Upload task output 1
Browse files- config.json +11 -15
- generation_config.json +10 -0
- loss.txt +1 -1
- model.safetensors +2 -2
- trainer_state.json +981 -420
- training_args.bin +1 -1
config.json
CHANGED
|
@@ -3,13 +3,13 @@
|
|
| 3 |
"Qwen2ForCausalLM"
|
| 4 |
],
|
| 5 |
"attention_dropout": 0.0,
|
| 6 |
-
"bos_token_id":
|
| 7 |
"dtype": "bfloat16",
|
| 8 |
"eos_token_id": 151645,
|
| 9 |
"hidden_act": "silu",
|
| 10 |
-
"hidden_size":
|
| 11 |
"initializer_range": 0.02,
|
| 12 |
-
"intermediate_size":
|
| 13 |
"layer_types": [
|
| 14 |
"full_attention",
|
| 15 |
"full_attention",
|
|
@@ -34,28 +34,24 @@
|
|
| 34 |
"full_attention",
|
| 35 |
"full_attention",
|
| 36 |
"full_attention",
|
| 37 |
-
"full_attention",
|
| 38 |
-
"full_attention",
|
| 39 |
-
"full_attention",
|
| 40 |
-
"full_attention",
|
| 41 |
"full_attention"
|
| 42 |
],
|
| 43 |
"max_position_embeddings": 32768,
|
| 44 |
-
"max_window_layers":
|
| 45 |
"model_type": "qwen2",
|
| 46 |
-
"num_attention_heads":
|
| 47 |
-
"num_hidden_layers":
|
| 48 |
-
"num_key_value_heads":
|
| 49 |
-
"pad_token_id":
|
| 50 |
"rms_norm_eps": 1e-06,
|
| 51 |
"rope_parameters": {
|
| 52 |
"rope_theta": 1000000.0,
|
| 53 |
"rope_type": "default"
|
| 54 |
},
|
| 55 |
"sliding_window": null,
|
| 56 |
-
"tie_word_embeddings":
|
| 57 |
"transformers_version": "5.1.0",
|
| 58 |
-
"use_cache":
|
| 59 |
"use_sliding_window": false,
|
| 60 |
-
"vocab_size":
|
| 61 |
}
|
|
|
|
| 3 |
"Qwen2ForCausalLM"
|
| 4 |
],
|
| 5 |
"attention_dropout": 0.0,
|
| 6 |
+
"bos_token_id": null,
|
| 7 |
"dtype": "bfloat16",
|
| 8 |
"eos_token_id": 151645,
|
| 9 |
"hidden_act": "silu",
|
| 10 |
+
"hidden_size": 896,
|
| 11 |
"initializer_range": 0.02,
|
| 12 |
+
"intermediate_size": 4864,
|
| 13 |
"layer_types": [
|
| 14 |
"full_attention",
|
| 15 |
"full_attention",
|
|
|
|
| 34 |
"full_attention",
|
| 35 |
"full_attention",
|
| 36 |
"full_attention",
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
"full_attention"
|
| 38 |
],
|
| 39 |
"max_position_embeddings": 32768,
|
| 40 |
+
"max_window_layers": 21,
|
| 41 |
"model_type": "qwen2",
|
| 42 |
+
"num_attention_heads": 14,
|
| 43 |
+
"num_hidden_layers": 24,
|
| 44 |
+
"num_key_value_heads": 2,
|
| 45 |
+
"pad_token_id": 151643,
|
| 46 |
"rms_norm_eps": 1e-06,
|
| 47 |
"rope_parameters": {
|
| 48 |
"rope_theta": 1000000.0,
|
| 49 |
"rope_type": "default"
|
| 50 |
},
|
| 51 |
"sliding_window": null,
|
| 52 |
+
"tie_word_embeddings": true,
|
| 53 |
"transformers_version": "5.1.0",
|
| 54 |
+
"use_cache": false,
|
| 55 |
"use_sliding_window": false,
|
| 56 |
+
"vocab_size": 151936
|
| 57 |
}
|
generation_config.json
CHANGED
|
@@ -1,3 +1,13 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"transformers_version": "5.1.0"
|
| 3 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"do_sample": true,
|
| 3 |
+
"eos_token_id": [
|
| 4 |
+
151645,
|
| 5 |
+
151643
|
| 6 |
+
],
|
| 7 |
+
"pad_token_id": 151643,
|
| 8 |
+
"repetition_penalty": 1.1,
|
| 9 |
+
"temperature": 0.7,
|
| 10 |
+
"top_k": 20,
|
| 11 |
+
"top_p": 0.8,
|
| 12 |
"transformers_version": "5.1.0"
|
| 13 |
}
|
loss.txt
CHANGED
|
@@ -1 +1 @@
|
|
| 1 |
-
|
|
|
|
| 1 |
+
152,-0.06000000238418579
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df164ec05eae3899c60a65c9089d6d4eedccdc33078b9a90ec77d9672e240f89
|
| 3 |
+
size 988097824
|
trainer_state.json
CHANGED
|
@@ -2,507 +2,507 @@
|
|
| 2 |
"best_global_step": null,
|
| 3 |
"best_metric": null,
|
| 4 |
"best_model_checkpoint": null,
|
| 5 |
-
"epoch": 0.
|
| 6 |
"eval_steps": 500,
|
| 7 |
-
"global_step":
|
| 8 |
"is_hyper_param_search": false,
|
| 9 |
"is_local_process_zero": true,
|
| 10 |
"is_world_process_zero": true,
|
| 11 |
"log_history": [
|
| 12 |
{
|
| 13 |
-
"clip_ratio/high_max": 0.
|
| 14 |
-
"clip_ratio/high_mean": 0.
|
| 15 |
-
"clip_ratio/low_mean": 0.
|
| 16 |
"clip_ratio/low_min": 0.0,
|
| 17 |
-
"clip_ratio/region_mean": 0.
|
| 18 |
"completions/clipped_ratio": 0.0,
|
| 19 |
-
"completions/max_length": 374.
|
| 20 |
-
"completions/max_terminated_length": 374.
|
| 21 |
-
"completions/mean_length":
|
| 22 |
-
"completions/mean_terminated_length":
|
| 23 |
-
"completions/min_length":
|
| 24 |
-
"completions/min_terminated_length":
|
| 25 |
-
"entropy": 0.
|
| 26 |
"epoch": 0.0008,
|
| 27 |
-
"frac_reward_zero_std": 0.
|
| 28 |
-
"grad_norm": 0.
|
| 29 |
-
"kl": 0.
|
| 30 |
"learning_rate": 1.137216e-06,
|
| 31 |
-
"loss":
|
| 32 |
-
"num_tokens":
|
| 33 |
-
"reward": 0.
|
| 34 |
-
"reward_std": 0.
|
| 35 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 36 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 37 |
-
"sampling/importance_sampling_ratio/max": 1.
|
| 38 |
-
"sampling/importance_sampling_ratio/mean":
|
| 39 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 40 |
-
"sampling/sampling_logp_difference/max":
|
| 41 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 42 |
"step": 5,
|
| 43 |
-
"step_time":
|
| 44 |
},
|
| 45 |
{
|
| 46 |
-
"clip_ratio/high_max": 0.
|
| 47 |
-
"clip_ratio/high_mean": 0.
|
| 48 |
-
"clip_ratio/low_mean": 0.
|
| 49 |
"clip_ratio/low_min": 0.0,
|
| 50 |
-
"clip_ratio/region_mean": 0.
|
| 51 |
"completions/clipped_ratio": 0.0,
|
| 52 |
-
"completions/max_length": 374.
|
| 53 |
-
"completions/max_terminated_length": 374.
|
| 54 |
-
"completions/mean_length": 290.
|
| 55 |
-
"completions/mean_terminated_length": 290.
|
| 56 |
-
"completions/min_length":
|
| 57 |
-
"completions/min_terminated_length":
|
| 58 |
-
"entropy": 0.
|
| 59 |
"epoch": 0.0016,
|
| 60 |
-
"frac_reward_zero_std": 0.
|
| 61 |
-
"grad_norm": 0.
|
| 62 |
-
"kl": 0.
|
| 63 |
"learning_rate": 2.5587359999999995e-06,
|
| 64 |
-
"loss":
|
| 65 |
-
"num_tokens":
|
| 66 |
-
"reward": 0.
|
| 67 |
-
"reward_std": 0.
|
| 68 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 69 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 70 |
-
"sampling/importance_sampling_ratio/max": 2.
|
| 71 |
-
"sampling/importance_sampling_ratio/mean":
|
| 72 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 73 |
-
"sampling/sampling_logp_difference/max":
|
| 74 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 75 |
"step": 10,
|
| 76 |
-
"step_time":
|
| 77 |
},
|
| 78 |
{
|
| 79 |
-
"clip_ratio/high_max": 0.
|
| 80 |
-
"clip_ratio/high_mean": 0.
|
| 81 |
-
"clip_ratio/low_mean": 0.
|
| 82 |
"clip_ratio/low_min": 0.0,
|
| 83 |
-
"clip_ratio/region_mean": 0.
|
| 84 |
"completions/clipped_ratio": 0.0,
|
| 85 |
-
"completions/max_length":
|
| 86 |
-
"completions/max_terminated_length":
|
| 87 |
-
"completions/mean_length":
|
| 88 |
-
"completions/mean_terminated_length":
|
| 89 |
-
"completions/min_length":
|
| 90 |
-
"completions/min_terminated_length":
|
| 91 |
-
"entropy": 0.
|
| 92 |
"epoch": 0.0024,
|
| 93 |
-
"frac_reward_zero_std": 0.
|
| 94 |
-
"grad_norm": 0.
|
| 95 |
-
"kl": 0.
|
| 96 |
"learning_rate": 3.9802559999999995e-06,
|
| 97 |
-
"loss": 0.
|
| 98 |
-
"num_tokens":
|
| 99 |
-
"reward": 0.
|
| 100 |
-
"reward_std": 0.
|
| 101 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 102 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 103 |
-
"sampling/importance_sampling_ratio/max": 1.
|
| 104 |
-
"sampling/importance_sampling_ratio/mean": 0.
|
| 105 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 106 |
-
"sampling/sampling_logp_difference/max":
|
| 107 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 108 |
"step": 15,
|
| 109 |
-
"step_time":
|
| 110 |
},
|
| 111 |
{
|
| 112 |
-
"clip_ratio/high_max": 0.
|
| 113 |
-
"clip_ratio/high_mean": 0.
|
| 114 |
-
"clip_ratio/low_mean": 0.
|
| 115 |
"clip_ratio/low_min": 0.0,
|
| 116 |
-
"clip_ratio/region_mean": 0.
|
| 117 |
"completions/clipped_ratio": 0.0,
|
| 118 |
-
"completions/max_length":
|
| 119 |
-
"completions/max_terminated_length":
|
| 120 |
-
"completions/mean_length":
|
| 121 |
-
"completions/mean_terminated_length":
|
| 122 |
-
"completions/min_length":
|
| 123 |
-
"completions/min_terminated_length":
|
| 124 |
-
"entropy": 0.
|
| 125 |
"epoch": 0.0032,
|
| 126 |
-
"frac_reward_zero_std": 0.
|
| 127 |
-
"grad_norm": 0.
|
| 128 |
-
"kl": 0.
|
| 129 |
"learning_rate": 5.401775999999999e-06,
|
| 130 |
-
"loss": -0.
|
| 131 |
-
"num_tokens":
|
| 132 |
-
"reward": 0.
|
| 133 |
-
"reward_std": 0.
|
| 134 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 135 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 136 |
-
"sampling/importance_sampling_ratio/max":
|
| 137 |
-
"sampling/importance_sampling_ratio/mean":
|
| 138 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 139 |
-
"sampling/sampling_logp_difference/max":
|
| 140 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 141 |
"step": 20,
|
| 142 |
-
"step_time":
|
| 143 |
},
|
| 144 |
{
|
| 145 |
-
"clip_ratio/high_max": 0.
|
| 146 |
-
"clip_ratio/high_mean": 0.
|
| 147 |
-
"clip_ratio/low_mean": 0.
|
| 148 |
-
"clip_ratio/low_min": 0.
|
| 149 |
-
"clip_ratio/region_mean": 0.
|
| 150 |
"completions/clipped_ratio": 0.0,
|
| 151 |
-
"completions/max_length": 374.
|
| 152 |
-
"completions/max_terminated_length": 374.
|
| 153 |
-
"completions/mean_length":
|
| 154 |
-
"completions/mean_terminated_length":
|
| 155 |
-
"completions/min_length":
|
| 156 |
-
"completions/min_terminated_length":
|
| 157 |
-
"entropy": 0.
|
| 158 |
"epoch": 0.004,
|
| 159 |
-
"frac_reward_zero_std": 0.
|
| 160 |
-
"grad_norm": 0.
|
| 161 |
-
"kl": 0.
|
| 162 |
"learning_rate": 6.8232959999999994e-06,
|
| 163 |
-
"loss":
|
| 164 |
-
"num_tokens":
|
| 165 |
-
"reward": 0.
|
| 166 |
-
"reward_std": 0.
|
| 167 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 168 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 169 |
-
"sampling/importance_sampling_ratio/max":
|
| 170 |
-
"sampling/importance_sampling_ratio/mean": 0.
|
| 171 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 172 |
-
"sampling/sampling_logp_difference/max":
|
| 173 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 174 |
"step": 25,
|
| 175 |
-
"step_time":
|
| 176 |
},
|
| 177 |
{
|
| 178 |
-
"clip_ratio/high_max": 0.
|
| 179 |
-
"clip_ratio/high_mean": 0.
|
| 180 |
-
"clip_ratio/low_mean": 0.
|
| 181 |
-
"clip_ratio/low_min": 0.
|
| 182 |
-
"clip_ratio/region_mean": 0.
|
| 183 |
"completions/clipped_ratio": 0.0,
|
| 184 |
-
"completions/max_length":
|
| 185 |
-
"completions/max_terminated_length":
|
| 186 |
-
"completions/mean_length": 283.
|
| 187 |
-
"completions/mean_terminated_length": 283.
|
| 188 |
"completions/min_length": 212.0,
|
| 189 |
"completions/min_terminated_length": 212.0,
|
| 190 |
-
"entropy": 0.
|
| 191 |
"epoch": 0.0048,
|
| 192 |
-
"frac_reward_zero_std": 0.
|
| 193 |
-
"grad_norm": 0.
|
| 194 |
-
"kl":
|
| 195 |
"learning_rate": 8.244816e-06,
|
| 196 |
-
"loss": 3.
|
| 197 |
-
"num_tokens":
|
| 198 |
-
"reward": 0.
|
| 199 |
-
"reward_std": 0.
|
| 200 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 201 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 202 |
-
"sampling/importance_sampling_ratio/max":
|
| 203 |
-
"sampling/importance_sampling_ratio/mean": 0.
|
| 204 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 205 |
-
"sampling/sampling_logp_difference/max":
|
| 206 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 207 |
"step": 30,
|
| 208 |
-
"step_time":
|
| 209 |
},
|
| 210 |
{
|
| 211 |
-
"clip_ratio/high_max": 0.
|
| 212 |
-
"clip_ratio/high_mean": 0.
|
| 213 |
-
"clip_ratio/low_mean": 0.
|
| 214 |
-
"clip_ratio/low_min": 0.
|
| 215 |
-
"clip_ratio/region_mean": 0.
|
| 216 |
"completions/clipped_ratio": 0.0,
|
| 217 |
-
"completions/max_length":
|
| 218 |
-
"completions/max_terminated_length":
|
| 219 |
-
"completions/mean_length": 290.
|
| 220 |
-
"completions/mean_terminated_length": 290.
|
| 221 |
"completions/min_length": 212.0,
|
| 222 |
"completions/min_terminated_length": 212.0,
|
| 223 |
-
"entropy": 0.
|
| 224 |
"epoch": 0.0056,
|
| 225 |
-
"frac_reward_zero_std": 0.
|
| 226 |
-
"grad_norm": 0.
|
| 227 |
-
"kl":
|
| 228 |
"learning_rate": 9.666336e-06,
|
| 229 |
-
"loss": -
|
| 230 |
-
"num_tokens":
|
| 231 |
-
"reward": 0.
|
| 232 |
-
"reward_std": 0.
|
| 233 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 234 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 235 |
-
"sampling/importance_sampling_ratio/max":
|
| 236 |
-
"sampling/importance_sampling_ratio/mean": 0.
|
| 237 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 238 |
-
"sampling/sampling_logp_difference/max":
|
| 239 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 240 |
"step": 35,
|
| 241 |
-
"step_time":
|
| 242 |
},
|
| 243 |
{
|
| 244 |
-
"clip_ratio/high_max": 0.
|
| 245 |
-
"clip_ratio/high_mean": 0.
|
| 246 |
-
"clip_ratio/low_mean": 0.
|
| 247 |
"clip_ratio/low_min": 0.0,
|
| 248 |
-
"clip_ratio/region_mean": 0.
|
| 249 |
"completions/clipped_ratio": 0.0,
|
| 250 |
-
"completions/max_length":
|
| 251 |
-
"completions/max_terminated_length":
|
| 252 |
-
"completions/mean_length": 289.
|
| 253 |
-
"completions/mean_terminated_length": 289.
|
| 254 |
-
"completions/min_length":
|
| 255 |
-
"completions/min_terminated_length":
|
| 256 |
-
"entropy": 0.
|
| 257 |
"epoch": 0.0064,
|
| 258 |
-
"frac_reward_zero_std": 0.
|
| 259 |
-
"grad_norm": 0.
|
| 260 |
-
"kl": 0.
|
| 261 |
"learning_rate": 9.95063915881342e-06,
|
| 262 |
-
"loss": 0.
|
| 263 |
-
"num_tokens":
|
| 264 |
-
"reward": 0.
|
| 265 |
-
"reward_std": 0.
|
| 266 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 267 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 268 |
-
"sampling/importance_sampling_ratio/max":
|
| 269 |
-
"sampling/importance_sampling_ratio/mean":
|
| 270 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 271 |
-
"sampling/sampling_logp_difference/max":
|
| 272 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 273 |
"step": 40,
|
| 274 |
-
"step_time":
|
| 275 |
},
|
| 276 |
{
|
| 277 |
-
"clip_ratio/high_max": 0.
|
| 278 |
-
"clip_ratio/high_mean": 0.
|
| 279 |
-
"clip_ratio/low_mean": 0.
|
| 280 |
-
"clip_ratio/low_min": 0.
|
| 281 |
-
"clip_ratio/region_mean": 0.
|
| 282 |
-
"completions/clipped_ratio": 0.
|
| 283 |
-
"completions/max_length":
|
| 284 |
-
"completions/max_terminated_length": 374.
|
| 285 |
-
"completions/mean_length":
|
| 286 |
-
"completions/mean_terminated_length":
|
| 287 |
-
"completions/min_length":
|
| 288 |
-
"completions/min_terminated_length":
|
| 289 |
-
"entropy": 0.
|
| 290 |
"epoch": 0.0072,
|
| 291 |
-
"frac_reward_zero_std": 0.
|
| 292 |
-
"grad_norm": 0.
|
| 293 |
-
"kl": 0.
|
| 294 |
"learning_rate": 9.950635741493589e-06,
|
| 295 |
-
"loss": 0.
|
| 296 |
-
"num_tokens":
|
| 297 |
-
"reward": 0.
|
| 298 |
-
"reward_std": 0.
|
| 299 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 300 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 301 |
-
"sampling/importance_sampling_ratio/max": 2.
|
| 302 |
-
"sampling/importance_sampling_ratio/mean":
|
| 303 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 304 |
-
"sampling/sampling_logp_difference/max":
|
| 305 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 306 |
"step": 45,
|
| 307 |
-
"step_time":
|
| 308 |
},
|
| 309 |
{
|
| 310 |
-
"clip_ratio/high_max": 0.
|
| 311 |
-
"clip_ratio/high_mean": 0.
|
| 312 |
-
"clip_ratio/low_mean": 0.
|
| 313 |
-
"clip_ratio/low_min": 0.
|
| 314 |
-
"clip_ratio/region_mean": 0.
|
| 315 |
"completions/clipped_ratio": 0.0,
|
| 316 |
-
"completions/max_length":
|
| 317 |
-
"completions/max_terminated_length":
|
| 318 |
-
"completions/mean_length":
|
| 319 |
-
"completions/mean_terminated_length":
|
| 320 |
-
"completions/min_length":
|
| 321 |
-
"completions/min_terminated_length":
|
| 322 |
-
"entropy": 0.
|
| 323 |
"epoch": 0.008,
|
| 324 |
-
"frac_reward_zero_std": 0.
|
| 325 |
-
"grad_norm": 0.
|
| 326 |
-
"kl":
|
| 327 |
"learning_rate": 9.950629695468755e-06,
|
| 328 |
-
"loss": 0.
|
| 329 |
-
"num_tokens":
|
| 330 |
-
"reward": 0.
|
| 331 |
-
"reward_std": 0.
|
| 332 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 333 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 334 |
-
"sampling/importance_sampling_ratio/max":
|
| 335 |
-
"sampling/importance_sampling_ratio/mean": 0.
|
| 336 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 337 |
-
"sampling/sampling_logp_difference/max":
|
| 338 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 339 |
"step": 50,
|
| 340 |
-
"step_time":
|
| 341 |
},
|
| 342 |
{
|
| 343 |
-
"clip_ratio/high_max": 0.
|
| 344 |
-
"clip_ratio/high_mean": 0.
|
| 345 |
-
"clip_ratio/low_mean": 0.
|
| 346 |
"clip_ratio/low_min": 0.0,
|
| 347 |
-
"clip_ratio/region_mean": 0.
|
| 348 |
"completions/clipped_ratio": 0.0,
|
| 349 |
-
"completions/max_length":
|
| 350 |
-
"completions/max_terminated_length":
|
| 351 |
-
"completions/mean_length": 284.
|
| 352 |
-
"completions/mean_terminated_length": 284.
|
| 353 |
-
"completions/min_length":
|
| 354 |
-
"completions/min_terminated_length":
|
| 355 |
-
"entropy": 0.
|
| 356 |
"epoch": 0.0088,
|
| 357 |
-
"frac_reward_zero_std": 0.
|
| 358 |
-
"grad_norm": 0.
|
| 359 |
-
"kl": 0.
|
| 360 |
"learning_rate": 9.950621020743173e-06,
|
| 361 |
-
"loss":
|
| 362 |
-
"num_tokens":
|
| 363 |
-
"reward": 0.
|
| 364 |
-
"reward_std": 0.
|
| 365 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 366 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 367 |
-
"sampling/importance_sampling_ratio/max":
|
| 368 |
-
"sampling/importance_sampling_ratio/mean": 0.
|
| 369 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 370 |
-
"sampling/sampling_logp_difference/max":
|
| 371 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 372 |
"step": 55,
|
| 373 |
-
"step_time":
|
| 374 |
},
|
| 375 |
{
|
| 376 |
-
"clip_ratio/high_max": 0.
|
| 377 |
-
"clip_ratio/high_mean": 0.
|
| 378 |
-
"clip_ratio/low_mean": 0.
|
| 379 |
"clip_ratio/low_min": 0.0,
|
| 380 |
-
"clip_ratio/region_mean": 0.
|
| 381 |
"completions/clipped_ratio": 0.0,
|
| 382 |
-
"completions/max_length":
|
| 383 |
-
"completions/max_terminated_length":
|
| 384 |
-
"completions/mean_length": 294.
|
| 385 |
-
"completions/mean_terminated_length": 294.
|
| 386 |
-
"completions/min_length":
|
| 387 |
-
"completions/min_terminated_length":
|
| 388 |
-
"entropy": 0.
|
| 389 |
"epoch": 0.0096,
|
| 390 |
-
"frac_reward_zero_std": 0.
|
| 391 |
-
"grad_norm": 0.
|
| 392 |
-
"kl": 0.
|
| 393 |
"learning_rate": 9.950609717322956e-06,
|
| 394 |
-
"loss": 0.
|
| 395 |
-
"num_tokens":
|
| 396 |
-
"reward": 0.
|
| 397 |
-
"reward_std": 0.
|
| 398 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 399 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 400 |
-
"sampling/importance_sampling_ratio/max":
|
| 401 |
-
"sampling/importance_sampling_ratio/mean": 1.
|
| 402 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 403 |
-
"sampling/sampling_logp_difference/max":
|
| 404 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 405 |
"step": 60,
|
| 406 |
-
"step_time":
|
| 407 |
},
|
| 408 |
{
|
| 409 |
-
"clip_ratio/high_max": 0.
|
| 410 |
-
"clip_ratio/high_mean": 0.
|
| 411 |
-
"clip_ratio/low_mean": 0.
|
| 412 |
"clip_ratio/low_min": 0.0,
|
| 413 |
-
"clip_ratio/region_mean": 0.
|
| 414 |
"completions/clipped_ratio": 0.0,
|
| 415 |
-
"completions/max_length":
|
| 416 |
-
"completions/max_terminated_length":
|
| 417 |
-
"completions/mean_length":
|
| 418 |
-
"completions/mean_terminated_length":
|
| 419 |
-
"completions/min_length":
|
| 420 |
-
"completions/min_terminated_length":
|
| 421 |
-
"entropy": 0.
|
| 422 |
"epoch": 0.0104,
|
| 423 |
-
"frac_reward_zero_std": 0.
|
| 424 |
-
"grad_norm": 0.
|
| 425 |
-
"kl": 0.
|
| 426 |
"learning_rate": 9.950595785216067e-06,
|
| 427 |
-
"loss":
|
| 428 |
-
"num_tokens":
|
| 429 |
-
"reward": 0.
|
| 430 |
-
"reward_std": 0.
|
| 431 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 432 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 433 |
-
"sampling/importance_sampling_ratio/max":
|
| 434 |
-
"sampling/importance_sampling_ratio/mean":
|
| 435 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 436 |
-
"sampling/sampling_logp_difference/max":
|
| 437 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 438 |
"step": 65,
|
| 439 |
-
"step_time":
|
| 440 |
},
|
| 441 |
{
|
| 442 |
-
"clip_ratio/high_max": 0.
|
| 443 |
-
"clip_ratio/high_mean": 0.
|
| 444 |
-
"clip_ratio/low_mean": 0.
|
| 445 |
"clip_ratio/low_min": 0.0,
|
| 446 |
-
"clip_ratio/region_mean": 0.
|
| 447 |
"completions/clipped_ratio": 0.0,
|
| 448 |
-
"completions/max_length":
|
| 449 |
-
"completions/max_terminated_length":
|
| 450 |
-
"completions/mean_length":
|
| 451 |
-
"completions/mean_terminated_length":
|
| 452 |
-
"completions/min_length":
|
| 453 |
-
"completions/min_terminated_length":
|
| 454 |
-
"entropy": 0.
|
| 455 |
"epoch": 0.0112,
|
| 456 |
-
"frac_reward_zero_std": 0.
|
| 457 |
-
"grad_norm": 0.
|
| 458 |
-
"kl":
|
| 459 |
"learning_rate": 9.950579224432321e-06,
|
| 460 |
-
"loss":
|
| 461 |
-
"num_tokens":
|
| 462 |
-
"reward": 0.
|
| 463 |
-
"reward_std": 0.
|
| 464 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 465 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 466 |
-
"sampling/importance_sampling_ratio/max":
|
| 467 |
-
"sampling/importance_sampling_ratio/mean": 0.
|
| 468 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 469 |
-
"sampling/sampling_logp_difference/max":
|
| 470 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 471 |
"step": 70,
|
| 472 |
-
"step_time": 5.
|
| 473 |
},
|
| 474 |
{
|
| 475 |
-
"clip_ratio/high_max": 0.
|
| 476 |
-
"clip_ratio/high_mean": 0.
|
| 477 |
-
"clip_ratio/low_mean": 0.
|
| 478 |
"clip_ratio/low_min": 0.0,
|
| 479 |
-
"clip_ratio/region_mean": 0.
|
| 480 |
"completions/clipped_ratio": 0.0,
|
| 481 |
-
"completions/max_length":
|
| 482 |
-
"completions/max_terminated_length":
|
| 483 |
-
"completions/mean_length":
|
| 484 |
-
"completions/mean_terminated_length":
|
| 485 |
-
"completions/min_length":
|
| 486 |
-
"completions/min_terminated_length":
|
| 487 |
-
"entropy": 0.
|
| 488 |
"epoch": 0.012,
|
| 489 |
-
"frac_reward_zero_std": 0.
|
| 490 |
-
"grad_norm": 0.
|
| 491 |
-
"kl": 0.
|
| 492 |
"learning_rate": 9.950560034983382e-06,
|
| 493 |
-
"loss":
|
| 494 |
-
"num_tokens":
|
| 495 |
-
"reward": 0.
|
| 496 |
-
"reward_std": 0.
|
| 497 |
-
"rewards/env_goofspiel_reward/mean": 0.
|
| 498 |
-
"rewards/env_goofspiel_reward/std": 0.
|
| 499 |
-
"sampling/importance_sampling_ratio/max":
|
| 500 |
-
"sampling/importance_sampling_ratio/mean": 0.
|
| 501 |
-
"sampling/importance_sampling_ratio/min": 0.
|
| 502 |
-
"sampling/sampling_logp_difference/max":
|
| 503 |
-
"sampling/sampling_logp_difference/mean": 0.
|
| 504 |
"step": 75,
|
| 505 |
-
"step_time": 5.
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.012,
|
|
@@ -512,35 +512,596 @@
|
|
| 512 |
"eval_clip_ratio/low_min": 0.0,
|
| 513 |
"eval_clip_ratio/region_mean": 0.0,
|
| 514 |
"eval_completions/clipped_ratio": 0.0,
|
| 515 |
-
"eval_completions/max_length":
|
| 516 |
-
"eval_completions/max_terminated_length":
|
| 517 |
-
"eval_completions/mean_length":
|
| 518 |
-
"eval_completions/mean_terminated_length":
|
| 519 |
-
"eval_completions/min_length":
|
| 520 |
-
"eval_completions/min_terminated_length":
|
| 521 |
-
"eval_entropy": 0.
|
| 522 |
-
"eval_frac_reward_zero_std": 0.
|
| 523 |
-
"eval_kl": 0.
|
| 524 |
-
"eval_loss": -
|
| 525 |
-
"eval_num_tokens":
|
| 526 |
-
"eval_reward": 0.
|
| 527 |
-
"eval_reward_std": 0.
|
| 528 |
-
"eval_rewards/env_goofspiel_reward/mean": 0.
|
| 529 |
-
"eval_rewards/env_goofspiel_reward/std": 0.
|
| 530 |
-
"eval_runtime": 2.
|
| 531 |
-
"eval_samples_per_second": 3.
|
| 532 |
-
"eval_sampling/importance_sampling_ratio/max": 1.
|
| 533 |
-
"eval_sampling/importance_sampling_ratio/mean":
|
| 534 |
-
"eval_sampling/importance_sampling_ratio/min": 0.
|
| 535 |
-
"eval_sampling/sampling_logp_difference/max":
|
| 536 |
-
"eval_sampling/sampling_logp_difference/mean": 0.
|
| 537 |
-
"eval_steps_per_second": 1.
|
| 538 |
"step": 75
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 539 |
}
|
| 540 |
],
|
| 541 |
"logging_steps": 5,
|
| 542 |
"max_steps": 18750,
|
| 543 |
-
"num_input_tokens_seen":
|
| 544 |
"num_train_epochs": 3,
|
| 545 |
"save_steps": 500,
|
| 546 |
"stateful_callbacks": {
|
|
@@ -550,13 +1111,13 @@
|
|
| 550 |
"should_evaluate": false,
|
| 551 |
"should_log": false,
|
| 552 |
"should_save": true,
|
| 553 |
-
"should_training_stop":
|
| 554 |
},
|
| 555 |
"attributes": {}
|
| 556 |
}
|
| 557 |
},
|
| 558 |
"total_flos": 0.0,
|
| 559 |
-
"train_batch_size":
|
| 560 |
"trial_name": null,
|
| 561 |
"trial_params": null
|
| 562 |
}
|
|
|
|
| 2 |
"best_global_step": null,
|
| 3 |
"best_metric": null,
|
| 4 |
"best_model_checkpoint": null,
|
| 5 |
+
"epoch": 0.02432,
|
| 6 |
"eval_steps": 500,
|
| 7 |
+
"global_step": 152,
|
| 8 |
"is_hyper_param_search": false,
|
| 9 |
"is_local_process_zero": true,
|
| 10 |
"is_world_process_zero": true,
|
| 11 |
"log_history": [
|
| 12 |
{
|
| 13 |
+
"clip_ratio/high_max": 0.0,
|
| 14 |
+
"clip_ratio/high_mean": 0.0,
|
| 15 |
+
"clip_ratio/low_mean": 0.0,
|
| 16 |
"clip_ratio/low_min": 0.0,
|
| 17 |
+
"clip_ratio/region_mean": 0.0,
|
| 18 |
"completions/clipped_ratio": 0.0,
|
| 19 |
+
"completions/max_length": 374.4,
|
| 20 |
+
"completions/max_terminated_length": 374.4,
|
| 21 |
+
"completions/mean_length": 293.55,
|
| 22 |
+
"completions/mean_terminated_length": 293.55,
|
| 23 |
+
"completions/min_length": 181.8,
|
| 24 |
+
"completions/min_terminated_length": 181.8,
|
| 25 |
+
"entropy": 0.7695603430271148,
|
| 26 |
"epoch": 0.0008,
|
| 27 |
+
"frac_reward_zero_std": 0.675,
|
| 28 |
+
"grad_norm": 0.1396484375,
|
| 29 |
+
"kl": 0.007634526048786938,
|
| 30 |
"learning_rate": 1.137216e-06,
|
| 31 |
+
"loss": 9.18293430004269e-05,
|
| 32 |
+
"num_tokens": 135798.0,
|
| 33 |
+
"reward": 0.051062504202127455,
|
| 34 |
+
"reward_std": 0.07574881687760353,
|
| 35 |
+
"rewards/env_goofspiel_reward/mean": 0.051062504202127455,
|
| 36 |
+
"rewards/env_goofspiel_reward/std": 0.195309117436409,
|
| 37 |
+
"sampling/importance_sampling_ratio/max": 1.719690752029419,
|
| 38 |
+
"sampling/importance_sampling_ratio/mean": 1.0014273881912232,
|
| 39 |
+
"sampling/importance_sampling_ratio/min": 0.48566103279590606,
|
| 40 |
+
"sampling/sampling_logp_difference/max": 0.7150738000869751,
|
| 41 |
+
"sampling/sampling_logp_difference/mean": 0.0632629081606865,
|
| 42 |
"step": 5,
|
| 43 |
+
"step_time": 4.057331793400226
|
| 44 |
},
|
| 45 |
{
|
| 46 |
+
"clip_ratio/high_max": 0.0,
|
| 47 |
+
"clip_ratio/high_mean": 0.0,
|
| 48 |
+
"clip_ratio/low_mean": 0.0,
|
| 49 |
"clip_ratio/low_min": 0.0,
|
| 50 |
+
"clip_ratio/region_mean": 0.0,
|
| 51 |
"completions/clipped_ratio": 0.0,
|
| 52 |
+
"completions/max_length": 374.0,
|
| 53 |
+
"completions/max_terminated_length": 374.0,
|
| 54 |
+
"completions/mean_length": 290.5375,
|
| 55 |
+
"completions/mean_terminated_length": 290.5375,
|
| 56 |
+
"completions/min_length": 193.0,
|
| 57 |
+
"completions/min_terminated_length": 193.0,
|
| 58 |
+
"entropy": 0.6640829563140869,
|
| 59 |
"epoch": 0.0016,
|
| 60 |
+
"frac_reward_zero_std": 0.7375,
|
| 61 |
+
"grad_norm": 0.296875,
|
| 62 |
+
"kl": 0.009508041350636631,
|
| 63 |
"learning_rate": 2.5587359999999995e-06,
|
| 64 |
+
"loss": -5.3186528384685515e-05,
|
| 65 |
+
"num_tokens": 271318.0,
|
| 66 |
+
"reward": 0.05906250327825546,
|
| 67 |
+
"reward_std": 0.06461188569664955,
|
| 68 |
+
"rewards/env_goofspiel_reward/mean": 0.05906250327825546,
|
| 69 |
+
"rewards/env_goofspiel_reward/std": 0.16751175224781037,
|
| 70 |
+
"sampling/importance_sampling_ratio/max": 2.0673105001449583,
|
| 71 |
+
"sampling/importance_sampling_ratio/mean": 1.0338432788848877,
|
| 72 |
+
"sampling/importance_sampling_ratio/min": 0.6319645524024964,
|
| 73 |
+
"sampling/sampling_logp_difference/max": 0.6566181659698487,
|
| 74 |
+
"sampling/sampling_logp_difference/mean": 0.05573496893048287,
|
| 75 |
"step": 10,
|
| 76 |
+
"step_time": 3.559387542600052
|
| 77 |
},
|
| 78 |
{
|
| 79 |
+
"clip_ratio/high_max": 0.0,
|
| 80 |
+
"clip_ratio/high_mean": 0.0,
|
| 81 |
+
"clip_ratio/low_mean": 0.0,
|
| 82 |
"clip_ratio/low_min": 0.0,
|
| 83 |
+
"clip_ratio/region_mean": 0.0,
|
| 84 |
"completions/clipped_ratio": 0.0,
|
| 85 |
+
"completions/max_length": 378.8,
|
| 86 |
+
"completions/max_terminated_length": 378.8,
|
| 87 |
+
"completions/mean_length": 285.15,
|
| 88 |
+
"completions/mean_terminated_length": 285.15,
|
| 89 |
+
"completions/min_length": 212.0,
|
| 90 |
+
"completions/min_terminated_length": 212.0,
|
| 91 |
+
"entropy": 0.5863359421491623,
|
| 92 |
"epoch": 0.0024,
|
| 93 |
+
"frac_reward_zero_std": 0.8,
|
| 94 |
+
"grad_norm": 0.1650390625,
|
| 95 |
+
"kl": 0.05467459289357066,
|
| 96 |
"learning_rate": 3.9802559999999995e-06,
|
| 97 |
+
"loss": 0.00011974496301263571,
|
| 98 |
+
"num_tokens": 403708.0,
|
| 99 |
+
"reward": 0.037000001792330296,
|
| 100 |
+
"reward_std": 0.0535633388790302,
|
| 101 |
+
"rewards/env_goofspiel_reward/mean": 0.037000001792330296,
|
| 102 |
+
"rewards/env_goofspiel_reward/std": 0.11216981350444258,
|
| 103 |
+
"sampling/importance_sampling_ratio/max": 1.5559733867645265,
|
| 104 |
+
"sampling/importance_sampling_ratio/mean": 0.9780893206596375,
|
| 105 |
+
"sampling/importance_sampling_ratio/min": 0.604676628112793,
|
| 106 |
+
"sampling/sampling_logp_difference/max": 0.4918365955352783,
|
| 107 |
+
"sampling/sampling_logp_difference/mean": 0.04973898231983185,
|
| 108 |
"step": 15,
|
| 109 |
+
"step_time": 3.5593893944003865
|
| 110 |
},
|
| 111 |
{
|
| 112 |
+
"clip_ratio/high_max": 0.0,
|
| 113 |
+
"clip_ratio/high_mean": 0.0,
|
| 114 |
+
"clip_ratio/low_mean": 0.0,
|
| 115 |
"clip_ratio/low_min": 0.0,
|
| 116 |
+
"clip_ratio/region_mean": 0.0,
|
| 117 |
"completions/clipped_ratio": 0.0,
|
| 118 |
+
"completions/max_length": 374.0,
|
| 119 |
+
"completions/max_terminated_length": 374.0,
|
| 120 |
+
"completions/mean_length": 281.6375,
|
| 121 |
+
"completions/mean_terminated_length": 281.6375,
|
| 122 |
+
"completions/min_length": 212.0,
|
| 123 |
+
"completions/min_terminated_length": 212.0,
|
| 124 |
+
"entropy": 0.4126308411359787,
|
| 125 |
"epoch": 0.0032,
|
| 126 |
+
"frac_reward_zero_std": 0.7375,
|
| 127 |
+
"grad_norm": 0.31640625,
|
| 128 |
+
"kl": 0.139003873616457,
|
| 129 |
"learning_rate": 5.401775999999999e-06,
|
| 130 |
+
"loss": -0.00015898187411949037,
|
| 131 |
+
"num_tokens": 536243.0,
|
| 132 |
+
"reward": 0.06343750283122063,
|
| 133 |
+
"reward_std": 0.0905980572104454,
|
| 134 |
+
"rewards/env_goofspiel_reward/mean": 0.06343750283122063,
|
| 135 |
+
"rewards/env_goofspiel_reward/std": 0.19052477180957794,
|
| 136 |
+
"sampling/importance_sampling_ratio/max": 1.6631618976593017,
|
| 137 |
+
"sampling/importance_sampling_ratio/mean": 1.0170260667800903,
|
| 138 |
+
"sampling/importance_sampling_ratio/min": 0.643394160270691,
|
| 139 |
+
"sampling/sampling_logp_difference/max": 0.5536829710006714,
|
| 140 |
+
"sampling/sampling_logp_difference/mean": 0.03567908257246018,
|
| 141 |
"step": 20,
|
| 142 |
+
"step_time": 3.513668759400389
|
| 143 |
},
|
| 144 |
{
|
| 145 |
+
"clip_ratio/high_max": 0.0,
|
| 146 |
+
"clip_ratio/high_mean": 0.0,
|
| 147 |
+
"clip_ratio/low_mean": 0.0,
|
| 148 |
+
"clip_ratio/low_min": 0.0,
|
| 149 |
+
"clip_ratio/region_mean": 0.0,
|
| 150 |
"completions/clipped_ratio": 0.0,
|
| 151 |
+
"completions/max_length": 374.0,
|
| 152 |
+
"completions/max_terminated_length": 374.0,
|
| 153 |
+
"completions/mean_length": 300.975,
|
| 154 |
+
"completions/mean_terminated_length": 300.975,
|
| 155 |
+
"completions/min_length": 219.0,
|
| 156 |
+
"completions/min_terminated_length": 219.0,
|
| 157 |
+
"entropy": 0.5057466819882392,
|
| 158 |
"epoch": 0.004,
|
| 159 |
+
"frac_reward_zero_std": 0.8625,
|
| 160 |
+
"grad_norm": 0.201171875,
|
| 161 |
+
"kl": 0.14662861209362746,
|
| 162 |
"learning_rate": 6.8232959999999994e-06,
|
| 163 |
+
"loss": 0.00011511286720633507,
|
| 164 |
+
"num_tokens": 674802.0,
|
| 165 |
+
"reward": 0.056125002296175806,
|
| 166 |
+
"reward_std": 0.05851308616111055,
|
| 167 |
+
"rewards/env_goofspiel_reward/mean": 0.056125002296175806,
|
| 168 |
+
"rewards/env_goofspiel_reward/std": 0.1724803472403437,
|
| 169 |
+
"sampling/importance_sampling_ratio/max": 1.8632315158843995,
|
| 170 |
+
"sampling/importance_sampling_ratio/mean": 0.9876452207565307,
|
| 171 |
+
"sampling/importance_sampling_ratio/min": 0.5998193681240082,
|
| 172 |
+
"sampling/sampling_logp_difference/max": 0.5798031091690063,
|
| 173 |
+
"sampling/sampling_logp_difference/mean": 0.049172034859657286,
|
| 174 |
"step": 25,
|
| 175 |
+
"step_time": 3.5666124442001093
|
| 176 |
},
|
| 177 |
{
|
| 178 |
+
"clip_ratio/high_max": 0.0,
|
| 179 |
+
"clip_ratio/high_mean": 0.0,
|
| 180 |
+
"clip_ratio/low_mean": 0.0,
|
| 181 |
+
"clip_ratio/low_min": 0.0,
|
| 182 |
+
"clip_ratio/region_mean": 0.0,
|
| 183 |
"completions/clipped_ratio": 0.0,
|
| 184 |
+
"completions/max_length": 373.8,
|
| 185 |
+
"completions/max_terminated_length": 373.8,
|
| 186 |
+
"completions/mean_length": 283.69375,
|
| 187 |
+
"completions/mean_terminated_length": 283.69375,
|
| 188 |
"completions/min_length": 212.0,
|
| 189 |
"completions/min_terminated_length": 212.0,
|
| 190 |
+
"entropy": 0.5599821880459785,
|
| 191 |
"epoch": 0.0048,
|
| 192 |
+
"frac_reward_zero_std": 0.775,
|
| 193 |
+
"grad_norm": 0.30859375,
|
| 194 |
+
"kl": 0.14428229751065375,
|
| 195 |
"learning_rate": 8.244816e-06,
|
| 196 |
+
"loss": -3.07892682030797e-05,
|
| 197 |
+
"num_tokens": 807598.0,
|
| 198 |
+
"reward": 0.08625000044703483,
|
| 199 |
+
"reward_std": 0.10076271444559097,
|
| 200 |
+
"rewards/env_goofspiel_reward/mean": 0.08625000044703483,
|
| 201 |
+
"rewards/env_goofspiel_reward/std": 0.21401541233062743,
|
| 202 |
+
"sampling/importance_sampling_ratio/max": 1.4588377714157104,
|
| 203 |
+
"sampling/importance_sampling_ratio/mean": 0.9682739973068237,
|
| 204 |
+
"sampling/importance_sampling_ratio/min": 0.5809138238430023,
|
| 205 |
+
"sampling/sampling_logp_difference/max": 0.5705435633659363,
|
| 206 |
+
"sampling/sampling_logp_difference/mean": 0.04744169861078262,
|
| 207 |
"step": 30,
|
| 208 |
+
"step_time": 3.545922210400022
|
| 209 |
},
|
| 210 |
{
|
| 211 |
+
"clip_ratio/high_max": 0.0,
|
| 212 |
+
"clip_ratio/high_mean": 0.0,
|
| 213 |
+
"clip_ratio/low_mean": 0.0,
|
| 214 |
+
"clip_ratio/low_min": 0.0,
|
| 215 |
+
"clip_ratio/region_mean": 0.0,
|
| 216 |
"completions/clipped_ratio": 0.0,
|
| 217 |
+
"completions/max_length": 365.6,
|
| 218 |
+
"completions/max_terminated_length": 365.6,
|
| 219 |
+
"completions/mean_length": 290.56875,
|
| 220 |
+
"completions/mean_terminated_length": 290.56875,
|
| 221 |
"completions/min_length": 212.0,
|
| 222 |
"completions/min_terminated_length": 212.0,
|
| 223 |
+
"entropy": 0.5238839864730835,
|
| 224 |
"epoch": 0.0056,
|
| 225 |
+
"frac_reward_zero_std": 0.9,
|
| 226 |
+
"grad_norm": 0.0791015625,
|
| 227 |
+
"kl": 0.22365115247666836,
|
| 228 |
"learning_rate": 9.666336e-06,
|
| 229 |
+
"loss": -1.7423409735783936e-05,
|
| 230 |
+
"num_tokens": 941209.0,
|
| 231 |
+
"reward": 0.04125000163912773,
|
| 232 |
+
"reward_std": 0.04772970825433731,
|
| 233 |
+
"rewards/env_goofspiel_reward/mean": 0.04125000163912773,
|
| 234 |
+
"rewards/env_goofspiel_reward/std": 0.16298522651195527,
|
| 235 |
+
"sampling/importance_sampling_ratio/max": 1.4255825757980347,
|
| 236 |
+
"sampling/importance_sampling_ratio/mean": 0.973192298412323,
|
| 237 |
+
"sampling/importance_sampling_ratio/min": 0.5944858670234681,
|
| 238 |
+
"sampling/sampling_logp_difference/max": 0.4666349172592163,
|
| 239 |
+
"sampling/sampling_logp_difference/mean": 0.04362327083945274,
|
| 240 |
"step": 35,
|
| 241 |
+
"step_time": 3.448241228200095
|
| 242 |
},
|
| 243 |
{
|
| 244 |
+
"clip_ratio/high_max": 0.0,
|
| 245 |
+
"clip_ratio/high_mean": 0.0,
|
| 246 |
+
"clip_ratio/low_mean": 0.0,
|
| 247 |
"clip_ratio/low_min": 0.0,
|
| 248 |
+
"clip_ratio/region_mean": 0.0,
|
| 249 |
"completions/clipped_ratio": 0.0,
|
| 250 |
+
"completions/max_length": 373.8,
|
| 251 |
+
"completions/max_terminated_length": 373.8,
|
| 252 |
+
"completions/mean_length": 289.4875,
|
| 253 |
+
"completions/mean_terminated_length": 289.4875,
|
| 254 |
+
"completions/min_length": 206.4,
|
| 255 |
+
"completions/min_terminated_length": 206.4,
|
| 256 |
+
"entropy": 0.5457443177700043,
|
| 257 |
"epoch": 0.0064,
|
| 258 |
+
"frac_reward_zero_std": 0.8125,
|
| 259 |
+
"grad_norm": 0.05517578125,
|
| 260 |
+
"kl": 0.21872444674372674,
|
| 261 |
"learning_rate": 9.95063915881342e-06,
|
| 262 |
+
"loss": 0.00012433364754542707,
|
| 263 |
+
"num_tokens": 1076167.0,
|
| 264 |
+
"reward": 0.05981250181794166,
|
| 265 |
+
"reward_std": 0.07451137900352478,
|
| 266 |
+
"rewards/env_goofspiel_reward/mean": 0.05981250181794166,
|
| 267 |
+
"rewards/env_goofspiel_reward/std": 0.1769598752260208,
|
| 268 |
+
"sampling/importance_sampling_ratio/max": 1.5738928318023682,
|
| 269 |
+
"sampling/importance_sampling_ratio/mean": 1.0145540714263916,
|
| 270 |
+
"sampling/importance_sampling_ratio/min": 0.6491193056106568,
|
| 271 |
+
"sampling/sampling_logp_difference/max": 0.46645350456237794,
|
| 272 |
+
"sampling/sampling_logp_difference/mean": 0.04598992168903351,
|
| 273 |
"step": 40,
|
| 274 |
+
"step_time": 3.4551121715998305
|
| 275 |
},
|
| 276 |
{
|
| 277 |
+
"clip_ratio/high_max": 0.0,
|
| 278 |
+
"clip_ratio/high_mean": 0.0,
|
| 279 |
+
"clip_ratio/low_mean": 0.0,
|
| 280 |
+
"clip_ratio/low_min": 0.0,
|
| 281 |
+
"clip_ratio/region_mean": 0.0,
|
| 282 |
+
"completions/clipped_ratio": 0.00625,
|
| 283 |
+
"completions/max_length": 462.0,
|
| 284 |
+
"completions/max_terminated_length": 374.0,
|
| 285 |
+
"completions/mean_length": 300.95625,
|
| 286 |
+
"completions/mean_terminated_length": 297.8118957519531,
|
| 287 |
+
"completions/min_length": 212.0,
|
| 288 |
+
"completions/min_terminated_length": 212.0,
|
| 289 |
+
"entropy": 0.5893479824066162,
|
| 290 |
"epoch": 0.0072,
|
| 291 |
+
"frac_reward_zero_std": 0.7875,
|
| 292 |
+
"grad_norm": 0.224609375,
|
| 293 |
+
"kl": 0.2417328185401857,
|
| 294 |
"learning_rate": 9.950635741493589e-06,
|
| 295 |
+
"loss": 0.02391626685857773,
|
| 296 |
+
"num_tokens": 1212925.0,
|
| 297 |
+
"reward": 0.05993750244379044,
|
| 298 |
+
"reward_std": 0.08494120314717293,
|
| 299 |
+
"rewards/env_goofspiel_reward/mean": 0.05993750244379044,
|
| 300 |
+
"rewards/env_goofspiel_reward/std": 0.17770446538925172,
|
| 301 |
+
"sampling/importance_sampling_ratio/max": 2.169111466407776,
|
| 302 |
+
"sampling/importance_sampling_ratio/mean": 1.020458698272705,
|
| 303 |
+
"sampling/importance_sampling_ratio/min": 0.6463833510875702,
|
| 304 |
+
"sampling/sampling_logp_difference/max": 0.5775727272033692,
|
| 305 |
+
"sampling/sampling_logp_difference/mean": 0.04204721674323082,
|
| 306 |
"step": 45,
|
| 307 |
+
"step_time": 4.194754163199832
|
| 308 |
},
|
| 309 |
{
|
| 310 |
+
"clip_ratio/high_max": 0.0,
|
| 311 |
+
"clip_ratio/high_mean": 0.0,
|
| 312 |
+
"clip_ratio/low_mean": 0.0,
|
| 313 |
+
"clip_ratio/low_min": 0.0,
|
| 314 |
+
"clip_ratio/region_mean": 0.0,
|
| 315 |
"completions/clipped_ratio": 0.0,
|
| 316 |
+
"completions/max_length": 373.4,
|
| 317 |
+
"completions/max_terminated_length": 373.4,
|
| 318 |
+
"completions/mean_length": 299.31875,
|
| 319 |
+
"completions/mean_terminated_length": 299.31875,
|
| 320 |
+
"completions/min_length": 212.0,
|
| 321 |
+
"completions/min_terminated_length": 212.0,
|
| 322 |
+
"entropy": 0.4704709783196449,
|
| 323 |
"epoch": 0.008,
|
| 324 |
+
"frac_reward_zero_std": 0.85,
|
| 325 |
+
"grad_norm": 0.103515625,
|
| 326 |
+
"kl": 0.4402174398303032,
|
| 327 |
"learning_rate": 9.950629695468755e-06,
|
| 328 |
+
"loss": 0.00011011158348992467,
|
| 329 |
+
"num_tokens": 1350825.0,
|
| 330 |
+
"reward": 0.04487500190734863,
|
| 331 |
+
"reward_std": 0.063816387206316,
|
| 332 |
+
"rewards/env_goofspiel_reward/mean": 0.04487500190734863,
|
| 333 |
+
"rewards/env_goofspiel_reward/std": 0.17575940787792205,
|
| 334 |
+
"sampling/importance_sampling_ratio/max": 1.630899429321289,
|
| 335 |
+
"sampling/importance_sampling_ratio/mean": 0.9812780380249023,
|
| 336 |
+
"sampling/importance_sampling_ratio/min": 0.3998015284538269,
|
| 337 |
+
"sampling/sampling_logp_difference/max": 0.8225874900817871,
|
| 338 |
+
"sampling/sampling_logp_difference/mean": 0.057265565544366834,
|
| 339 |
"step": 50,
|
| 340 |
+
"step_time": 3.4821775794000134
|
| 341 |
},
|
| 342 |
{
|
| 343 |
+
"clip_ratio/high_max": 0.0,
|
| 344 |
+
"clip_ratio/high_mean": 0.0,
|
| 345 |
+
"clip_ratio/low_mean": 0.0,
|
| 346 |
"clip_ratio/low_min": 0.0,
|
| 347 |
+
"clip_ratio/region_mean": 0.0,
|
| 348 |
"completions/clipped_ratio": 0.0,
|
| 349 |
+
"completions/max_length": 374.0,
|
| 350 |
+
"completions/max_terminated_length": 374.0,
|
| 351 |
+
"completions/mean_length": 284.0375,
|
| 352 |
+
"completions/mean_terminated_length": 284.0375,
|
| 353 |
+
"completions/min_length": 206.8,
|
| 354 |
+
"completions/min_terminated_length": 206.8,
|
| 355 |
+
"entropy": 0.23580820970237254,
|
| 356 |
"epoch": 0.0088,
|
| 357 |
+
"frac_reward_zero_std": 0.7875,
|
| 358 |
+
"grad_norm": 0.4765625,
|
| 359 |
+
"kl": 0.6652570061385632,
|
| 360 |
"learning_rate": 9.950621020743173e-06,
|
| 361 |
+
"loss": 8.13461490906775e-05,
|
| 362 |
+
"num_tokens": 1483464.0,
|
| 363 |
+
"reward": 0.05993750244379044,
|
| 364 |
+
"reward_std": 0.08494120314717293,
|
| 365 |
+
"rewards/env_goofspiel_reward/mean": 0.05993750244379044,
|
| 366 |
+
"rewards/env_goofspiel_reward/std": 0.18124795854091644,
|
| 367 |
+
"sampling/importance_sampling_ratio/max": 1.8012218236923219,
|
| 368 |
+
"sampling/importance_sampling_ratio/mean": 0.9974164485931396,
|
| 369 |
+
"sampling/importance_sampling_ratio/min": 0.6288759648799896,
|
| 370 |
+
"sampling/sampling_logp_difference/max": 0.6000619411468506,
|
| 371 |
+
"sampling/sampling_logp_difference/mean": 0.03408294580876827,
|
| 372 |
"step": 55,
|
| 373 |
+
"step_time": 3.518262126200534
|
| 374 |
},
|
| 375 |
{
|
| 376 |
+
"clip_ratio/high_max": 0.0,
|
| 377 |
+
"clip_ratio/high_mean": 0.0,
|
| 378 |
+
"clip_ratio/low_mean": 0.0,
|
| 379 |
"clip_ratio/low_min": 0.0,
|
| 380 |
+
"clip_ratio/region_mean": 0.0,
|
| 381 |
"completions/clipped_ratio": 0.0,
|
| 382 |
+
"completions/max_length": 373.8,
|
| 383 |
+
"completions/max_terminated_length": 373.8,
|
| 384 |
+
"completions/mean_length": 294.53125,
|
| 385 |
+
"completions/mean_terminated_length": 294.53125,
|
| 386 |
+
"completions/min_length": 219.0,
|
| 387 |
+
"completions/min_terminated_length": 219.0,
|
| 388 |
+
"entropy": 0.06949742138385773,
|
| 389 |
"epoch": 0.0096,
|
| 390 |
+
"frac_reward_zero_std": 0.8125,
|
| 391 |
+
"grad_norm": 0.033447265625,
|
| 392 |
+
"kl": 0.8152253001928329,
|
| 393 |
"learning_rate": 9.950609717322956e-06,
|
| 394 |
+
"loss": 0.00011886359425261617,
|
| 395 |
+
"num_tokens": 1618570.0,
|
| 396 |
+
"reward": 0.07500000223517418,
|
| 397 |
+
"reward_std": 0.08485281318426133,
|
| 398 |
+
"rewards/env_goofspiel_reward/mean": 0.07500000223517418,
|
| 399 |
+
"rewards/env_goofspiel_reward/std": 0.1962749719619751,
|
| 400 |
+
"sampling/importance_sampling_ratio/max": 1.434102201461792,
|
| 401 |
+
"sampling/importance_sampling_ratio/mean": 1.0074862837791443,
|
| 402 |
+
"sampling/importance_sampling_ratio/min": 0.8809547781944275,
|
| 403 |
+
"sampling/sampling_logp_difference/max": 0.35270476043224336,
|
| 404 |
+
"sampling/sampling_logp_difference/mean": 0.007789037330076099,
|
| 405 |
"step": 60,
|
| 406 |
+
"step_time": 3.4527530410003235
|
| 407 |
},
|
| 408 |
{
|
| 409 |
+
"clip_ratio/high_max": 0.0,
|
| 410 |
+
"clip_ratio/high_mean": 0.0,
|
| 411 |
+
"clip_ratio/low_mean": 0.0,
|
| 412 |
"clip_ratio/low_min": 0.0,
|
| 413 |
+
"clip_ratio/region_mean": 0.0,
|
| 414 |
"completions/clipped_ratio": 0.0,
|
| 415 |
+
"completions/max_length": 445.2,
|
| 416 |
+
"completions/max_terminated_length": 445.2,
|
| 417 |
+
"completions/mean_length": 325.83125,
|
| 418 |
+
"completions/mean_terminated_length": 325.83125,
|
| 419 |
+
"completions/min_length": 251.2,
|
| 420 |
+
"completions/min_terminated_length": 251.2,
|
| 421 |
+
"entropy": 0.08617601059377193,
|
| 422 |
"epoch": 0.0104,
|
| 423 |
+
"frac_reward_zero_std": 0.875,
|
| 424 |
+
"grad_norm": 0.018798828125,
|
| 425 |
+
"kl": 0.7552784413099289,
|
| 426 |
"learning_rate": 9.950595785216067e-06,
|
| 427 |
+
"loss": 0.00014277141308411957,
|
| 428 |
+
"num_tokens": 1755265.0,
|
| 429 |
+
"reward": 0.03750000149011612,
|
| 430 |
+
"reward_std": 0.0530330091714859,
|
| 431 |
+
"rewards/env_goofspiel_reward/mean": 0.03750000149011612,
|
| 432 |
+
"rewards/env_goofspiel_reward/std": 0.13009902238845825,
|
| 433 |
+
"sampling/importance_sampling_ratio/max": 1.3271136283874512,
|
| 434 |
+
"sampling/importance_sampling_ratio/mean": 1.004535722732544,
|
| 435 |
+
"sampling/importance_sampling_ratio/min": 0.7700567066669464,
|
| 436 |
+
"sampling/sampling_logp_difference/max": 0.3696352869272232,
|
| 437 |
+
"sampling/sampling_logp_difference/mean": 0.008786045084707438,
|
| 438 |
"step": 65,
|
| 439 |
+
"step_time": 3.861200922199896
|
| 440 |
},
|
| 441 |
{
|
| 442 |
+
"clip_ratio/high_max": 0.0,
|
| 443 |
+
"clip_ratio/high_mean": 0.0,
|
| 444 |
+
"clip_ratio/low_mean": 0.0,
|
| 445 |
"clip_ratio/low_min": 0.0,
|
| 446 |
+
"clip_ratio/region_mean": 0.0,
|
| 447 |
"completions/clipped_ratio": 0.0,
|
| 448 |
+
"completions/max_length": 732.0,
|
| 449 |
+
"completions/max_terminated_length": 732.0,
|
| 450 |
+
"completions/mean_length": 544.625,
|
| 451 |
+
"completions/mean_terminated_length": 544.625,
|
| 452 |
+
"completions/min_length": 383.6,
|
| 453 |
+
"completions/min_terminated_length": 383.6,
|
| 454 |
+
"entropy": 0.33896628841757775,
|
| 455 |
"epoch": 0.0112,
|
| 456 |
+
"frac_reward_zero_std": 0.75,
|
| 457 |
+
"grad_norm": 0.2109375,
|
| 458 |
+
"kl": 1.354549203068018,
|
| 459 |
"learning_rate": 9.950579224432321e-06,
|
| 460 |
+
"loss": 0.00035492791794240476,
|
| 461 |
+
"num_tokens": 1929034.0,
|
| 462 |
+
"reward": 0.04277083389461041,
|
| 463 |
+
"reward_std": 0.061901307106018065,
|
| 464 |
+
"rewards/env_goofspiel_reward/mean": 0.04277083389461041,
|
| 465 |
+
"rewards/env_goofspiel_reward/std": 0.14327263236045837,
|
| 466 |
+
"sampling/importance_sampling_ratio/max": 1.457657814025879,
|
| 467 |
+
"sampling/importance_sampling_ratio/mean": 0.9943998098373413,
|
| 468 |
+
"sampling/importance_sampling_ratio/min": 0.6945780634880065,
|
| 469 |
+
"sampling/sampling_logp_difference/max": 0.450562047958374,
|
| 470 |
+
"sampling/sampling_logp_difference/mean": 0.02044728323817253,
|
| 471 |
"step": 70,
|
| 472 |
+
"step_time": 5.568116331400051
|
| 473 |
},
|
| 474 |
{
|
| 475 |
+
"clip_ratio/high_max": 0.0,
|
| 476 |
+
"clip_ratio/high_mean": 0.0,
|
| 477 |
+
"clip_ratio/low_mean": 0.0,
|
| 478 |
"clip_ratio/low_min": 0.0,
|
| 479 |
+
"clip_ratio/region_mean": 0.0,
|
| 480 |
"completions/clipped_ratio": 0.0,
|
| 481 |
+
"completions/max_length": 731.6,
|
| 482 |
+
"completions/max_terminated_length": 731.6,
|
| 483 |
+
"completions/mean_length": 564.09375,
|
| 484 |
+
"completions/mean_terminated_length": 564.09375,
|
| 485 |
+
"completions/min_length": 408.0,
|
| 486 |
+
"completions/min_terminated_length": 408.0,
|
| 487 |
+
"entropy": 0.490242238342762,
|
| 488 |
"epoch": 0.012,
|
| 489 |
+
"frac_reward_zero_std": 0.625,
|
| 490 |
+
"grad_norm": 0.17578125,
|
| 491 |
+
"kl": 0.48342139422893526,
|
| 492 |
"learning_rate": 9.950560034983382e-06,
|
| 493 |
+
"loss": 0.00017674852861091495,
|
| 494 |
+
"num_tokens": 2107362.0,
|
| 495 |
+
"reward": 0.055687499977648255,
|
| 496 |
+
"reward_std": 0.07409889809787273,
|
| 497 |
+
"rewards/env_goofspiel_reward/mean": 0.055687499977648255,
|
| 498 |
+
"rewards/env_goofspiel_reward/std": 0.16855775713920593,
|
| 499 |
+
"sampling/importance_sampling_ratio/max": 1.2386622190475465,
|
| 500 |
+
"sampling/importance_sampling_ratio/mean": 0.9834899187088013,
|
| 501 |
+
"sampling/importance_sampling_ratio/min": 0.776045274734497,
|
| 502 |
+
"sampling/sampling_logp_difference/max": 0.22788455486297607,
|
| 503 |
+
"sampling/sampling_logp_difference/mean": 0.01976088173687458,
|
| 504 |
"step": 75,
|
| 505 |
+
"step_time": 5.500796792399524
|
| 506 |
},
|
| 507 |
{
|
| 508 |
"epoch": 0.012,
|
|
|
|
| 512 |
"eval_clip_ratio/low_min": 0.0,
|
| 513 |
"eval_clip_ratio/region_mean": 0.0,
|
| 514 |
"eval_completions/clipped_ratio": 0.0,
|
| 515 |
+
"eval_completions/max_length": 614.4,
|
| 516 |
+
"eval_completions/max_terminated_length": 614.4,
|
| 517 |
+
"eval_completions/mean_length": 572.65,
|
| 518 |
+
"eval_completions/mean_terminated_length": 572.65,
|
| 519 |
+
"eval_completions/min_length": 531.8,
|
| 520 |
+
"eval_completions/min_terminated_length": 531.8,
|
| 521 |
+
"eval_entropy": 0.5134166538715362,
|
| 522 |
+
"eval_frac_reward_zero_std": 0.8,
|
| 523 |
+
"eval_kl": 0.5596067428588867,
|
| 524 |
+
"eval_loss": -1.7431222659070045e-05,
|
| 525 |
+
"eval_num_tokens": 2107362.0,
|
| 526 |
+
"eval_reward": 0.021166666876524687,
|
| 527 |
+
"eval_reward_std": 0.03134840028360486,
|
| 528 |
+
"eval_rewards/env_goofspiel_reward/mean": 0.021166666876524687,
|
| 529 |
+
"eval_rewards/env_goofspiel_reward/std": 0.04433333072811365,
|
| 530 |
+
"eval_runtime": 2.6749,
|
| 531 |
+
"eval_samples_per_second": 3.738,
|
| 532 |
+
"eval_sampling/importance_sampling_ratio/max": 1.0850860834121705,
|
| 533 |
+
"eval_sampling/importance_sampling_ratio/mean": 1.0015697360038758,
|
| 534 |
+
"eval_sampling/importance_sampling_ratio/min": 0.9188181281089782,
|
| 535 |
+
"eval_sampling/sampling_logp_difference/max": 0.10906529426574707,
|
| 536 |
+
"eval_sampling/sampling_logp_difference/mean": 0.017179742455482483,
|
| 537 |
+
"eval_steps_per_second": 1.122,
|
| 538 |
"step": 75
|
| 539 |
+
},
|
| 540 |
+
{
|
| 541 |
+
"clip_ratio/high_max": 0.0,
|
| 542 |
+
"clip_ratio/high_mean": 0.0,
|
| 543 |
+
"clip_ratio/low_mean": 0.0,
|
| 544 |
+
"clip_ratio/low_min": 0.0,
|
| 545 |
+
"clip_ratio/region_mean": 0.0,
|
| 546 |
+
"completions/clipped_ratio": 0.0,
|
| 547 |
+
"completions/max_length": 733.2,
|
| 548 |
+
"completions/max_terminated_length": 733.2,
|
| 549 |
+
"completions/mean_length": 549.075,
|
| 550 |
+
"completions/mean_terminated_length": 549.075,
|
| 551 |
+
"completions/min_length": 395.8,
|
| 552 |
+
"completions/min_terminated_length": 395.8,
|
| 553 |
+
"entropy": 0.527910877764225,
|
| 554 |
+
"epoch": 0.0128,
|
| 555 |
+
"frac_reward_zero_std": 0.7,
|
| 556 |
+
"grad_norm": 0.255859375,
|
| 557 |
+
"kl": 0.47208923250436785,
|
| 558 |
+
"learning_rate": 9.95053821688277e-06,
|
| 559 |
+
"loss": 0.0007011178880929947,
|
| 560 |
+
"num_tokens": 2282715.0,
|
| 561 |
+
"reward": 0.05902083367109299,
|
| 562 |
+
"reward_std": 0.07722195237874985,
|
| 563 |
+
"rewards/env_goofspiel_reward/mean": 0.05902083367109299,
|
| 564 |
+
"rewards/env_goofspiel_reward/std": 0.15880293548107147,
|
| 565 |
+
"sampling/importance_sampling_ratio/max": 1.3604092836380004,
|
| 566 |
+
"sampling/importance_sampling_ratio/mean": 1.0046632289886475,
|
| 567 |
+
"sampling/importance_sampling_ratio/min": 0.7511004090309144,
|
| 568 |
+
"sampling/sampling_logp_difference/max": 0.3062611103057861,
|
| 569 |
+
"sampling/sampling_logp_difference/mean": 0.022707394510507583,
|
| 570 |
+
"step": 80,
|
| 571 |
+
"step_time": 5.533078667399787
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"clip_ratio/high_max": 0.0,
|
| 575 |
+
"clip_ratio/high_mean": 0.0,
|
| 576 |
+
"clip_ratio/low_mean": 0.0,
|
| 577 |
+
"clip_ratio/low_min": 0.0,
|
| 578 |
+
"clip_ratio/region_mean": 0.0,
|
| 579 |
+
"completions/clipped_ratio": 0.0,
|
| 580 |
+
"completions/max_length": 732.2,
|
| 581 |
+
"completions/max_terminated_length": 732.2,
|
| 582 |
+
"completions/mean_length": 561.8875,
|
| 583 |
+
"completions/mean_terminated_length": 561.8875,
|
| 584 |
+
"completions/min_length": 408.0,
|
| 585 |
+
"completions/min_terminated_length": 408.0,
|
| 586 |
+
"entropy": 0.46153847873210907,
|
| 587 |
+
"epoch": 0.0136,
|
| 588 |
+
"frac_reward_zero_std": 0.825,
|
| 589 |
+
"grad_norm": 0.095703125,
|
| 590 |
+
"kl": 0.5197702750563622,
|
| 591 |
+
"learning_rate": 9.950513770145857e-06,
|
| 592 |
+
"loss": 0.00010828666854649782,
|
| 593 |
+
"num_tokens": 2459616.0,
|
| 594 |
+
"reward": 0.04585416615009308,
|
| 595 |
+
"reward_std": 0.057717590034008025,
|
| 596 |
+
"rewards/env_goofspiel_reward/mean": 0.04585416615009308,
|
| 597 |
+
"rewards/env_goofspiel_reward/std": 0.1589788019657135,
|
| 598 |
+
"sampling/importance_sampling_ratio/max": 1.5443368434906006,
|
| 599 |
+
"sampling/importance_sampling_ratio/mean": 1.0077369570732118,
|
| 600 |
+
"sampling/importance_sampling_ratio/min": 0.7487894654273987,
|
| 601 |
+
"sampling/sampling_logp_difference/max": 0.36205780506134033,
|
| 602 |
+
"sampling/sampling_logp_difference/mean": 0.021098615229129793,
|
| 603 |
+
"step": 85,
|
| 604 |
+
"step_time": 5.460692007999751
|
| 605 |
+
},
|
| 606 |
+
{
|
| 607 |
+
"clip_ratio/high_max": 0.0,
|
| 608 |
+
"clip_ratio/high_mean": 0.0,
|
| 609 |
+
"clip_ratio/low_mean": 0.0,
|
| 610 |
+
"clip_ratio/low_min": 0.0,
|
| 611 |
+
"clip_ratio/region_mean": 0.0,
|
| 612 |
+
"completions/clipped_ratio": 0.0,
|
| 613 |
+
"completions/max_length": 732.0,
|
| 614 |
+
"completions/max_terminated_length": 732.0,
|
| 615 |
+
"completions/mean_length": 581.36875,
|
| 616 |
+
"completions/mean_terminated_length": 581.36875,
|
| 617 |
+
"completions/min_length": 408.0,
|
| 618 |
+
"completions/min_terminated_length": 408.0,
|
| 619 |
+
"entropy": 0.48351355642080307,
|
| 620 |
+
"epoch": 0.0144,
|
| 621 |
+
"frac_reward_zero_std": 0.8375,
|
| 622 |
+
"grad_norm": 0.1630859375,
|
| 623 |
+
"kl": 0.6207657802850008,
|
| 624 |
+
"learning_rate": 9.950486694789862e-06,
|
| 625 |
+
"loss": 0.00018907544435933232,
|
| 626 |
+
"num_tokens": 2641146.0,
|
| 627 |
+
"reward": 0.026770833879709244,
|
| 628 |
+
"reward_std": 0.03856678232550621,
|
| 629 |
+
"rewards/env_goofspiel_reward/mean": 0.026770833879709244,
|
| 630 |
+
"rewards/env_goofspiel_reward/std": 0.10960558950901031,
|
| 631 |
+
"sampling/importance_sampling_ratio/max": 1.2647137641906738,
|
| 632 |
+
"sampling/importance_sampling_ratio/mean": 1.004696297645569,
|
| 633 |
+
"sampling/importance_sampling_ratio/min": 0.7099822998046875,
|
| 634 |
+
"sampling/sampling_logp_difference/max": 0.2817555904388428,
|
| 635 |
+
"sampling/sampling_logp_difference/mean": 0.022446612268686293,
|
| 636 |
+
"step": 90,
|
| 637 |
+
"step_time": 5.498646953999923
|
| 638 |
+
},
|
| 639 |
+
{
|
| 640 |
+
"clip_ratio/high_max": 0.0,
|
| 641 |
+
"clip_ratio/high_mean": 0.0,
|
| 642 |
+
"clip_ratio/low_mean": 0.0,
|
| 643 |
+
"clip_ratio/low_min": 0.0,
|
| 644 |
+
"clip_ratio/region_mean": 0.0,
|
| 645 |
+
"completions/clipped_ratio": 0.0,
|
| 646 |
+
"completions/max_length": 732.0,
|
| 647 |
+
"completions/max_terminated_length": 732.0,
|
| 648 |
+
"completions/mean_length": 549.825,
|
| 649 |
+
"completions/mean_terminated_length": 549.825,
|
| 650 |
+
"completions/min_length": 408.0,
|
| 651 |
+
"completions/min_terminated_length": 408.0,
|
| 652 |
+
"entropy": 0.5179941833019257,
|
| 653 |
+
"epoch": 0.0152,
|
| 654 |
+
"frac_reward_zero_std": 0.8125,
|
| 655 |
+
"grad_norm": 0.0908203125,
|
| 656 |
+
"kl": 0.46866228580474856,
|
| 657 |
+
"learning_rate": 9.95045699083386e-06,
|
| 658 |
+
"loss": 0.00028545891400426625,
|
| 659 |
+
"num_tokens": 2815323.0,
|
| 660 |
+
"reward": 0.03772916682064533,
|
| 661 |
+
"reward_std": 0.05388742834329605,
|
| 662 |
+
"rewards/env_goofspiel_reward/mean": 0.03772916682064533,
|
| 663 |
+
"rewards/env_goofspiel_reward/std": 0.13617317676544188,
|
| 664 |
+
"sampling/importance_sampling_ratio/max": 1.2603100776672362,
|
| 665 |
+
"sampling/importance_sampling_ratio/mean": 1.0017661929130555,
|
| 666 |
+
"sampling/importance_sampling_ratio/min": 0.7478124976158143,
|
| 667 |
+
"sampling/sampling_logp_difference/max": 0.25655815601348875,
|
| 668 |
+
"sampling/sampling_logp_difference/mean": 0.023887046799063682,
|
| 669 |
+
"step": 95,
|
| 670 |
+
"step_time": 5.659158172999559
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"clip_ratio/high_max": 0.0,
|
| 674 |
+
"clip_ratio/high_mean": 0.0,
|
| 675 |
+
"clip_ratio/low_mean": 0.0,
|
| 676 |
+
"clip_ratio/low_min": 0.0,
|
| 677 |
+
"clip_ratio/region_mean": 0.0,
|
| 678 |
+
"completions/clipped_ratio": 0.0,
|
| 679 |
+
"completions/max_length": 1003.6,
|
| 680 |
+
"completions/max_terminated_length": 1003.6,
|
| 681 |
+
"completions/mean_length": 754.59375,
|
| 682 |
+
"completions/mean_terminated_length": 754.59375,
|
| 683 |
+
"completions/min_length": 550.8,
|
| 684 |
+
"completions/min_terminated_length": 550.8,
|
| 685 |
+
"entropy": 0.5820856660604476,
|
| 686 |
+
"epoch": 0.016,
|
| 687 |
+
"frac_reward_zero_std": 0.8125,
|
| 688 |
+
"grad_norm": 0.171875,
|
| 689 |
+
"kl": 0.4048814922571182,
|
| 690 |
+
"learning_rate": 9.950424658298776e-06,
|
| 691 |
+
"loss": 0.0002957838121801615,
|
| 692 |
+
"num_tokens": 3022802.0,
|
| 693 |
+
"reward": 0.03360416684299707,
|
| 694 |
+
"reward_std": 0.0420433908700943,
|
| 695 |
+
"rewards/env_goofspiel_reward/mean": 0.03360416684299707,
|
| 696 |
+
"rewards/env_goofspiel_reward/std": 0.12295188903808593,
|
| 697 |
+
"sampling/importance_sampling_ratio/max": 1.3984624147415161,
|
| 698 |
+
"sampling/importance_sampling_ratio/mean": 0.980130672454834,
|
| 699 |
+
"sampling/importance_sampling_ratio/min": 0.6590921759605408,
|
| 700 |
+
"sampling/sampling_logp_difference/max": 0.3471600294113159,
|
| 701 |
+
"sampling/sampling_logp_difference/mean": 0.02802230753004551,
|
| 702 |
+
"step": 100,
|
| 703 |
+
"step_time": 7.31073631659965
|
| 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 |
+
"completions/clipped_ratio": 0.0,
|
| 712 |
+
"completions/max_length": 1075.0,
|
| 713 |
+
"completions/max_terminated_length": 1075.0,
|
| 714 |
+
"completions/mean_length": 814.73125,
|
| 715 |
+
"completions/mean_terminated_length": 814.73125,
|
| 716 |
+
"completions/min_length": 590.0,
|
| 717 |
+
"completions/min_terminated_length": 590.0,
|
| 718 |
+
"entropy": 0.6229563593864441,
|
| 719 |
+
"epoch": 0.0168,
|
| 720 |
+
"frac_reward_zero_std": 0.775,
|
| 721 |
+
"grad_norm": 0.09326171875,
|
| 722 |
+
"kl": 0.4244683228433132,
|
| 723 |
+
"learning_rate": 9.950389697207388e-06,
|
| 724 |
+
"loss": 0.00020875066984444858,
|
| 725 |
+
"num_tokens": 3239290.0,
|
| 726 |
+
"reward": 0.038875000365078446,
|
| 727 |
+
"reward_std": 0.049674246832728385,
|
| 728 |
+
"rewards/env_goofspiel_reward/mean": 0.038875000365078446,
|
| 729 |
+
"rewards/env_goofspiel_reward/std": 0.11504672318696976,
|
| 730 |
+
"sampling/importance_sampling_ratio/max": 1.3492220401763917,
|
| 731 |
+
"sampling/importance_sampling_ratio/mean": 0.9891064882278442,
|
| 732 |
+
"sampling/importance_sampling_ratio/min": 0.5561659098671476,
|
| 733 |
+
"sampling/sampling_logp_difference/max": 4.213754487037659,
|
| 734 |
+
"sampling/sampling_logp_difference/mean": 0.044166411831974985,
|
| 735 |
+
"step": 105,
|
| 736 |
+
"step_time": 7.745885893800368
|
| 737 |
+
},
|
| 738 |
+
{
|
| 739 |
+
"clip_ratio/high_max": 0.0,
|
| 740 |
+
"clip_ratio/high_mean": 0.0,
|
| 741 |
+
"clip_ratio/low_mean": 0.0,
|
| 742 |
+
"clip_ratio/low_min": 0.0,
|
| 743 |
+
"clip_ratio/region_mean": 0.0,
|
| 744 |
+
"completions/clipped_ratio": 0.0,
|
| 745 |
+
"completions/max_length": 1075.6,
|
| 746 |
+
"completions/max_terminated_length": 1075.6,
|
| 747 |
+
"completions/mean_length": 814.6375,
|
| 748 |
+
"completions/mean_terminated_length": 814.6375,
|
| 749 |
+
"completions/min_length": 590.0,
|
| 750 |
+
"completions/min_terminated_length": 590.0,
|
| 751 |
+
"entropy": 0.6465200364589692,
|
| 752 |
+
"epoch": 0.0176,
|
| 753 |
+
"frac_reward_zero_std": 0.7875,
|
| 754 |
+
"grad_norm": 0.007080078125,
|
| 755 |
+
"kl": 0.3995703622698784,
|
| 756 |
+
"learning_rate": 9.950352107584324e-06,
|
| 757 |
+
"loss": 0.0007265920285135508,
|
| 758 |
+
"num_tokens": 3456558.0,
|
| 759 |
+
"reward": 0.023437499813735486,
|
| 760 |
+
"reward_std": 0.03473661988973618,
|
| 761 |
+
"rewards/env_goofspiel_reward/mean": 0.023437499813735486,
|
| 762 |
+
"rewards/env_goofspiel_reward/std": 0.0773700624704361,
|
| 763 |
+
"sampling/importance_sampling_ratio/max": 1.4152228832244873,
|
| 764 |
+
"sampling/importance_sampling_ratio/mean": 0.9911394953727722,
|
| 765 |
+
"sampling/importance_sampling_ratio/min": 0.4911388456960605,
|
| 766 |
+
"sampling/sampling_logp_difference/max": 4.677203369140625,
|
| 767 |
+
"sampling/sampling_logp_difference/mean": 0.04649800434708595,
|
| 768 |
+
"step": 110,
|
| 769 |
+
"step_time": 7.710595274399202
|
| 770 |
+
},
|
| 771 |
+
{
|
| 772 |
+
"clip_ratio/high_max": 0.0,
|
| 773 |
+
"clip_ratio/high_mean": 0.0,
|
| 774 |
+
"clip_ratio/low_mean": 0.0,
|
| 775 |
+
"clip_ratio/low_min": 0.0,
|
| 776 |
+
"clip_ratio/region_mean": 0.0,
|
| 777 |
+
"completions/clipped_ratio": 0.0,
|
| 778 |
+
"completions/max_length": 1075.0,
|
| 779 |
+
"completions/max_terminated_length": 1075.0,
|
| 780 |
+
"completions/mean_length": 831.425,
|
| 781 |
+
"completions/mean_terminated_length": 831.425,
|
| 782 |
+
"completions/min_length": 590.0,
|
| 783 |
+
"completions/min_terminated_length": 590.0,
|
| 784 |
+
"entropy": 0.6319158881902694,
|
| 785 |
+
"epoch": 0.0184,
|
| 786 |
+
"frac_reward_zero_std": 0.7625,
|
| 787 |
+
"grad_norm": 0.30078125,
|
| 788 |
+
"kl": 0.4417851775884628,
|
| 789 |
+
"learning_rate": 9.950311889456064e-06,
|
| 790 |
+
"loss": -0.00014657191932201385,
|
| 791 |
+
"num_tokens": 3677316.0,
|
| 792 |
+
"reward": 0.03468749914318323,
|
| 793 |
+
"reward_std": 0.0499394167214632,
|
| 794 |
+
"rewards/env_goofspiel_reward/mean": 0.03468749914318323,
|
| 795 |
+
"rewards/env_goofspiel_reward/std": 0.10438641458749771,
|
| 796 |
+
"sampling/importance_sampling_ratio/max": 1.4034383773803711,
|
| 797 |
+
"sampling/importance_sampling_ratio/mean": 1.003822433948517,
|
| 798 |
+
"sampling/importance_sampling_ratio/min": 0.7400717735290527,
|
| 799 |
+
"sampling/sampling_logp_difference/max": 0.2535316705703735,
|
| 800 |
+
"sampling/sampling_logp_difference/mean": 0.02623956575989723,
|
| 801 |
+
"step": 115,
|
| 802 |
+
"step_time": 7.7139906279999195
|
| 803 |
+
},
|
| 804 |
+
{
|
| 805 |
+
"clip_ratio/high_max": 0.0,
|
| 806 |
+
"clip_ratio/high_mean": 0.0,
|
| 807 |
+
"clip_ratio/low_mean": 0.0,
|
| 808 |
+
"clip_ratio/low_min": 0.0,
|
| 809 |
+
"clip_ratio/region_mean": 0.0,
|
| 810 |
+
"completions/clipped_ratio": 0.0,
|
| 811 |
+
"completions/max_length": 1075.2,
|
| 812 |
+
"completions/max_terminated_length": 1075.2,
|
| 813 |
+
"completions/mean_length": 809.6125,
|
| 814 |
+
"completions/mean_terminated_length": 809.6125,
|
| 815 |
+
"completions/min_length": 578.0,
|
| 816 |
+
"completions/min_terminated_length": 578.0,
|
| 817 |
+
"entropy": 0.6280775606632233,
|
| 818 |
+
"epoch": 0.0192,
|
| 819 |
+
"frac_reward_zero_std": 0.825,
|
| 820 |
+
"grad_norm": 0.1328125,
|
| 821 |
+
"kl": 0.39189435467123984,
|
| 822 |
+
"learning_rate": 9.950269042850943e-06,
|
| 823 |
+
"loss": 0.00020067014265805482,
|
| 824 |
+
"num_tokens": 3893277.0,
|
| 825 |
+
"reward": 0.028249999321997166,
|
| 826 |
+
"reward_std": 0.0342946782708168,
|
| 827 |
+
"rewards/env_goofspiel_reward/mean": 0.028249999321997166,
|
| 828 |
+
"rewards/env_goofspiel_reward/std": 0.10060491263866425,
|
| 829 |
+
"sampling/importance_sampling_ratio/max": 1.449436855316162,
|
| 830 |
+
"sampling/importance_sampling_ratio/mean": 0.9966808676719665,
|
| 831 |
+
"sampling/importance_sampling_ratio/min": 0.6293053507804871,
|
| 832 |
+
"sampling/sampling_logp_difference/max": 0.3306601524353027,
|
| 833 |
+
"sampling/sampling_logp_difference/mean": 0.027687131240963937,
|
| 834 |
+
"step": 120,
|
| 835 |
+
"step_time": 7.60252943919968
|
| 836 |
+
},
|
| 837 |
+
{
|
| 838 |
+
"clip_ratio/high_max": 0.0,
|
| 839 |
+
"clip_ratio/high_mean": 0.0,
|
| 840 |
+
"clip_ratio/low_mean": 0.0,
|
| 841 |
+
"clip_ratio/low_min": 0.0,
|
| 842 |
+
"clip_ratio/region_mean": 0.0,
|
| 843 |
+
"completions/clipped_ratio": 0.0,
|
| 844 |
+
"completions/max_length": 1075.0,
|
| 845 |
+
"completions/max_terminated_length": 1075.0,
|
| 846 |
+
"completions/mean_length": 831.9125,
|
| 847 |
+
"completions/mean_terminated_length": 831.9125,
|
| 848 |
+
"completions/min_length": 590.0,
|
| 849 |
+
"completions/min_terminated_length": 590.0,
|
| 850 |
+
"entropy": 0.6203011125326157,
|
| 851 |
+
"epoch": 0.02,
|
| 852 |
+
"frac_reward_zero_std": 0.8,
|
| 853 |
+
"grad_norm": 0.0888671875,
|
| 854 |
+
"kl": 0.48651044964790346,
|
| 855 |
+
"learning_rate": 9.95022356779914e-06,
|
| 856 |
+
"loss": -4.498030175454914e-05,
|
| 857 |
+
"num_tokens": 4114505.0,
|
| 858 |
+
"reward": 0.0279999990016222,
|
| 859 |
+
"reward_std": 0.04065863937139511,
|
| 860 |
+
"rewards/env_goofspiel_reward/mean": 0.0279999990016222,
|
| 861 |
+
"rewards/env_goofspiel_reward/std": 0.09859532788395882,
|
| 862 |
+
"sampling/importance_sampling_ratio/max": 1.3767565965652466,
|
| 863 |
+
"sampling/importance_sampling_ratio/mean": 1.0073092341423036,
|
| 864 |
+
"sampling/importance_sampling_ratio/min": 0.6863486647605896,
|
| 865 |
+
"sampling/sampling_logp_difference/max": 0.2979677677154541,
|
| 866 |
+
"sampling/sampling_logp_difference/mean": 0.026996534690260886,
|
| 867 |
+
"step": 125,
|
| 868 |
+
"step_time": 7.651589208400037
|
| 869 |
+
},
|
| 870 |
+
{
|
| 871 |
+
"clip_ratio/high_max": 0.0,
|
| 872 |
+
"clip_ratio/high_mean": 0.0,
|
| 873 |
+
"clip_ratio/low_mean": 0.0,
|
| 874 |
+
"clip_ratio/low_min": 0.0,
|
| 875 |
+
"clip_ratio/region_mean": 0.0,
|
| 876 |
+
"completions/clipped_ratio": 0.0,
|
| 877 |
+
"completions/max_length": 1206.4,
|
| 878 |
+
"completions/max_terminated_length": 1206.4,
|
| 879 |
+
"completions/mean_length": 933.89375,
|
| 880 |
+
"completions/mean_terminated_length": 933.89375,
|
| 881 |
+
"completions/min_length": 657.2,
|
| 882 |
+
"completions/min_terminated_length": 657.2,
|
| 883 |
+
"entropy": 0.6028641879558563,
|
| 884 |
+
"epoch": 0.0208,
|
| 885 |
+
"frac_reward_zero_std": 0.85,
|
| 886 |
+
"grad_norm": 0.12890625,
|
| 887 |
+
"kl": 0.3837214097380638,
|
| 888 |
+
"learning_rate": 9.950175464332696e-06,
|
| 889 |
+
"loss": 3.275293856859207e-05,
|
| 890 |
+
"num_tokens": 4351587.0,
|
| 891 |
+
"reward": 0.02824999988079071,
|
| 892 |
+
"reward_std": 0.040128308534622195,
|
| 893 |
+
"rewards/env_goofspiel_reward/mean": 0.02824999988079071,
|
| 894 |
+
"rewards/env_goofspiel_reward/std": 0.107171730697155,
|
| 895 |
+
"sampling/importance_sampling_ratio/max": 1.4337808609008789,
|
| 896 |
+
"sampling/importance_sampling_ratio/mean": 1.0062252044677735,
|
| 897 |
+
"sampling/importance_sampling_ratio/min": 0.7049606084823609,
|
| 898 |
+
"sampling/sampling_logp_difference/max": 0.29388277530670165,
|
| 899 |
+
"sampling/sampling_logp_difference/mean": 0.027071699127554895,
|
| 900 |
+
"step": 130,
|
| 901 |
+
"step_time": 8.72059853139981
|
| 902 |
+
},
|
| 903 |
+
{
|
| 904 |
+
"clip_ratio/high_max": 0.0,
|
| 905 |
+
"clip_ratio/high_mean": 0.0,
|
| 906 |
+
"clip_ratio/low_mean": 0.0,
|
| 907 |
+
"clip_ratio/low_min": 0.0,
|
| 908 |
+
"clip_ratio/region_mean": 0.0,
|
| 909 |
+
"completions/clipped_ratio": 0.0,
|
| 910 |
+
"completions/max_length": 1403.2,
|
| 911 |
+
"completions/max_terminated_length": 1403.2,
|
| 912 |
+
"completions/mean_length": 1043.53125,
|
| 913 |
+
"completions/mean_terminated_length": 1043.53125,
|
| 914 |
+
"completions/min_length": 758.0,
|
| 915 |
+
"completions/min_terminated_length": 758.0,
|
| 916 |
+
"entropy": 0.5716616868972778,
|
| 917 |
+
"epoch": 0.0216,
|
| 918 |
+
"frac_reward_zero_std": 0.825,
|
| 919 |
+
"grad_norm": 0.11865234375,
|
| 920 |
+
"kl": 0.3439621731638908,
|
| 921 |
+
"learning_rate": 9.950124732485496e-06,
|
| 922 |
+
"loss": -0.00015461102593690158,
|
| 923 |
+
"num_tokens": 4604604.0,
|
| 924 |
+
"reward": 0.02993750013411045,
|
| 925 |
+
"reward_std": 0.04251479506492615,
|
| 926 |
+
"rewards/env_goofspiel_reward/mean": 0.02993750013411045,
|
| 927 |
+
"rewards/env_goofspiel_reward/std": 0.10386608839035034,
|
| 928 |
+
"sampling/importance_sampling_ratio/max": 1.560789942741394,
|
| 929 |
+
"sampling/importance_sampling_ratio/mean": 1.0229114413261413,
|
| 930 |
+
"sampling/importance_sampling_ratio/min": 0.656338346004486,
|
| 931 |
+
"sampling/sampling_logp_difference/max": 0.33734931945800783,
|
| 932 |
+
"sampling/sampling_logp_difference/mean": 0.026270415633916855,
|
| 933 |
+
"step": 135,
|
| 934 |
+
"step_time": 10.42263705259993
|
| 935 |
+
},
|
| 936 |
+
{
|
| 937 |
+
"clip_ratio/high_max": 0.0,
|
| 938 |
+
"clip_ratio/high_mean": 0.0,
|
| 939 |
+
"clip_ratio/low_mean": 0.0,
|
| 940 |
+
"clip_ratio/low_min": 0.0,
|
| 941 |
+
"clip_ratio/region_mean": 0.0,
|
| 942 |
+
"completions/clipped_ratio": 0.0,
|
| 943 |
+
"completions/max_length": 1404.2,
|
| 944 |
+
"completions/max_terminated_length": 1404.2,
|
| 945 |
+
"completions/mean_length": 1106.9125,
|
| 946 |
+
"completions/mean_terminated_length": 1106.9125,
|
| 947 |
+
"completions/min_length": 746.4,
|
| 948 |
+
"completions/min_terminated_length": 746.4,
|
| 949 |
+
"entropy": 0.6015262037515641,
|
| 950 |
+
"epoch": 0.0224,
|
| 951 |
+
"frac_reward_zero_std": 0.7875,
|
| 952 |
+
"grad_norm": 0.1142578125,
|
| 953 |
+
"kl": 0.4554178059101105,
|
| 954 |
+
"learning_rate": 9.95007137229328e-06,
|
| 955 |
+
"loss": 4.9450411461293696e-05,
|
| 956 |
+
"num_tokens": 4870794.0,
|
| 957 |
+
"reward": 0.02393750089686364,
|
| 958 |
+
"reward_std": 0.03491339806932956,
|
| 959 |
+
"rewards/env_goofspiel_reward/mean": 0.02393750089686364,
|
| 960 |
+
"rewards/env_goofspiel_reward/std": 0.07218290558084846,
|
| 961 |
+
"sampling/importance_sampling_ratio/max": 1.4548166275024415,
|
| 962 |
+
"sampling/importance_sampling_ratio/mean": 0.9814130544662476,
|
| 963 |
+
"sampling/importance_sampling_ratio/min": 0.6002501428127289,
|
| 964 |
+
"sampling/sampling_logp_difference/max": 0.33309857845306395,
|
| 965 |
+
"sampling/sampling_logp_difference/mean": 0.028818363696336745,
|
| 966 |
+
"step": 140,
|
| 967 |
+
"step_time": 10.392815717999474
|
| 968 |
+
},
|
| 969 |
+
{
|
| 970 |
+
"clip_ratio/high_max": 0.0,
|
| 971 |
+
"clip_ratio/high_mean": 0.0,
|
| 972 |
+
"clip_ratio/low_mean": 0.0,
|
| 973 |
+
"clip_ratio/low_min": 0.0,
|
| 974 |
+
"clip_ratio/region_mean": 0.0,
|
| 975 |
+
"completions/clipped_ratio": 0.0,
|
| 976 |
+
"completions/max_length": 1403.4,
|
| 977 |
+
"completions/max_terminated_length": 1403.4,
|
| 978 |
+
"completions/mean_length": 1064.275,
|
| 979 |
+
"completions/mean_terminated_length": 1064.275,
|
| 980 |
+
"completions/min_length": 746.6,
|
| 981 |
+
"completions/min_terminated_length": 746.6,
|
| 982 |
+
"entropy": 0.5568872570991517,
|
| 983 |
+
"epoch": 0.0232,
|
| 984 |
+
"frac_reward_zero_std": 0.775,
|
| 985 |
+
"grad_norm": 0.2041015625,
|
| 986 |
+
"kl": 0.3555284239351749,
|
| 987 |
+
"learning_rate": 9.950015383793636e-06,
|
| 988 |
+
"loss": -0.00010267798788845539,
|
| 989 |
+
"num_tokens": 5127626.0,
|
| 990 |
+
"reward": 0.03343750163912773,
|
| 991 |
+
"reward_std": 0.04799487330019474,
|
| 992 |
+
"rewards/env_goofspiel_reward/mean": 0.03343750163912773,
|
| 993 |
+
"rewards/env_goofspiel_reward/std": 0.10812449753284455,
|
| 994 |
+
"sampling/importance_sampling_ratio/max": 1.587186908721924,
|
| 995 |
+
"sampling/importance_sampling_ratio/mean": 1.0198248624801636,
|
| 996 |
+
"sampling/importance_sampling_ratio/min": 0.7297749638557434,
|
| 997 |
+
"sampling/sampling_logp_difference/max": 0.2695728540420532,
|
| 998 |
+
"sampling/sampling_logp_difference/mean": 0.0254237774759531,
|
| 999 |
+
"step": 145,
|
| 1000 |
+
"step_time": 10.32607021879976
|
| 1001 |
+
},
|
| 1002 |
+
{
|
| 1003 |
+
"clip_ratio/high_max": 0.0,
|
| 1004 |
+
"clip_ratio/high_mean": 0.0,
|
| 1005 |
+
"clip_ratio/low_mean": 0.0,
|
| 1006 |
+
"clip_ratio/low_min": 0.0,
|
| 1007 |
+
"clip_ratio/region_mean": 0.0,
|
| 1008 |
+
"completions/clipped_ratio": 0.0,
|
| 1009 |
+
"completions/max_length": 1400.4,
|
| 1010 |
+
"completions/max_terminated_length": 1400.4,
|
| 1011 |
+
"completions/mean_length": 1040.89375,
|
| 1012 |
+
"completions/mean_terminated_length": 1040.89375,
|
| 1013 |
+
"completions/min_length": 758.0,
|
| 1014 |
+
"completions/min_terminated_length": 758.0,
|
| 1015 |
+
"entropy": 0.5438663050532341,
|
| 1016 |
+
"epoch": 0.024,
|
| 1017 |
+
"frac_reward_zero_std": 0.8,
|
| 1018 |
+
"grad_norm": 0.11181640625,
|
| 1019 |
+
"kl": 0.313924690335989,
|
| 1020 |
+
"learning_rate": 9.949956767026006e-06,
|
| 1021 |
+
"loss": 0.0002234043786302209,
|
| 1022 |
+
"num_tokens": 5379923.0,
|
| 1023 |
+
"reward": 0.03181250132620335,
|
| 1024 |
+
"reward_std": 0.0451664462685585,
|
| 1025 |
+
"rewards/env_goofspiel_reward/mean": 0.03181250132620335,
|
| 1026 |
+
"rewards/env_goofspiel_reward/std": 0.10608797073364258,
|
| 1027 |
+
"sampling/importance_sampling_ratio/max": 1.3631083726882935,
|
| 1028 |
+
"sampling/importance_sampling_ratio/mean": 1.0006531119346618,
|
| 1029 |
+
"sampling/importance_sampling_ratio/min": 0.691138219833374,
|
| 1030 |
+
"sampling/sampling_logp_difference/max": 0.26457092761993406,
|
| 1031 |
+
"sampling/sampling_logp_difference/mean": 0.024369171261787413,
|
| 1032 |
+
"step": 150,
|
| 1033 |
+
"step_time": 10.337828036600513
|
| 1034 |
+
},
|
| 1035 |
+
{
|
| 1036 |
+
"epoch": 0.024,
|
| 1037 |
+
"eval_clip_ratio/high_max": 0.0,
|
| 1038 |
+
"eval_clip_ratio/high_mean": 0.0,
|
| 1039 |
+
"eval_clip_ratio/low_mean": 0.0,
|
| 1040 |
+
"eval_clip_ratio/low_min": 0.0,
|
| 1041 |
+
"eval_clip_ratio/region_mean": 0.0,
|
| 1042 |
+
"eval_completions/clipped_ratio": 0.0,
|
| 1043 |
+
"eval_completions/max_length": 1167.6,
|
| 1044 |
+
"eval_completions/max_terminated_length": 1167.6,
|
| 1045 |
+
"eval_completions/mean_length": 1085.75,
|
| 1046 |
+
"eval_completions/mean_terminated_length": 1085.75,
|
| 1047 |
+
"eval_completions/min_length": 1003.6,
|
| 1048 |
+
"eval_completions/min_terminated_length": 1003.6,
|
| 1049 |
+
"eval_entropy": 0.5619663238525391,
|
| 1050 |
+
"eval_frac_reward_zero_std": 0.9,
|
| 1051 |
+
"eval_kl": 0.4370845973491669,
|
| 1052 |
+
"eval_loss": -0.00014827963605057448,
|
| 1053 |
+
"eval_num_tokens": 5379923.0,
|
| 1054 |
+
"eval_reward": -0.0004999999888241291,
|
| 1055 |
+
"eval_reward_std": 0.0007071067579090595,
|
| 1056 |
+
"eval_rewards/env_goofspiel_reward/mean": -0.0004999999888241291,
|
| 1057 |
+
"eval_rewards/env_goofspiel_reward/std": 0.0009999999776482583,
|
| 1058 |
+
"eval_runtime": 4.3907,
|
| 1059 |
+
"eval_samples_per_second": 2.278,
|
| 1060 |
+
"eval_sampling/importance_sampling_ratio/max": 1.2002264738082886,
|
| 1061 |
+
"eval_sampling/importance_sampling_ratio/mean": 1.0190080761909486,
|
| 1062 |
+
"eval_sampling/importance_sampling_ratio/min": 0.8775287628173828,
|
| 1063 |
+
"eval_sampling/sampling_logp_difference/max": 0.19612762928009034,
|
| 1064 |
+
"eval_sampling/sampling_logp_difference/mean": 0.025080177932977676,
|
| 1065 |
+
"eval_steps_per_second": 0.683,
|
| 1066 |
+
"step": 150
|
| 1067 |
+
},
|
| 1068 |
+
{
|
| 1069 |
+
"epoch": 0.02432,
|
| 1070 |
+
"eval_clip_ratio/high_max": 0.0,
|
| 1071 |
+
"eval_clip_ratio/high_mean": 0.0,
|
| 1072 |
+
"eval_clip_ratio/low_mean": 0.0,
|
| 1073 |
+
"eval_clip_ratio/low_min": 0.0,
|
| 1074 |
+
"eval_clip_ratio/region_mean": 0.0,
|
| 1075 |
+
"eval_completions/clipped_ratio": 0.0,
|
| 1076 |
+
"eval_completions/max_length": 1166.0,
|
| 1077 |
+
"eval_completions/max_terminated_length": 1166.0,
|
| 1078 |
+
"eval_completions/mean_length": 1082.75,
|
| 1079 |
+
"eval_completions/mean_terminated_length": 1082.75,
|
| 1080 |
+
"eval_completions/min_length": 999.8,
|
| 1081 |
+
"eval_completions/min_terminated_length": 999.8,
|
| 1082 |
+
"eval_entropy": 0.5707040071487427,
|
| 1083 |
+
"eval_frac_reward_zero_std": 1.0,
|
| 1084 |
+
"eval_kl": 0.40999372601509093,
|
| 1085 |
+
"eval_loss": 3.846790423267521e-05,
|
| 1086 |
+
"eval_num_tokens": 5476501.0,
|
| 1087 |
+
"eval_reward": 0.06000000238418579,
|
| 1088 |
+
"eval_reward_std": 0.0,
|
| 1089 |
+
"eval_rewards/env_goofspiel_reward/mean": 0.06000000238418579,
|
| 1090 |
+
"eval_rewards/env_goofspiel_reward/std": 0.06928203105926514,
|
| 1091 |
+
"eval_runtime": 4.0973,
|
| 1092 |
+
"eval_samples_per_second": 2.441,
|
| 1093 |
+
"eval_sampling/importance_sampling_ratio/max": 1.1712106943130494,
|
| 1094 |
+
"eval_sampling/importance_sampling_ratio/mean": 1.011259377002716,
|
| 1095 |
+
"eval_sampling/importance_sampling_ratio/min": 0.8288854002952576,
|
| 1096 |
+
"eval_sampling/sampling_logp_difference/max": 0.17412886619567872,
|
| 1097 |
+
"eval_sampling/sampling_logp_difference/mean": 0.024767952039837837,
|
| 1098 |
+
"eval_steps_per_second": 0.732,
|
| 1099 |
+
"step": 152
|
| 1100 |
}
|
| 1101 |
],
|
| 1102 |
"logging_steps": 5,
|
| 1103 |
"max_steps": 18750,
|
| 1104 |
+
"num_input_tokens_seen": 5476501,
|
| 1105 |
"num_train_epochs": 3,
|
| 1106 |
"save_steps": 500,
|
| 1107 |
"stateful_callbacks": {
|
|
|
|
| 1111 |
"should_evaluate": false,
|
| 1112 |
"should_log": false,
|
| 1113 |
"should_save": true,
|
| 1114 |
+
"should_training_stop": true
|
| 1115 |
},
|
| 1116 |
"attributes": {}
|
| 1117 |
}
|
| 1118 |
},
|
| 1119 |
"total_flos": 0.0,
|
| 1120 |
+
"train_batch_size": 8,
|
| 1121 |
"trial_name": null,
|
| 1122 |
"trial_params": null
|
| 1123 |
}
|
training_args.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 7185
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1b300496f72b512d9eb82d58bc70e9cfecf1e6725146e612c791121039cde76d
|
| 3 |
size 7185
|