": 63.68,
|
| 796 |
+
"completions/mean_length": 63.6,
|
| 797 |
+
"completions/mean_terminated_length": 63.6,
|
| 798 |
+
"completions/min_length": 63.48,
|
| 799 |
+
"completions/min_terminated_length": 63.48,
|
| 800 |
+
"epoch": 0.002575442709720964,
|
| 801 |
+
"frac_reward_zero_std": 0.8,
|
| 802 |
"grad_norm": 0.0,
|
| 803 |
"kl": 0.0,
|
| 804 |
+
"learning_rate": 1.6085344463932862e-07,
|
| 805 |
+
"loss": 2.284705638885498e-05,
|
| 806 |
+
"num_tokens": 1366199.0,
|
| 807 |
+
"reward": 1.7881393432617187e-09,
|
| 808 |
+
"reward_std": 0.21380898952484131,
|
| 809 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 1.7881393432617187e-09,
|
| 810 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.21380898952484131,
|
| 811 |
"step": 1450
|
| 812 |
},
|
| 813 |
{
|
| 814 |
"clip_ratio/high_max": 0.0,
|
| 815 |
"clip_ratio/high_mean": 0.0,
|
| 816 |
+
"clip_ratio/low_mean": 0.0,
|
| 817 |
+
"clip_ratio/low_min": 0.0,
|
| 818 |
+
"clip_ratio/region_mean": 0.0,
|
| 819 |
+
"completion_length": 68.165,
|
| 820 |
"completions/clipped_ratio": 0.0,
|
| 821 |
+
"completions/max_length": 68.36,
|
| 822 |
+
"completions/max_terminated_length": 68.36,
|
| 823 |
+
"completions/mean_length": 68.165,
|
| 824 |
+
"completions/mean_terminated_length": 68.165,
|
| 825 |
+
"completions/min_length": 67.88,
|
| 826 |
+
"completions/min_terminated_length": 67.88,
|
| 827 |
+
"epoch": 0.002664251079021687,
|
| 828 |
+
"frac_reward_zero_std": 0.56,
|
| 829 |
+
"grad_norm": 0.32475942373275757,
|
| 830 |
"kl": 0.0,
|
| 831 |
+
"learning_rate": 1.6640394307408808e-07,
|
| 832 |
+
"loss": -0.0001750713586807251,
|
| 833 |
+
"num_tokens": 1415240.0,
|
| 834 |
+
"reward": 8.940696716308593e-10,
|
| 835 |
+
"reward_std": 0.4703797721862793,
|
| 836 |
+
"rewards/TranscriptCorrectionGrpoReward/mean": 8.940696716308593e-10,
|
| 837 |
+
"rewards/TranscriptCorrectionGrpoReward/std": 0.4703797769546509,
|
| 838 |
"step": 1500
|
| 839 |
},
|
| 840 |
{
|
| 841 |
+
"eval_cer_subset": 0.01286498281954826,
|
| 842 |
+
"eval_cer_subset_edit_distance": 790,
|
| 843 |
"eval_cer_subset_groups": 250,
|
| 844 |
"eval_cer_subset_items": 250,
|
| 845 |
"eval_cer_subset_ref_chars": 61407,
|
| 846 |
+
"eval_loss": 0.03962622743110548,
|
| 847 |
"step": 1521
|
| 848 |
}
|
| 849 |
],
|
| 850 |
"logging_steps": 50,
|
| 851 |
+
"max_steps": 2252040,
|
| 852 |
+
"num_input_tokens_seen": 1438098,
|
| 853 |
"num_train_epochs": 4,
|
| 854 |
"save_steps": 507,
|
| 855 |
"stateful_callbacks": {
|
|
|
|
| 865 |
}
|
| 866 |
},
|
| 867 |
"total_flos": 0.0,
|
| 868 |
+
"train_batch_size": 1,
|
| 869 |
"trial_name": null,
|
| 870 |
"trial_params": null
|
| 871 |
}
|
meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521/training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 6737
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9bdad4bf60f02650b7130e20d3de1e81654197d408ff77954683d133ac0bce86
|
| 3 |
size 6737
|
meta-llama__llama-3.2-1b/grpo/run_manifest.json
CHANGED
|
@@ -6,10 +6,10 @@
|
|
| 6 |
"model_name": "meta-llama/Llama-3.2-1B",
|
| 7 |
"status": "running",
|
| 8 |
"best_metric_name": "eval_cer_subset",
|
| 9 |
-
"best_metric_value": 0.
|
| 10 |
-
"best_step":
|
| 11 |
-
"best_checkpoint_step":
|
| 12 |
-
"latest_checkpoint_step":
|
| 13 |
"checkpoints": [
|
| 14 |
{
|
| 15 |
"step": 507,
|
|
@@ -22,10 +22,16 @@
|
|
| 22 |
"epoch": 0.0018010337294186605,
|
| 23 |
"repo_path": "meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1014",
|
| 24 |
"timestamp": "2026-04-09T03:49:15.799552+00:00"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
}
|
| 26 |
],
|
| 27 |
"epoch_artifacts": [],
|
| 28 |
"resume_history": [],
|
| 29 |
"created_at": "2026-04-09T03:35:11.674696+00:00",
|
| 30 |
-
"updated_at": "2026-04-09T03:
|
| 31 |
}
|
|
|
|
| 6 |
"model_name": "meta-llama/Llama-3.2-1B",
|
| 7 |
"status": "running",
|
| 8 |
"best_metric_name": "eval_cer_subset",
|
| 9 |
+
"best_metric_value": 0.01286498281954826,
|
| 10 |
+
"best_step": 1521,
|
| 11 |
+
"best_checkpoint_step": 1521,
|
| 12 |
+
"latest_checkpoint_step": 1521,
|
| 13 |
"checkpoints": [
|
| 14 |
{
|
| 15 |
"step": 507,
|
|
|
|
| 22 |
"epoch": 0.0018010337294186605,
|
| 23 |
"repo_path": "meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1014",
|
| 24 |
"timestamp": "2026-04-09T03:49:15.799552+00:00"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"step": 1521,
|
| 28 |
+
"epoch": 0.0027015505941279908,
|
| 29 |
+
"repo_path": "meta-llama__llama-3.2-1b/grpo/checkpoints/checkpoint-1521",
|
| 30 |
+
"timestamp": "2026-04-09T03:55:44.298843+00:00"
|
| 31 |
}
|
| 32 |
],
|
| 33 |
"epoch_artifacts": [],
|
| 34 |
"resume_history": [],
|
| 35 |
"created_at": "2026-04-09T03:35:11.674696+00:00",
|
| 36 |
+
"updated_at": "2026-04-09T03:55:44.298859+00:00"
|
| 37 |
}
|