Instructions to use TokenBender/issue51-r32-more-online-codex-repro-551-full with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use TokenBender/issue51-r32-more-online-codex-repro-551-full with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
Add reproduction receipts
Browse files- receipts/codex_source_ref.txt +3 -0
- receipts/eval.log +270 -0
- receipts/eval_r32_more_online_masked.jsonl +0 -0
- receipts/eval_r32_more_online_masked.summary.json +15 -0
- receipts/latest_codex_r32_more_online.txt +1 -0
- receipts/run_config.json +65 -0
- receipts/train.log +143 -0
- receipts/train_summary.json +0 -0
receipts/codex_source_ref.txt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
remote_ref=origin/krishna/func
|
| 2 |
+
sha=138e4af08f7fe95ad54c2378ad8e3e375c3f5f83
|
| 3 |
+
source=local git archive from tokenbender mac
|
receipts/eval.log
ADDED
|
@@ -0,0 +1,270 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[transformers] `torch_dtype` is deprecated! Use `dtype` instead!
|
| 2 |
+
|
| 3 |
+
evaluated 4/1007
|
| 4 |
+
evaluated 8/1007
|
| 5 |
+
evaluated 12/1007
|
| 6 |
+
evaluated 16/1007
|
| 7 |
+
evaluated 20/1007
|
| 8 |
+
evaluated 24/1007
|
| 9 |
+
evaluated 28/1007
|
| 10 |
+
evaluated 32/1007
|
| 11 |
+
evaluated 36/1007
|
| 12 |
+
evaluated 40/1007
|
| 13 |
+
evaluated 44/1007
|
| 14 |
+
evaluated 48/1007
|
| 15 |
+
evaluated 52/1007
|
| 16 |
+
evaluated 56/1007
|
| 17 |
+
evaluated 60/1007
|
| 18 |
+
evaluated 64/1007
|
| 19 |
+
evaluated 68/1007
|
| 20 |
+
evaluated 72/1007
|
| 21 |
+
evaluated 76/1007
|
| 22 |
+
evaluated 80/1007
|
| 23 |
+
evaluated 84/1007
|
| 24 |
+
evaluated 88/1007
|
| 25 |
+
evaluated 92/1007
|
| 26 |
+
evaluated 96/1007
|
| 27 |
+
evaluated 100/1007
|
| 28 |
+
evaluated 104/1007
|
| 29 |
+
evaluated 108/1007
|
| 30 |
+
evaluated 112/1007
|
| 31 |
+
evaluated 116/1007
|
| 32 |
+
evaluated 120/1007
|
| 33 |
+
evaluated 124/1007
|
| 34 |
+
evaluated 128/1007
|
| 35 |
+
evaluated 132/1007
|
| 36 |
+
evaluated 136/1007
|
| 37 |
+
evaluated 140/1007
|
| 38 |
+
evaluated 144/1007
|
| 39 |
+
evaluated 148/1007
|
| 40 |
+
evaluated 152/1007
|
| 41 |
+
evaluated 156/1007
|
| 42 |
+
evaluated 160/1007
|
| 43 |
+
evaluated 164/1007
|
| 44 |
+
evaluated 168/1007
|
| 45 |
+
evaluated 172/1007
|
| 46 |
+
evaluated 176/1007
|
| 47 |
+
evaluated 180/1007
|
| 48 |
+
evaluated 184/1007
|
| 49 |
+
evaluated 188/1007
|
| 50 |
+
evaluated 192/1007
|
| 51 |
+
evaluated 196/1007
|
| 52 |
+
evaluated 200/1007
|
| 53 |
+
evaluated 204/1007
|
| 54 |
+
evaluated 208/1007
|
| 55 |
+
evaluated 212/1007
|
| 56 |
+
evaluated 216/1007
|
| 57 |
+
evaluated 220/1007
|
| 58 |
+
evaluated 224/1007
|
| 59 |
+
evaluated 228/1007
|
| 60 |
+
evaluated 232/1007
|
| 61 |
+
evaluated 236/1007
|
| 62 |
+
evaluated 240/1007
|
| 63 |
+
evaluated 244/1007
|
| 64 |
+
evaluated 248/1007
|
| 65 |
+
evaluated 252/1007
|
| 66 |
+
evaluated 256/1007
|
| 67 |
+
evaluated 260/1007
|
| 68 |
+
evaluated 264/1007
|
| 69 |
+
evaluated 268/1007
|
| 70 |
+
evaluated 272/1007
|
| 71 |
+
evaluated 276/1007
|
| 72 |
+
evaluated 280/1007
|
| 73 |
+
evaluated 284/1007
|
| 74 |
+
evaluated 288/1007
|
| 75 |
+
evaluated 292/1007
|
| 76 |
+
evaluated 296/1007
|
| 77 |
+
evaluated 300/1007
|
| 78 |
+
evaluated 304/1007
|
| 79 |
+
evaluated 308/1007
|
| 80 |
+
evaluated 312/1007
|
| 81 |
+
evaluated 316/1007
|
| 82 |
+
evaluated 320/1007
|
| 83 |
+
evaluated 324/1007
|
| 84 |
+
evaluated 328/1007
|
| 85 |
+
evaluated 332/1007
|
| 86 |
+
evaluated 336/1007
|
| 87 |
+
evaluated 340/1007
|
| 88 |
+
evaluated 344/1007
|
| 89 |
+
evaluated 348/1007
|
| 90 |
+
evaluated 352/1007
|
| 91 |
+
evaluated 356/1007
|
| 92 |
+
evaluated 360/1007
|
| 93 |
+
evaluated 364/1007
|
| 94 |
+
evaluated 368/1007
|
| 95 |
+
evaluated 372/1007
|
| 96 |
+
evaluated 376/1007
|
| 97 |
+
evaluated 380/1007
|
| 98 |
+
evaluated 384/1007
|
| 99 |
+
evaluated 388/1007
|
| 100 |
+
evaluated 392/1007
|
| 101 |
+
evaluated 396/1007
|
| 102 |
+
evaluated 400/1007
|
| 103 |
+
evaluated 404/1007
|
| 104 |
+
evaluated 408/1007
|
| 105 |
+
evaluated 412/1007
|
| 106 |
+
evaluated 416/1007
|
| 107 |
+
evaluated 420/1007
|
| 108 |
+
evaluated 424/1007
|
| 109 |
+
evaluated 428/1007
|
| 110 |
+
evaluated 432/1007
|
| 111 |
+
evaluated 436/1007
|
| 112 |
+
evaluated 440/1007
|
| 113 |
+
evaluated 444/1007
|
| 114 |
+
evaluated 448/1007
|
| 115 |
+
evaluated 452/1007
|
| 116 |
+
evaluated 456/1007
|
| 117 |
+
evaluated 460/1007
|
| 118 |
+
evaluated 464/1007
|
| 119 |
+
evaluated 468/1007
|
| 120 |
+
evaluated 472/1007
|
| 121 |
+
evaluated 476/1007
|
| 122 |
+
evaluated 480/1007
|
| 123 |
+
evaluated 484/1007
|
| 124 |
+
evaluated 488/1007
|
| 125 |
+
evaluated 492/1007
|
| 126 |
+
evaluated 496/1007
|
| 127 |
+
evaluated 500/1007
|
| 128 |
+
evaluated 504/1007
|
| 129 |
+
evaluated 508/1007
|
| 130 |
+
evaluated 512/1007
|
| 131 |
+
evaluated 516/1007
|
| 132 |
+
evaluated 520/1007
|
| 133 |
+
evaluated 524/1007
|
| 134 |
+
evaluated 528/1007
|
| 135 |
+
evaluated 532/1007
|
| 136 |
+
evaluated 536/1007
|
| 137 |
+
evaluated 540/1007
|
| 138 |
+
evaluated 544/1007
|
| 139 |
+
evaluated 548/1007
|
| 140 |
+
evaluated 552/1007
|
| 141 |
+
evaluated 556/1007
|
| 142 |
+
evaluated 560/1007
|
| 143 |
+
evaluated 564/1007
|
| 144 |
+
evaluated 568/1007
|
| 145 |
+
evaluated 572/1007
|
| 146 |
+
evaluated 576/1007
|
| 147 |
+
evaluated 580/1007
|
| 148 |
+
evaluated 584/1007
|
| 149 |
+
evaluated 588/1007
|
| 150 |
+
evaluated 592/1007
|
| 151 |
+
evaluated 596/1007
|
| 152 |
+
evaluated 600/1007
|
| 153 |
+
evaluated 604/1007
|
| 154 |
+
evaluated 608/1007
|
| 155 |
+
evaluated 612/1007
|
| 156 |
+
evaluated 616/1007
|
| 157 |
+
evaluated 620/1007
|
| 158 |
+
evaluated 624/1007
|
| 159 |
+
evaluated 628/1007
|
| 160 |
+
evaluated 632/1007
|
| 161 |
+
evaluated 636/1007
|
| 162 |
+
evaluated 640/1007
|
| 163 |
+
evaluated 644/1007
|
| 164 |
+
evaluated 648/1007
|
| 165 |
+
evaluated 652/1007
|
| 166 |
+
evaluated 656/1007
|
| 167 |
+
evaluated 660/1007
|
| 168 |
+
evaluated 664/1007
|
| 169 |
+
evaluated 668/1007
|
| 170 |
+
evaluated 672/1007
|
| 171 |
+
evaluated 676/1007
|
| 172 |
+
evaluated 680/1007
|
| 173 |
+
evaluated 684/1007
|
| 174 |
+
evaluated 688/1007
|
| 175 |
+
evaluated 692/1007
|
| 176 |
+
evaluated 696/1007
|
| 177 |
+
evaluated 700/1007
|
| 178 |
+
evaluated 704/1007
|
| 179 |
+
evaluated 708/1007
|
| 180 |
+
evaluated 712/1007
|
| 181 |
+
evaluated 716/1007
|
| 182 |
+
evaluated 720/1007
|
| 183 |
+
evaluated 724/1007
|
| 184 |
+
evaluated 728/1007
|
| 185 |
+
evaluated 732/1007
|
| 186 |
+
evaluated 736/1007
|
| 187 |
+
evaluated 740/1007
|
| 188 |
+
evaluated 744/1007
|
| 189 |
+
evaluated 748/1007
|
| 190 |
+
evaluated 752/1007
|
| 191 |
+
evaluated 756/1007
|
| 192 |
+
evaluated 760/1007
|
| 193 |
+
evaluated 764/1007
|
| 194 |
+
evaluated 768/1007
|
| 195 |
+
evaluated 772/1007
|
| 196 |
+
evaluated 776/1007
|
| 197 |
+
evaluated 780/1007
|
| 198 |
+
evaluated 784/1007
|
| 199 |
+
evaluated 788/1007
|
| 200 |
+
evaluated 792/1007
|
| 201 |
+
evaluated 796/1007
|
| 202 |
+
evaluated 800/1007
|
| 203 |
+
evaluated 804/1007
|
| 204 |
+
evaluated 808/1007
|
| 205 |
+
evaluated 812/1007
|
| 206 |
+
evaluated 816/1007
|
| 207 |
+
evaluated 820/1007
|
| 208 |
+
evaluated 824/1007
|
| 209 |
+
evaluated 828/1007
|
| 210 |
+
evaluated 832/1007
|
| 211 |
+
evaluated 836/1007
|
| 212 |
+
evaluated 840/1007
|
| 213 |
+
evaluated 844/1007
|
| 214 |
+
evaluated 848/1007
|
| 215 |
+
evaluated 852/1007
|
| 216 |
+
evaluated 856/1007
|
| 217 |
+
evaluated 860/1007
|
| 218 |
+
evaluated 864/1007
|
| 219 |
+
evaluated 868/1007
|
| 220 |
+
evaluated 872/1007
|
| 221 |
+
evaluated 876/1007
|
| 222 |
+
evaluated 880/1007
|
| 223 |
+
evaluated 884/1007
|
| 224 |
+
evaluated 888/1007
|
| 225 |
+
evaluated 892/1007
|
| 226 |
+
evaluated 896/1007
|
| 227 |
+
evaluated 900/1007
|
| 228 |
+
evaluated 904/1007
|
| 229 |
+
evaluated 908/1007
|
| 230 |
+
evaluated 912/1007
|
| 231 |
+
evaluated 916/1007
|
| 232 |
+
evaluated 920/1007
|
| 233 |
+
evaluated 924/1007
|
| 234 |
+
evaluated 928/1007
|
| 235 |
+
evaluated 932/1007
|
| 236 |
+
evaluated 936/1007
|
| 237 |
+
evaluated 940/1007
|
| 238 |
+
evaluated 944/1007
|
| 239 |
+
evaluated 948/1007
|
| 240 |
+
evaluated 952/1007
|
| 241 |
+
evaluated 956/1007
|
| 242 |
+
evaluated 960/1007
|
| 243 |
+
evaluated 964/1007
|
| 244 |
+
evaluated 968/1007
|
| 245 |
+
evaluated 972/1007
|
| 246 |
+
evaluated 976/1007
|
| 247 |
+
evaluated 980/1007
|
| 248 |
+
evaluated 984/1007
|
| 249 |
+
evaluated 988/1007
|
| 250 |
+
evaluated 992/1007
|
| 251 |
+
evaluated 996/1007
|
| 252 |
+
evaluated 1000/1007
|
| 253 |
+
evaluated 1004/1007
|
| 254 |
+
evaluated 1007/1007
|
| 255 |
+
{
|
| 256 |
+
"examples": 1007,
|
| 257 |
+
"exact_correct": 551,
|
| 258 |
+
"exact_accuracy": 0.5471698113207547,
|
| 259 |
+
"raw_exact_correct": 537,
|
| 260 |
+
"raw_exact_accuracy": 0.5332671300893744,
|
| 261 |
+
"normalized_exact_correct": 551,
|
| 262 |
+
"normalized_exact_accuracy": 0.5471698113207547,
|
| 263 |
+
"reported_metric": "normalized_exact",
|
| 264 |
+
"bfcl_canonicalization_prompt": true,
|
| 265 |
+
"mask_topk": 160000,
|
| 266 |
+
"attribution": "runs/issue49_bfcl_single_call/relp_full.npz",
|
| 267 |
+
"generations": "runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/eval_r32_more_online_masked.jsonl",
|
| 268 |
+
"note": "crude exact structured match against BFCL simple possible answers"
|
| 269 |
+
}
|
| 270 |
+
EVAL_STATUS=0
|
receipts/eval_r32_more_online_masked.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
receipts/eval_r32_more_online_masked.summary.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"examples": 1007,
|
| 3 |
+
"exact_correct": 551,
|
| 4 |
+
"exact_accuracy": 0.5471698113207547,
|
| 5 |
+
"raw_exact_correct": 537,
|
| 6 |
+
"raw_exact_accuracy": 0.5332671300893744,
|
| 7 |
+
"normalized_exact_correct": 551,
|
| 8 |
+
"normalized_exact_accuracy": 0.5471698113207547,
|
| 9 |
+
"reported_metric": "normalized_exact",
|
| 10 |
+
"bfcl_canonicalization_prompt": true,
|
| 11 |
+
"mask_topk": 160000,
|
| 12 |
+
"attribution": "runs/issue49_bfcl_single_call/relp_full.npz",
|
| 13 |
+
"generations": "runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/eval_r32_more_online_masked.jsonl",
|
| 14 |
+
"note": "crude exact structured match against BFCL simple possible answers"
|
| 15 |
+
}
|
receipts/latest_codex_r32_more_online.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857
|
receipts/run_config.json
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"method": "gkd_aopd_hybrid_sampled_token_v0",
|
| 3 |
+
"contract": {
|
| 4 |
+
"rollout": "current LoRA student under MLP zero-isolated keep mask",
|
| 5 |
+
"teacher": "same base model with adapter disabled and no mask",
|
| 6 |
+
"loss": "offline gold-prefix teacher KL/CE plus positive-advantage sampled-token AOPD"
|
| 7 |
+
},
|
| 8 |
+
"args": {
|
| 9 |
+
"base_model": "/root/models/Qwen3-8B",
|
| 10 |
+
"init_adapter": "/root/hf_artifacts/issue49-k160-r32-len1024-adapter",
|
| 11 |
+
"lora_r": 32,
|
| 12 |
+
"lora_alpha": 64,
|
| 13 |
+
"lora_dropout": 0.05,
|
| 14 |
+
"target_modules": "all-linear",
|
| 15 |
+
"use_rslora": true,
|
| 16 |
+
"train_jsonl": "data/bfcl_strict_10k_mix_len1024/train.jsonl",
|
| 17 |
+
"attribution": "runs/issue49_bfcl_single_call/relp_full.npz",
|
| 18 |
+
"topk": 160000,
|
| 19 |
+
"out_dir": "runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/r32_more_online",
|
| 20 |
+
"device": "cuda",
|
| 21 |
+
"dtype": "bfloat16",
|
| 22 |
+
"seed": 42,
|
| 23 |
+
"max_rows": 1000,
|
| 24 |
+
"max_seq_length": 1024,
|
| 25 |
+
"max_new_tokens": 192,
|
| 26 |
+
"n_calib": 128,
|
| 27 |
+
"rollout_batch_size": 8,
|
| 28 |
+
"train_batch_size": 1,
|
| 29 |
+
"offline_batch_size": 1,
|
| 30 |
+
"grad_accum": 8,
|
| 31 |
+
"updates_per_rollout": 1,
|
| 32 |
+
"max_updates": 125,
|
| 33 |
+
"lr": 2e-05,
|
| 34 |
+
"weight_decay": 0.0,
|
| 35 |
+
"warmup_ratio": 0.05,
|
| 36 |
+
"max_grad_norm": 1.0,
|
| 37 |
+
"temperature": 0.7,
|
| 38 |
+
"top_p": 0.95,
|
| 39 |
+
"do_sample": true,
|
| 40 |
+
"adv_scale": 1.0,
|
| 41 |
+
"kl_tau": 0.001,
|
| 42 |
+
"online_opd_beta": 0.4,
|
| 43 |
+
"offline_kl_beta": 0.8,
|
| 44 |
+
"ce_beta": 0.05,
|
| 45 |
+
"kl_temperature": 1.0,
|
| 46 |
+
"positive_opd_only": true,
|
| 47 |
+
"dppo_threshold": 0.2,
|
| 48 |
+
"mask_mode": "zero",
|
| 49 |
+
"eval_rollout_every": 10,
|
| 50 |
+
"save_every": 25,
|
| 51 |
+
"save_merged": true,
|
| 52 |
+
"bfcl_canonicalization_prompt": true
|
| 53 |
+
},
|
| 54 |
+
"n_rows": 1000,
|
| 55 |
+
"n_layers": 36,
|
| 56 |
+
"d_ffn": 12288,
|
| 57 |
+
"mask_kept": 160000,
|
| 58 |
+
"adapter_init": {
|
| 59 |
+
"mode": "init_adapter",
|
| 60 |
+
"path": "/root/hf_artifacts/issue49-k160-r32-len1024-adapter"
|
| 61 |
+
},
|
| 62 |
+
"warmup_steps": 6,
|
| 63 |
+
"logs": [],
|
| 64 |
+
"checkpoints": []
|
| 65 |
+
}
|
receipts/train.log
ADDED
|
@@ -0,0 +1,143 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[data] rows=1000 train=data/bfcl_strict_10k_mix_len1024/train.jsonl
|
| 2 |
+
[data] kept=1000 dropped=0
|
| 3 |
+
[model] base=/root/models/Qwen3-8B init=/root/hf_artifacts/issue49-k160-r32-len1024-adapter
|
| 4 |
+
[transformers] `torch_dtype` is deprecated! Use `dtype` instead!
|
| 5 |
+
|
| 6 |
+
[mask] topk=160000 kept=160000 attribution=runs/issue49_bfcl_single_call/relp_full.npz
|
| 7 |
+
trainable params: 87,293,952 || all params: 8,278,029,312 || trainable%: 1.0545
|
| 8 |
+
{"rollout": 1, "update": 1, "updates_added": 1, "lr": 3.3333333333333333e-06, "elapsed_s": 27.74144220352173, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 277}, "train": {"loss": -0.010878981061978266, "pg_loss": -0.010878981061978266, "kl_loss": 0.0, "teacher_adv": -0.21376380375295412, "teacher_adv_pos_frac": 0.9483879804611206, "opd_selected_frac": 0.9483879804611206, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.025562286376953125, "offline_kl": 0.029056072235107422, "ce": 0.04706144332885742, "weighted_online_loss": -0.004351592475359212, "total_loss": 0.021210693899774924, "tokens": 34.625}}
|
| 9 |
+
{"rollout": 2, "update": 2, "updates_added": 1, "lr": 6.666666666666667e-06, "elapsed_s": 57.45381689071655, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 308}, "train": {"loss": -0.019993211142718792, "pg_loss": -0.019993211142718792, "kl_loss": 0.0, "teacher_adv": -0.4033845847006887, "teacher_adv_pos_frac": 0.946535274386406, "opd_selected_frac": 0.946535274386406, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.086578369140625, "offline_kl": 0.09735870361328125, "ce": 0.171600341796875, "weighted_online_loss": -0.007997284570592456, "total_loss": 0.0785810831002891, "tokens": 38.5}}
|
| 10 |
+
{"rollout": 3, "update": 3, "updates_added": 1, "lr": 1e-05, "elapsed_s": 82.55218124389648, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 249}, "train": {"loss": -0.023243495766109845, "pg_loss": -0.023243495766109845, "kl_loss": 0.0, "teacher_adv": -0.595445566120361, "teacher_adv_pos_frac": 0.9072228968143463, "opd_selected_frac": 0.9072228968143463, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.09975838661193848, "offline_kl": 0.1174854040145874, "ce": 0.11131983995437622, "weighted_online_loss": -0.009297398476519447, "total_loss": 0.09046098694670945, "tokens": 31.125}}
|
| 11 |
+
{"rollout": 4, "update": 4, "updates_added": 1, "lr": 1.3333333333333333e-05, "elapsed_s": 108.56254553794861, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 257}, "train": {"loss": -0.03088429150011507, "pg_loss": -0.03088429150011507, "kl_loss": 0.0, "teacher_adv": -0.20690712332361727, "teacher_adv_pos_frac": 0.9411146715283394, "opd_selected_frac": 0.9411146715283394, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.11254596710205078, "offline_kl": 0.12900304794311523, "ce": 0.18909978866577148, "weighted_online_loss": -0.012353716883808374, "total_loss": 0.10019224882125854, "tokens": 32.125}}
|
| 12 |
+
{"rollout": 5, "update": 5, "updates_added": 1, "lr": 1.6666666666666667e-05, "elapsed_s": 137.8877878189087, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 303}, "train": {"loss": -0.009125907090492547, "pg_loss": -0.009125907090492547, "kl_loss": 0.0, "teacher_adv": -0.20826974607189186, "teacher_adv_pos_frac": 0.9531974494457245, "opd_selected_frac": 0.9531974494457245, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.11085343360900879, "offline_kl": 0.13287103176116943, "ce": 0.09356439113616943, "weighted_online_loss": -0.0036503629526123405, "total_loss": 0.1072030728100799, "tokens": 37.875}}
|
| 13 |
+
{"rollout": 6, "update": 6, "updates_added": 1, "lr": 2e-05, "elapsed_s": 166.94009709358215, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 296}, "train": {"loss": -0.005991066122078337, "pg_loss": -0.005991066122078337, "kl_loss": 0.0, "teacher_adv": -0.03611905185607611, "teacher_adv_pos_frac": 0.9502050653100014, "opd_selected_frac": 0.9502050653100014, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.14380693435668945, "offline_kl": 0.16127920150756836, "ce": 0.29424238204956055, "weighted_online_loss": -0.0023964265319591505, "total_loss": 0.14141050836769864, "tokens": 37.0}}
|
| 14 |
+
{"rollout": 7, "update": 7, "updates_added": 1, "lr": 1.9996515418688493e-05, "elapsed_s": 196.04023456573486, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 295}, "train": {"loss": -0.026489450014196336, "pg_loss": -0.026489450014196336, "kl_loss": 0.0, "teacher_adv": -0.3734949378413148, "teacher_adv_pos_frac": 0.929985374212265, "opd_selected_frac": 0.929985374212265, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.25345611572265625, "offline_kl": 0.292266845703125, "ce": 0.4022979736328125, "weighted_online_loss": -0.010595780448056757, "total_loss": 0.2428603389998898, "tokens": 36.875}}
|
| 15 |
+
{"rollout": 8, "update": 8, "updates_added": 1, "lr": 1.998606410321534e-05, "elapsed_s": 233.94550561904907, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 417}, "train": {"loss": -0.037320026982342824, "pg_loss": -0.037320026982342824, "kl_loss": 0.0, "teacher_adv": -0.34848144749412313, "teacher_adv_pos_frac": 0.8777374774217606, "opd_selected_frac": 0.8777374774217606, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.12473189830780029, "offline_kl": 0.14596056938171387, "ce": 0.15653234720230103, "weighted_online_loss": -0.014928010801668279, "total_loss": 0.10980388819734799, "tokens": 52.125}}
|
| 16 |
+
{"rollout": 9, "update": 9, "updates_added": 1, "lr": 1.9968653337272262e-05, "elapsed_s": 261.64218640327454, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 284}, "train": {"loss": -0.018826504801836563, "pg_loss": -0.018826504801836563, "kl_loss": 0.0, "teacher_adv": -0.12933818040437473, "teacher_adv_pos_frac": 0.9496986567974091, "opd_selected_frac": 0.9496986567974091, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.024078369140625, "offline_kl": 0.025438308715820312, "ce": 0.07521247863769531, "weighted_online_loss": -0.00753060201441258, "total_loss": 0.016547767561860383, "tokens": 35.5}}
|
| 17 |
+
{"rollout": 10, "update": 10, "updates_added": 1, "lr": 1.9944295254705187e-05, "elapsed_s": 306.3442749977112, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 507}, "train": {"loss": -0.02793700415350031, "pg_loss": -0.02793700415350031, "kl_loss": 0.0, "teacher_adv": -0.10557314603647683, "teacher_adv_pos_frac": 0.9407447725534439, "opd_selected_frac": 0.9407447725534439, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.08551549911499023, "offline_kl": 0.0964512825012207, "ce": 0.1671299934387207, "weighted_online_loss": -0.011174801729794126, "total_loss": 0.07434069603186799, "tokens": 63.375}}
|
| 18 |
+
{"rollout": 11, "update": 11, "updates_added": 1, "lr": 1.9913006831057967e-05, "elapsed_s": 333.6601507663727, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 274}, "train": {"loss": -0.026584985898807645, "pg_loss": -0.026584985898807645, "kl_loss": 0.0, "teacher_adv": -0.29928471718449146, "teacher_adv_pos_frac": 0.9301197156310081, "opd_selected_frac": 0.9301197156310081, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.008555889129638672, "offline_kl": 0.00857400894165039, "ce": 0.03350949287414551, "weighted_online_loss": -0.01063399447593838, "total_loss": -0.002078105288092047, "tokens": 34.25}}
|
| 19 |
+
{"rollout": 12, "update": 12, "updates_added": 1, "lr": 1.9874809871741877e-05, "elapsed_s": 377.1379773616791, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 484}, "train": {"loss": -0.01181748117232928, "pg_loss": -0.01181748117232928, "kl_loss": 0.0, "teacher_adv": -0.32131374671007507, "teacher_adv_pos_frac": 0.9270229414105415, "opd_selected_frac": 0.9270229414105415, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.022127151489257812, "offline_kl": 0.022919654846191406, "ce": 0.0751349925994873, "weighted_online_loss": -0.004726992534415331, "total_loss": 0.017400158954842482, "tokens": 60.5}}
|
| 20 |
+
{"rollout": 13, "update": 13, "updates_added": 1, "lr": 1.982973099683902e-05, "elapsed_s": 418.4671747684479, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 454}, "train": {"loss": -0.028932540044479538, "pg_loss": -0.028932540044479538, "kl_loss": 0.0, "teacher_adv": -0.34122188371839, "teacher_adv_pos_frac": 0.9075045138597488, "opd_selected_frac": 0.9075045138597488, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.025283455848693848, "offline_kl": 0.024255037307739258, "ce": 0.11723828315734863, "weighted_online_loss": -0.011573016379770706, "total_loss": 0.013710439729038626, "tokens": 56.75}}
|
| 21 |
+
{"rollout": 14, "update": 14, "updates_added": 1, "lr": 1.977780162255041e-05, "elapsed_s": 445.0928146839142, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 267}, "train": {"loss": -0.012170710293503362, "pg_loss": -0.012170710293503362, "kl_loss": 0.0, "teacher_adv": -0.32204647107027995, "teacher_adv_pos_frac": 0.927700899541378, "opd_selected_frac": 0.927700899541378, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.003037393093109131, "offline_kl": 0.0032862424850463867, "ce": 0.00803518295288086, "weighted_online_loss": -0.004868284207077522, "total_loss": -0.0018308911203348543, "tokens": 33.375}}
|
| 22 |
+
{"rollout": 15, "update": 15, "updates_added": 1, "lr": 1.9719057939301477e-05, "elapsed_s": 475.18583059310913, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 316}, "train": {"loss": -0.006443917187425541, "pg_loss": -0.006443917187425541, "kl_loss": 0.0, "teacher_adv": -0.057058752652665135, "teacher_adv_pos_frac": 0.9177775681018829, "opd_selected_frac": 0.9177775681018829, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.01888716220855713, "offline_kl": 0.020358562469482422, "ce": 0.05258655548095703, "weighted_online_loss": -0.0025775668436835986, "total_loss": 0.016309595404891297, "tokens": 39.5}}
|
| 23 |
+
{"rollout": 16, "update": 16, "updates_added": 1, "lr": 1.9653540886520387e-05, "elapsed_s": 499.6275577545166, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 235}, "train": {"loss": -0.005502504889591364, "pg_loss": -0.005502504889591364, "kl_loss": 0.0, "teacher_adv": -0.02562424904681393, "teacher_adv_pos_frac": 0.8746209740638733, "opd_selected_frac": 0.8746209740638733, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.07345366477966309, "offline_kl": 0.08304047584533691, "ce": 0.13593697547912598, "weighted_online_loss": -0.0022010019620211096, "total_loss": 0.07125266078219283, "tokens": 29.375}}
|
| 24 |
+
{"rollout": 17, "update": 17, "updates_added": 1, "lr": 1.9581296124106682e-05, "elapsed_s": 526.6508371829987, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 275}, "train": {"loss": -0.009509171422905638, "pg_loss": -0.009509171422905638, "kl_loss": 0.0, "teacher_adv": -0.16899409081997874, "teacher_adv_pos_frac": 0.8579553365707397, "opd_selected_frac": 0.8579553365707397, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.011006414890289307, "offline_kl": 0.009801477193832397, "ce": 0.06390342116355896, "weighted_online_loss": -0.003803668621912948, "total_loss": 0.007202746535767801, "tokens": 34.375}}
|
| 25 |
+
{"rollout": 18, "update": 18, "updates_added": 1, "lr": 1.9502374000610152e-05, "elapsed_s": 553.5995314121246, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 271}, "train": {"loss": -0.010264740325510502, "pg_loss": -0.010264740325510502, "kl_loss": 0.0, "teacher_adv": -0.3266243701800704, "teacher_adv_pos_frac": 0.8824442774057388, "opd_selected_frac": 0.8824442774057388, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.1257314682006836, "offline_kl": 0.14116549491882324, "ce": 0.252272367477417, "weighted_online_loss": -0.004105896114197094, "total_loss": 0.12162556980911177, "tokens": 33.875}}
|
| 26 |
+
{"rollout": 19, "update": 19, "updates_added": 1, "lr": 1.941682951814212e-05, "elapsed_s": 583.2185690402985, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 310}, "train": {"loss": -0.013121407610015012, "pg_loss": -0.013121407610015012, "kl_loss": 0.0, "teacher_adv": -0.31685128485696623, "teacher_adv_pos_frac": 0.8166598230600357, "opd_selected_frac": 0.8166598230600357, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.04409217834472656, "offline_kl": 0.04730987548828125, "ce": 0.12291908264160156, "weighted_online_loss": -0.005248563196801115, "total_loss": 0.03884361477685161, "tokens": 38.75}}
|
| 27 |
+
{"rollout": 20, "update": 20, "updates_added": 1, "lr": 1.932472229404356e-05, "elapsed_s": 613.4990367889404, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 318}, "train": {"loss": -0.0031482516205869615, "pg_loss": -0.0031482516205869615, "kl_loss": 0.0, "teacher_adv": -0.010011400459916331, "teacher_adv_pos_frac": 0.8847223967313766, "opd_selected_frac": 0.8847223967313766, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.009646773338317871, "offline_kl": 0.010191202163696289, "ce": 0.029886484146118164, "weighted_online_loss": -0.001259300659512519, "total_loss": 0.008387472578760935, "tokens": 39.75}}
|
| 28 |
+
{"rollout": 21, "update": 21, "updates_added": 1, "lr": 1.922611651933683e-05, "elapsed_s": 640.2883412837982, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 273}, "train": {"loss": -0.03049171663042216, "pg_loss": -0.03049171663042216, "kl_loss": 0.0, "teacher_adv": -0.19074536968764733, "teacher_adv_pos_frac": 0.8469238430261612, "opd_selected_frac": 0.8469238430261612, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.048754215240478516, "offline_kl": 0.0569305419921875, "ce": 0.06604909896850586, "weighted_online_loss": -0.012196686834613502, "total_loss": 0.03655752925260458, "tokens": 34.125}}
|
| 29 |
+
{"rollout": 22, "update": 22, "updates_added": 1, "lr": 1.912108091398988e-05, "elapsed_s": 665.8441827297211, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 256}, "train": {"loss": -0.007860905208872282, "pg_loss": -0.007860905208872282, "kl_loss": 0.0, "teacher_adv": -0.2681308827250177, "teacher_adv_pos_frac": 0.7832186967134476, "opd_selected_frac": 0.7832186967134476, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0399627685546875, "offline_kl": 0.046176910400390625, "ce": 0.061229705810546875, "weighted_online_loss": -0.003144362043713045, "total_loss": 0.03681840639910661, "tokens": 32.0}}
|
| 30 |
+
{"rollout": 23, "update": 23, "updates_added": 1, "lr": 1.900968867902419e-05, "elapsed_s": 692.0538473129272, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 265}, "train": {"loss": -0.0057855815248331055, "pg_loss": -0.0057855815248331055, "kl_loss": 0.0, "teacher_adv": -0.17204108566511422, "teacher_adv_pos_frac": 0.850994735956192, "opd_selected_frac": 0.850994735956192, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.05491971969604492, "offline_kl": 0.061252593994140625, "ce": 0.11898231506347656, "weighted_online_loss": -0.0023142326481320197, "total_loss": 0.05260548513251706, "tokens": 33.125}}
|
| 31 |
+
{"rollout": 24, "update": 24, "updates_added": 1, "lr": 1.8892017445499812e-05, "elapsed_s": 722.1788568496704, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 312}, "train": {"loss": -0.004671681877880474, "pg_loss": -0.004671681877880474, "kl_loss": 0.0, "teacher_adv": -0.11518581995642307, "teacher_adv_pos_frac": 0.8798587247729301, "opd_selected_frac": 0.8798587247729301, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03317582607269287, "offline_kl": 0.03519022464752197, "ce": 0.10017478466033936, "weighted_online_loss": -0.001868672786258685, "total_loss": 0.03130715325823985, "tokens": 39.0}}
|
| 32 |
+
{"rollout": 25, "update": 25, "updates_added": 1, "lr": 1.876814922041299e-05, "elapsed_s": 747.2936553955078, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 249}, "train": {"loss": -0.007617811163072474, "pg_loss": -0.007617811163072474, "kl_loss": 0.0, "teacher_adv": -0.2137641423723835, "teacher_adv_pos_frac": 0.8599288463592529, "opd_selected_frac": 0.8599288463592529, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0494382381439209, "offline_kl": 0.05492877960205078, "ce": 0.1135450005531311, "weighted_online_loss": -0.003047124486329267, "total_loss": 0.046391112642595544, "tokens": 31.125}}
|
| 33 |
+
/root/work/circuit-shotting/.venv/lib/python3.12/site-packages/peft/utils/save_and_load.py:372: UserWarning: Could not find a config file in /workspace/hf_cache/hub/models--Qwen--Qwen3-8B/snapshots/b968826d9c46dd6066d109eabc6255188de91218 - will assume that the vocabulary was not modified.
|
| 34 |
+
warnings.warn(
|
| 35 |
+
[checkpoint] update=25 adapter=runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/r32_more_online/checkpoints/update_000025/adapter
|
| 36 |
+
{"rollout": 26, "update": 26, "updates_added": 1, "lr": 1.8638170329544164e-05, "elapsed_s": 774.4814410209656, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 267}, "train": {"loss": -0.008694339252542704, "pg_loss": -0.008694339252542704, "kl_loss": 0.0, "teacher_adv": -0.021693322467399412, "teacher_adv_pos_frac": 0.8224471360445023, "opd_selected_frac": 0.8224471360445023, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.005158662796020508, "offline_kl": 0.003803253173828125, "ce": 0.04236006736755371, "weighted_online_loss": -0.0034777357659550034, "total_loss": 0.0016809270728117554, "tokens": 33.375}}
|
| 37 |
+
{"rollout": 27, "update": 27, "updates_added": 1, "lr": 1.8502171357296144e-05, "elapsed_s": 801.6256437301636, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 270}, "train": {"loss": -0.011821121312095784, "pg_loss": -0.011821121312095784, "kl_loss": 0.0, "teacher_adv": -0.13147640245733783, "teacher_adv_pos_frac": 0.8200302496552467, "opd_selected_frac": 0.8200302496552467, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0668521523475647, "offline_kl": 0.07780250906944275, "ce": 0.09194743633270264, "weighted_online_loss": -0.004728448648165795, "total_loss": 0.062123704028635984, "tokens": 33.75}}
|
| 38 |
+
{"rollout": 28, "update": 28, "updates_added": 1, "lr": 1.8360247083564343e-05, "elapsed_s": 829.896320104599, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 289}, "train": {"loss": -0.009311691628226981, "pg_loss": -0.009311691628226981, "kl_loss": 0.0, "teacher_adv": -0.3261171583280884, "teacher_adv_pos_frac": 0.8131098300218582, "opd_selected_frac": 0.8131098300218582, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.04703482985496521, "offline_kl": 0.048107683658599854, "ce": 0.16996440291404724, "weighted_online_loss": -0.0037246766794396535, "total_loss": 0.04331015301613661, "tokens": 36.125}}
|
| 39 |
+
{"rollout": 29, "update": 29, "updates_added": 1, "lr": 1.8212496417683135e-05, "elapsed_s": 853.7840025424957, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 232}, "train": {"loss": -0.036502278271655086, "pg_loss": -0.036502278271655086, "kl_loss": 0.0, "teacher_adv": -0.7356477647645079, "teacher_adv_pos_frac": 0.800651378929615, "opd_selected_frac": 0.800651378929615, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.028103947639465332, "offline_kl": 0.030510425567626953, "ce": 0.07303085923194885, "weighted_online_loss": -0.014600911178604292, "total_loss": 0.013503036454494577, "tokens": 29.0}}
|
| 40 |
+
{"rollout": 30, "update": 30, "updates_added": 1, "lr": 1.805902232949435e-05, "elapsed_s": 879.7541060447693, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 253}, "train": {"loss": -0.003943655960028991, "pg_loss": -0.003943655960028991, "kl_loss": 0.0, "teacher_adv": -0.173453407958732, "teacher_adv_pos_frac": 0.8106145560741425, "opd_selected_frac": 0.8106145560741425, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.15775132179260254, "offline_kl": 0.17460834980010986, "ce": 0.36119675636291504, "weighted_online_loss": -0.0015774624516780023, "total_loss": 0.1561738621967379, "tokens": 31.625}}
|
| 41 |
+
{"rollout": 31, "update": 31, "updates_added": 1, "lr": 1.789993177758588e-05, "elapsed_s": 906.8185706138611, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 273}, "train": {"loss": -0.017077673932362814, "pg_loss": -0.017077673932362814, "kl_loss": 0.0, "teacher_adv": -0.5846366627774842, "teacher_adv_pos_frac": 0.8134342730045319, "opd_selected_frac": 0.8134342730045319, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.008405447006225586, "offline_kl": 0.0083693265914917, "ce": 0.03445470333099365, "weighted_online_loss": -0.006831069702457171, "total_loss": 0.0015743772164569236, "tokens": 34.125}}
|
| 42 |
+
{"rollout": 32, "update": 32, "updates_added": 1, "lr": 1.773533563475053e-05, "elapsed_s": 932.5899419784546, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 250}, "train": {"loss": -0.010922932135144947, "pg_loss": -0.010922932135144947, "kl_loss": 0.0, "teacher_adv": -0.12914039642782882, "teacher_adv_pos_frac": 0.800095446407795, "opd_selected_frac": 0.800095446407795, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0046176910400390625, "offline_kl": 0.004938602447509766, "ce": 0.013469696044921875, "weighted_online_loss": -0.004369172993392567, "total_loss": 0.0002485181175870821, "tokens": 31.25}}
|
| 43 |
+
{"rollout": 33, "update": 33, "updates_added": 1, "lr": 1.7565348610716963e-05, "elapsed_s": 960.1315338611603, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 276}, "train": {"loss": -0.0028372966571623692, "pg_loss": -0.0028372966571623692, "kl_loss": 0.0, "teacher_adv": -0.13027930411135458, "teacher_adv_pos_frac": 0.8099153116345406, "opd_selected_frac": 0.8099153116345406, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.029887259006500244, "offline_kl": 0.02889573574066162, "ce": 0.13355207443237305, "weighted_online_loss": -0.0011349186802362965, "total_loss": 0.028752341357176192, "tokens": 34.5}}
|
| 44 |
+
{"rollout": 34, "update": 34, "updates_added": 1, "lr": 1.7390089172206594e-05, "elapsed_s": 984.6213049888611, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 239}, "train": {"loss": -0.0075084136333316565, "pg_loss": -0.0075084136333316565, "kl_loss": 0.0, "teacher_adv": -0.12357825855360716, "teacher_adv_pos_frac": 0.7675380930304527, "opd_selected_frac": 0.7675380930304527, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.007025003433227539, "offline_kl": 0.0064893364906311035, "ce": 0.03681468963623047, "weighted_online_loss": -0.003003365546646819, "total_loss": 0.004021637734695105, "tokens": 29.875}}
|
| 45 |
+
{"rollout": 35, "update": 35, "updates_added": 1, "lr": 1.720967946037225e-05, "elapsed_s": 1013.440726518631, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 293}, "train": {"loss": -0.008202193796023494, "pg_loss": -0.008202193796023494, "kl_loss": 0.0, "teacher_adv": -0.3906367107956612, "teacher_adv_pos_frac": 0.7527794763445854, "opd_selected_frac": 0.7527794763445854, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.037067711353302, "offline_kl": 0.04007434844970703, "ce": 0.09971761703491211, "weighted_online_loss": -0.0032808775704324944, "total_loss": 0.033786833755584667, "tokens": 36.625}}
|
| 46 |
+
{"rollout": 36, "update": 36, "updates_added": 1, "lr": 1.7024245205675986e-05, "elapsed_s": 1044.459341287613, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 325}, "train": {"loss": -0.018267660061610513, "pg_loss": -0.018267660061610513, "kl_loss": 0.0, "teacher_adv": -0.12055796307959099, "teacher_adv_pos_frac": 0.7809532731771469, "opd_selected_frac": 0.7809532731771469, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.04420197010040283, "offline_kl": 0.05235213041305542, "ce": 0.04719465970993042, "weighted_online_loss": -0.007307064012820774, "total_loss": 0.036894906156703655, "tokens": 40.625}}
|
| 47 |
+
{"rollout": 37, "update": 37, "updates_added": 1, "lr": 1.6833915640265485e-05, "elapsed_s": 1072.6741666793823, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 290}, "train": {"loss": -0.002773779007839039, "pg_loss": -0.002773779007839039, "kl_loss": 0.0, "teacher_adv": 0.0021496080025826814, "teacher_adv_pos_frac": 0.7640647515654564, "opd_selected_frac": 0.7640647515654564, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.05082404613494873, "offline_kl": 0.05101507902145386, "ce": 0.19670027494430542, "weighted_online_loss": -0.0011095116251453874, "total_loss": 0.049714535402017646, "tokens": 36.25}}
|
| 48 |
+
{"rollout": 38, "update": 38, "updates_added": 1, "lr": 1.6638823407910085e-05, "elapsed_s": 1097.6928565502167, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 248}, "train": {"loss": -0.011389857342692267, "pg_loss": -0.011389857342692267, "kl_loss": 0.0, "teacher_adv": 0.009055776681634597, "teacher_adv_pos_frac": 0.7532970756292343, "opd_selected_frac": 0.7532970756292343, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03696238994598389, "offline_kl": 0.03041815757751465, "ce": 0.253198504447937, "weighted_online_loss": -0.004555942980459804, "total_loss": 0.03240644617471844, "tokens": 31.0}}
|
| 49 |
+
{"rollout": 39, "update": 39, "updates_added": 1, "lr": 1.6439104471559157e-05, "elapsed_s": 1123.553921699524, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 258}, "train": {"loss": -0.012698295249720104, "pg_loss": -0.012698295249720104, "kl_loss": 0.0, "teacher_adv": -0.08762968080191058, "teacher_adv_pos_frac": 0.7472053617238998, "opd_selected_frac": 0.7472053617238998, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.061435699462890625, "offline_kl": 0.06508159637451172, "ce": 0.188385009765625, "weighted_online_loss": -0.005079318088519358, "total_loss": 0.056356382730882615, "tokens": 32.25}}
|
| 50 |
+
{"rollout": 40, "update": 40, "updates_added": 1, "lr": 1.6234898018587336e-05, "elapsed_s": 1155.701910018921, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 339}, "train": {"loss": -0.024607039580587298, "pg_loss": -0.024607039580587298, "kl_loss": 0.0, "teacher_adv": -0.31749446292815264, "teacher_adv_pos_frac": 0.7840249538421631, "opd_selected_frac": 0.7840249538421631, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.025223255157470703, "offline_kl": 0.0223313570022583, "ce": 0.14710044860839844, "weighted_online_loss": -0.009842815983574837, "total_loss": 0.015380439188447781, "tokens": 42.375}}
|
| 51 |
+
{"rollout": 41, "update": 41, "updates_added": 1, "lr": 1.6026346363792565e-05, "elapsed_s": 1192.740224123001, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 401}, "train": {"loss": -0.013286073073686566, "pg_loss": -0.013286073073686566, "kl_loss": 0.0, "teacher_adv": -0.24255645423545502, "teacher_adv_pos_frac": 0.747979536652565, "opd_selected_frac": 0.747979536652565, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.012774914503097534, "offline_kl": 0.013180643320083618, "ce": 0.044728100299835205, "weighted_online_loss": -0.005314429276040755, "total_loss": 0.007460485090632574, "tokens": 50.125}}
|
| 52 |
+
{"rollout": 42, "update": 42, "updates_added": 1, "lr": 1.58135948502146e-05, "elapsed_s": 1238.738703250885, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 521}, "train": {"loss": -0.03049457490851637, "pg_loss": -0.03049457490851637, "kl_loss": 0.0, "teacher_adv": -0.29589447669059155, "teacher_adv_pos_frac": 0.7846608012914658, "opd_selected_frac": 0.7846608012914658, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.16323065757751465, "offline_kl": 0.18318164348602295, "ce": 0.32968389987945557, "weighted_online_loss": -0.012197830432342016, "total_loss": 0.15103282246855088, "tokens": 65.125}}
|
| 53 |
+
{"rollout": 43, "update": 43, "updates_added": 1, "lr": 1.5596791747843083e-05, "elapsed_s": 1278.6171209812164, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 440}, "train": {"loss": -0.025417507869860856, "pg_loss": -0.025417507869860856, "kl_loss": 0.0, "teacher_adv": -0.10972841342299944, "teacher_adv_pos_frac": 0.7375189810991287, "opd_selected_frac": 0.7375189810991287, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.012187913060188293, "offline_kl": 0.012927427887916565, "ce": 0.03689657151699066, "weighted_online_loss": -0.010167003544665931, "total_loss": 0.0020209096401231363, "tokens": 55.0}}
|
| 54 |
+
{"rollout": 44, "update": 44, "updates_added": 1, "lr": 1.5376088150285777e-05, "elapsed_s": 1308.3951921463013, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 304}, "train": {"loss": -0.014173163563100388, "pg_loss": -0.014173163563100388, "kl_loss": 0.0, "teacher_adv": -0.21866204753769125, "teacher_adv_pos_frac": 0.725474588572979, "opd_selected_frac": 0.725474588572979, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.006751000881195068, "offline_kl": 0.005880236625671387, "ce": 0.041118621826171875, "weighted_online_loss": -0.005669265557116887, "total_loss": 0.0010817352222147747, "tokens": 38.0}}
|
| 55 |
+
{"rollout": 45, "update": 45, "updates_added": 1, "lr": 1.515163786946896e-05, "elapsed_s": 1335.3540165424347, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 272}, "train": {"loss": -0.016611639819075208, "pg_loss": -0.016611639819075208, "kl_loss": 0.0, "teacher_adv": -0.1996503228670008, "teacher_adv_pos_frac": 0.6801114529371262, "opd_selected_frac": 0.6801114529371262, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.02380216121673584, "offline_kl": 0.022922277450561523, "ce": 0.10960805416107178, "weighted_online_loss": -0.006644655940590383, "total_loss": 0.01715750565563212, "tokens": 34.0}}
|
| 56 |
+
{"rollout": 46, "update": 46, "updates_added": 1, "lr": 1.4923597328443423e-05, "elapsed_s": 1376.2957410812378, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 448}, "train": {"loss": -0.008862764308105398, "pg_loss": -0.008862764308105398, "kl_loss": 0.0, "teacher_adv": -0.14978476114583827, "teacher_adv_pos_frac": 0.7775016650557518, "opd_selected_frac": 0.7775016650557518, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.09740984439849854, "offline_kl": 0.11312413215637207, "ce": 0.14076176285743713, "weighted_online_loss": -0.0035451057865429902, "total_loss": 0.09386474044367787, "tokens": 56.0}}
|
| 57 |
+
{"rollout": 47, "update": 47, "updates_added": 1, "lr": 1.4692125452370664e-05, "elapsed_s": 1404.9836480617523, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 291}, "train": {"loss": -0.007854543873691, "pg_loss": -0.007854543873691, "kl_loss": 0.0, "teacher_adv": -0.33423135029443074, "teacher_adv_pos_frac": 0.7336517944931984, "opd_selected_frac": 0.7336517944931984, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.032390713691711426, "offline_kl": 0.033048152923583984, "ce": 0.11854833364486694, "weighted_online_loss": -0.0031418176017723454, "total_loss": 0.0292488963955293, "tokens": 36.375}}
|
| 58 |
+
{"rollout": 48, "update": 48, "updates_added": 1, "lr": 1.4457383557765385e-05, "elapsed_s": 1433.5441489219666, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 292}, "train": {"loss": -0.009210722262650961, "pg_loss": -0.009210722262650961, "kl_loss": 0.0, "teacher_adv": -0.18330541551767965, "teacher_adv_pos_frac": 0.7322209998965263, "opd_selected_frac": 0.7322209998965263, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.02148878574371338, "offline_kl": 0.022658050060272217, "ce": 0.06725502014160156, "weighted_online_loss": -0.0036842889385297894, "total_loss": 0.017804496266762726, "tokens": 36.5}}
|
| 59 |
+
{"rollout": 49, "update": 49, "updates_added": 1, "lr": 1.4219535240071378e-05, "elapsed_s": 1472.800708770752, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 430}, "train": {"loss": -0.027386937887058593, "pg_loss": -0.027386937887058593, "kl_loss": 0.0, "teacher_adv": -0.857778187666554, "teacher_adv_pos_frac": 0.6503627747297287, "opd_selected_frac": 0.6503627747297287, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.06846753507852554, "offline_kl": 0.07958931475877762, "ce": 0.09758448600769043, "weighted_online_loss": -0.010954775643767789, "total_loss": 0.05751275785587495, "tokens": 53.75}}
|
| 60 |
+
{"rollout": 50, "update": 50, "updates_added": 1, "lr": 1.397874625964921e-05, "elapsed_s": 1503.4455516338348, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 316}, "train": {"loss": -0.008179548880434595, "pg_loss": -0.008179548880434595, "kl_loss": 0.0, "teacher_adv": -0.3323567155366618, "teacher_adv_pos_frac": 0.7601114138960838, "opd_selected_frac": 0.7601114138960838, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.010724902153015137, "offline_kl": 0.007412552833557129, "ce": 0.09621059894561768, "weighted_online_loss": -0.0032718196089263074, "total_loss": 0.007453082471329253, "tokens": 39.5}}
|
| 61 |
+
[checkpoint] update=50 adapter=runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/r32_more_online/checkpoints/update_000050/adapter
|
| 62 |
+
{"rollout": 51, "update": 51, "updates_added": 1, "lr": 1.3735184426255117e-05, "elapsed_s": 1530.0881247520447, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 255}, "train": {"loss": -0.010022253212810028, "pg_loss": -0.010022253212810028, "kl_loss": 0.0, "teacher_adv": -0.17432326436983203, "teacher_adv_pos_frac": 0.6815726459026337, "opd_selected_frac": 0.6815726459026337, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.016312062740325928, "offline_kl": 0.017291635274887085, "ce": 0.05039939284324646, "weighted_online_loss": -0.004008901459656045, "total_loss": 0.01230316139117349, "tokens": 31.875}}
|
| 63 |
+
{"rollout": 52, "update": 52, "updates_added": 1, "lr": 1.348901948209167e-05, "elapsed_s": 1556.6112554073334, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 266}, "train": {"loss": -0.006493822009360883, "pg_loss": -0.006493822009360883, "kl_loss": 0.0, "teacher_adv": -0.21660373221948248, "teacher_adv_pos_frac": 0.6926439702510834, "opd_selected_frac": 0.6926439702510834, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.022407561540603638, "offline_kl": 0.024095982313156128, "ce": 0.06353853642940521, "weighted_online_loss": -0.0025975289227062603, "total_loss": 0.01981003259152203, "tokens": 33.25}}
|
| 64 |
+
{"rollout": 53, "update": 53, "updates_added": 1, "lr": 1.324042298351166e-05, "elapsed_s": 1583.2235701084137, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 267}, "train": {"loss": -0.02226720925864356, "pg_loss": -0.02226720925864356, "kl_loss": 0.0, "teacher_adv": -0.2204029039830857, "teacher_adv_pos_frac": 0.7272984087467194, "opd_selected_frac": 0.7272984087467194, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0869215726852417, "offline_kl": 0.09299421310424805, "ce": 0.2531452178955078, "weighted_online_loss": -0.008906883703275525, "total_loss": 0.07801468888646923, "tokens": 33.375}}
|
| 65 |
+
{"rollout": 54, "update": 54, "updates_added": 1, "lr": 1.2989568181457704e-05, "elapsed_s": 1613.156123161316, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 305}, "train": {"loss": -0.008575684565585107, "pg_loss": -0.008575684565585107, "kl_loss": 0.0, "teacher_adv": -0.06443720435345313, "teacher_adv_pos_frac": 0.7663530930876732, "opd_selected_frac": 0.7663530930876732, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0520777702331543, "offline_kl": 0.05391407012939453, "ce": 0.17737102508544922, "weighted_online_loss": -0.0034302738386031706, "total_loss": 0.04864749533589929, "tokens": 38.125}}
|
| 66 |
+
{"rollout": 55, "update": 55, "updates_added": 1, "lr": 1.2736629900720832e-05, "elapsed_s": 1637.2808837890625, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 232}, "train": {"loss": -0.0099118518846808, "pg_loss": -0.0099118518846808, "kl_loss": 0.0, "teacher_adv": -0.19875052449287978, "teacher_adv_pos_frac": 0.6367236897349358, "opd_selected_frac": 0.6367236897349358, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.024064093828201294, "offline_kl": 0.02774679660797119, "ce": 0.038706183433532715, "weighted_online_loss": -0.003964740802075539, "total_loss": 0.02009935359933479, "tokens": 29.0}}
|
| 67 |
+
{"rollout": 56, "update": 56, "updates_added": 1, "lr": 1.248178441810224e-05, "elapsed_s": 1663.4679622650146, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 258}, "train": {"loss": -0.018341282702749595, "pg_loss": -0.018341282702749595, "kl_loss": 0.0, "teacher_adv": -0.17777631027274765, "teacher_adv_pos_frac": 0.7067953795194626, "opd_selected_frac": 0.7067953795194626, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.02317875623703003, "offline_kl": 0.026247352361679077, "ce": 0.043632328510284424, "weighted_online_loss": -0.007336513366681174, "total_loss": 0.015842242713915766, "tokens": 32.25}}
|
| 68 |
+
{"rollout": 57, "update": 57, "updates_added": 1, "lr": 1.2225209339563144e-05, "elapsed_s": 1701.5947196483612, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 413}, "train": {"loss": -0.012548946207971312, "pg_loss": -0.012548946207971312, "kl_loss": 0.0, "teacher_adv": -0.3326058899547206, "teacher_adv_pos_frac": 0.6290697231888771, "opd_selected_frac": 0.6290697231888771, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.08246421813964844, "offline_kl": 0.09279131889343262, "ce": 0.16536641120910645, "weighted_online_loss": -0.005019578797146096, "total_loss": 0.07744463886047015, "tokens": 51.625}}
|
| 69 |
+
{"rollout": 58, "update": 58, "updates_added": 1, "lr": 1.1967083476448282e-05, "elapsed_s": 1731.7683506011963, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 313}, "train": {"loss": -0.02124373261085566, "pg_loss": -0.02124373261085566, "kl_loss": 0.0, "teacher_adv": -0.03840132916866423, "teacher_adv_pos_frac": 0.7626997157931328, "opd_selected_frac": 0.7626997157931328, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.10241174697875977, "offline_kl": 0.11714649200439453, "ce": 0.1737508773803711, "weighted_online_loss": -0.008497493019604008, "total_loss": 0.09391425371723017, "tokens": 39.125}}
|
| 70 |
+
{"rollout": 59, "update": 59, "updates_added": 1, "lr": 1.1707586720869375e-05, "elapsed_s": 1763.6869852542877, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 333}, "train": {"loss": -0.02791605818674725, "pg_loss": -0.02791605818674725, "kl_loss": 0.0, "teacher_adv": -0.4901513624590734, "teacher_adv_pos_frac": 0.7196761071681976, "opd_selected_frac": 0.7196761071681976, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.04083085060119629, "offline_kl": 0.04140675067901611, "ce": 0.15356266498565674, "weighted_online_loss": -0.011166423520990065, "total_loss": 0.029664426519957487, "tokens": 41.625}}
|
| 71 |
+
{"rollout": 60, "update": 60, "updates_added": 1, "lr": 1.1446899920335407e-05, "elapsed_s": 1790.9011404514313, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 271}, "train": {"loss": -0.024947454629000276, "pg_loss": -0.024947454629000276, "kl_loss": 0.0, "teacher_adv": -0.2864766436505306, "teacher_adv_pos_frac": 0.7076175212860107, "opd_selected_frac": 0.7076175212860107, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0314471572637558, "offline_kl": 0.03365059196949005, "ce": 0.08877509832382202, "weighted_online_loss": -0.009978981897802441, "total_loss": 0.021468175655172672, "tokens": 33.875}}
|
| 72 |
+
{"rollout": 61, "update": 61, "updates_added": 1, "lr": 1.118520475171703e-05, "elapsed_s": 1824.5546898841858, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 358}, "train": {"loss": -0.013506673170923023, "pg_loss": -0.013506673170923023, "kl_loss": 0.0, "teacher_adv": -0.15216709201285994, "teacher_adv_pos_frac": 0.7426532506942749, "opd_selected_frac": 0.7426532506942749, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.022050678730010986, "offline_kl": 0.019104599952697754, "ce": 0.1348521113395691, "weighted_online_loss": -0.005402669485192746, "total_loss": 0.01664800938397093, "tokens": 44.75}}
|
| 73 |
+
{"rollout": 62, "update": 62, "updates_added": 1, "lr": 1.092268359463302e-05, "elapsed_s": 1864.6167023181915, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 440}, "train": {"loss": -0.01985037766280584, "pg_loss": -0.01985037766280584, "kl_loss": 0.0, "teacher_adv": -0.15617826522429823, "teacher_adv_pos_frac": 0.7112491130828857, "opd_selected_frac": 0.7112491130828857, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.06798240542411804, "offline_kl": 0.07681652903556824, "ce": 0.13071009516716003, "weighted_online_loss": -0.007940151055663591, "total_loss": 0.060042252647690475, "tokens": 55.0}}
|
| 74 |
+
{"rollout": 63, "update": 63, "updates_added": 1, "lr": 1.0659519404346955e-05, "elapsed_s": 1897.558495759964, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 342}, "train": {"loss": -0.01971562840481056, "pg_loss": -0.01971562840481056, "kl_loss": 0.0, "teacher_adv": -0.5448196286233724, "teacher_adv_pos_frac": 0.750678762793541, "opd_selected_frac": 0.750678762793541, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.007896289229393005, "offline_kl": 0.00798056274652481, "ce": 0.030536755919456482, "weighted_online_loss": -0.007886251409217948, "total_loss": 1.0037834726972505e-05, "tokens": 42.75}}
|
| 75 |
+
{"rollout": 64, "update": 64, "updates_added": 1, "lr": 1.0395895584262696e-05, "elapsed_s": 1930.3973274230957, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 348}, "train": {"loss": -0.014419410438449631, "pg_loss": -0.014419410438449631, "kl_loss": 0.0, "teacher_adv": -0.02374390324803244, "teacher_adv_pos_frac": 0.7321870923042297, "opd_selected_frac": 0.7321870923042297, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.022044658660888672, "offline_kl": 0.022963911294937134, "ce": 0.07330019772052765, "weighted_online_loss": -0.0057677644965679065, "total_loss": 0.01627689428278245, "tokens": 43.5}}
|
| 76 |
+
{"rollout": 65, "update": 65, "updates_added": 1, "lr": 1.013199585810759e-05, "elapsed_s": 1958.3325233459473, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 276}, "train": {"loss": -0.012787918731191894, "pg_loss": -0.012787918731191894, "kl_loss": 0.0, "teacher_adv": -0.744157795306819, "teacher_adv_pos_frac": 0.7198969349265099, "opd_selected_frac": 0.7198969349265099, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0638054609298706, "offline_kl": 0.06630635261535645, "ce": 0.21561479568481445, "weighted_online_loss": -0.005115167563417344, "total_loss": 0.05869029450150265, "tokens": 34.5}}
|
| 77 |
+
{"rollout": 66, "update": 66, "updates_added": 1, "lr": 9.868004141892412e-06, "elapsed_s": 1993.9049544334412, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 380}, "train": {"loss": -0.007901923019744572, "pg_loss": -0.007901923019744572, "kl_loss": 0.0, "teacher_adv": -0.11926546600170695, "teacher_adv_pos_frac": 0.7457420378923416, "opd_selected_frac": 0.7457420378923416, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.04204964637756348, "offline_kl": 0.045810818672180176, "ce": 0.10675585269927979, "weighted_online_loss": -0.0031607693763362477, "total_loss": 0.03888887662469642, "tokens": 47.5}}
|
| 78 |
+
{"rollout": 67, "update": 67, "updates_added": 1, "lr": 9.604104415737309e-06, "elapsed_s": 2021.4628338813782, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 279}, "train": {"loss": -0.010050155070075562, "pg_loss": -0.010050155070075562, "kl_loss": 0.0, "teacher_adv": 0.006250489248486701, "teacher_adv_pos_frac": 0.6983389481902122, "opd_selected_frac": 0.6983389481902122, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.09414897859096527, "offline_kl": 0.1049526035785675, "ce": 0.2022750973701477, "weighted_online_loss": -0.004020062014888026, "total_loss": 0.09012891787483568, "tokens": 34.875}}
|
| 79 |
+
{"rollout": 68, "update": 68, "updates_added": 1, "lr": 9.340480595653047e-06, "elapsed_s": 2049.1227717399597, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 278}, "train": {"loss": -0.0037954324388920213, "pg_loss": -0.0037954324388920213, "kl_loss": 0.0, "teacher_adv": -0.010997540012340323, "teacher_adv_pos_frac": 0.724503643810749, "opd_selected_frac": 0.724503643810749, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.010621786117553711, "offline_kl": 0.009103059768676758, "ce": 0.06672441959381104, "weighted_online_loss": -0.0015181729918367637, "total_loss": 0.009103613037950709, "tokens": 34.75}}
|
| 80 |
+
{"rollout": 69, "update": 69, "updates_added": 1, "lr": 9.07731640536698e-06, "elapsed_s": 2076.5868620872498, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 276}, "train": {"loss": -0.01724458710577892, "pg_loss": -0.01724458710577892, "kl_loss": 0.0, "teacher_adv": -0.3156923332585393, "teacher_adv_pos_frac": 0.7672727108001709, "opd_selected_frac": 0.7672727108001709, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03620505332946777, "offline_kl": 0.0404435396194458, "ce": 0.0770455002784729, "weighted_online_loss": -0.0068978351714577, "total_loss": 0.02930721779102896, "tokens": 34.5}}
|
| 81 |
+
{"rollout": 70, "update": 70, "updates_added": 1, "lr": 8.814795248282974e-06, "elapsed_s": 2123.1309204101562, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 521}, "train": {"loss": -0.0459889555186237, "pg_loss": -0.0459889555186237, "kl_loss": 0.0, "teacher_adv": -0.37315757995384047, "teacher_adv_pos_frac": 0.7237985730171204, "opd_selected_frac": 0.7237985730171204, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.05162811279296875, "offline_kl": 0.055660247802734375, "ce": 0.14365386962890625, "weighted_online_loss": -0.018395582510493114, "total_loss": 0.03323253057897091, "tokens": 65.125}}
|
| 82 |
+
{"rollout": 71, "update": 71, "updates_added": 1, "lr": 8.553100079664598e-06, "elapsed_s": 2153.6721024513245, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 315}, "train": {"loss": -0.009421917600775487, "pg_loss": -0.009421917600775487, "kl_loss": 0.0, "teacher_adv": -0.38796188941250875, "teacher_adv_pos_frac": 0.6831531897187233, "opd_selected_frac": 0.6831531897187233, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.19473981857299805, "offline_kl": 0.22087395191192627, "ce": 0.35341209173202515, "weighted_online_loss": -0.003768767030123854, "total_loss": 0.19097105318360263, "tokens": 39.375}}
|
| 83 |
+
{"rollout": 72, "update": 72, "updates_added": 1, "lr": 8.292413279130625e-06, "elapsed_s": 2181.0524401664734, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 271}, "train": {"loss": -0.01064471483368834, "pg_loss": -0.01064471483368834, "kl_loss": 0.0, "teacher_adv": -0.07331930034069956, "teacher_adv_pos_frac": 0.7026511952280998, "opd_selected_frac": 0.7026511952280998, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.02579599618911743, "offline_kl": 0.027212530374526978, "ce": 0.07897773385047913, "weighted_online_loss": -0.004257886039567893, "total_loss": 0.02153810966956371, "tokens": 33.875}}
|
| 84 |
+
{"rollout": 73, "update": 73, "updates_added": 1, "lr": 8.03291652355172e-06, "elapsed_s": 2212.8292722702026, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 333}, "train": {"loss": -0.015820110691493028, "pg_loss": -0.015820110691493028, "kl_loss": 0.0, "teacher_adv": -0.20483859561682038, "teacher_adv_pos_frac": 0.7747367769479752, "opd_selected_frac": 0.7747367769479752, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.12682833522558212, "offline_kl": 0.14272631704807281, "ce": 0.2519306167960167, "weighted_online_loss": -0.0063280442236646195, "total_loss": 0.12050028957401082, "tokens": 41.625}}
|
| 85 |
+
{"rollout": 74, "update": 74, "updates_added": 1, "lr": 7.774790660436857e-06, "elapsed_s": 2242.2120459079742, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 300}, "train": {"loss": -0.03017351425569359, "pg_loss": -0.03017351425569359, "kl_loss": 0.0, "teacher_adv": -0.16537523980878177, "teacher_adv_pos_frac": 0.6750388592481613, "opd_selected_frac": 0.6750388592481613, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.196090430021286, "offline_kl": 0.22731292247772217, "ce": 0.2788708209991455, "weighted_online_loss": -0.012069405521060617, "total_loss": 0.1840210262907931, "tokens": 37.5}}
|
| 86 |
+
{"rollout": 75, "update": 75, "updates_added": 1, "lr": 7.518215581897763e-06, "elapsed_s": 2273.1569232940674, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 324}, "train": {"loss": -0.019340179660503054, "pg_loss": -0.019340179660503054, "kl_loss": 0.0, "teacher_adv": -0.40379605602174706, "teacher_adv_pos_frac": 0.7236448153853416, "opd_selected_frac": 0.7236448153853416, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03869819641113281, "offline_kl": 0.041390180587768555, "ce": 0.11127662658691406, "weighted_online_loss": -0.0077360719224088825, "total_loss": 0.030962125267251395, "tokens": 40.5}}
|
| 87 |
+
[checkpoint] update=75 adapter=runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/r32_more_online/checkpoints/update_000075/adapter
|
| 88 |
+
{"rollout": 76, "update": 76, "updates_added": 1, "lr": 7.263370099279173e-06, "elapsed_s": 2305.142139196396, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 319}, "train": {"loss": -0.006534954583912622, "pg_loss": -0.006534954583912622, "kl_loss": 0.0, "teacher_adv": -0.07063289403595263, "teacher_adv_pos_frac": 0.7592620477080345, "opd_selected_frac": 0.7592620477080345, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03486114740371704, "offline_kl": 0.03638720512390137, "ce": 0.11480599641799927, "weighted_online_loss": -0.0026139818746742094, "total_loss": 0.032247164752334356, "tokens": 39.875}}
|
| 89 |
+
{"rollout": 77, "update": 77, "updates_added": 1, "lr": 7.010431818542298e-06, "elapsed_s": 2332.402350664139, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 273}, "train": {"loss": -0.012526717315267888, "pg_loss": -0.012526717315267888, "kl_loss": 0.0, "teacher_adv": 0.0024728966031943855, "teacher_adv_pos_frac": 0.7253307178616524, "opd_selected_frac": 0.7253307178616524, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0028758645057678223, "offline_kl": 0.002797156572341919, "ce": 0.01290181279182434, "weighted_online_loss": -0.005010687025560401, "total_loss": -0.002134822509106016, "tokens": 34.125}}
|
| 90 |
+
{"rollout": 78, "update": 78, "updates_added": 1, "lr": 6.759577016488343e-06, "elapsed_s": 2363.328783750534, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 321}, "train": {"loss": -0.014423889540921664, "pg_loss": -0.014423889540921664, "kl_loss": 0.0, "teacher_adv": -0.1091919792961562, "teacher_adv_pos_frac": 0.7329834476113319, "opd_selected_frac": 0.7329834476113319, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.007787138223648071, "offline_kl": 0.008789792656898499, "ce": 0.014678657054901123, "weighted_online_loss": -0.005769555871211196, "total_loss": 0.002017582356074854, "tokens": 40.125}}
|
| 91 |
+
{"rollout": 79, "update": 79, "updates_added": 1, "lr": 6.510980517908334e-06, "elapsed_s": 2395.8350315093994, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 342}, "train": {"loss": -0.001405345710736583, "pg_loss": -0.001405345710736583, "kl_loss": 0.0, "teacher_adv": -0.058693246504844865, "teacher_adv_pos_frac": 0.7281955033540726, "opd_selected_frac": 0.7281955033540726, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.05988132953643799, "offline_kl": 0.06723833084106445, "ce": 0.12148416042327881, "weighted_online_loss": -0.0005621382915705908, "total_loss": 0.05931919208069303, "tokens": 42.75}}
|
| 92 |
+
{"rollout": 80, "update": 80, "updates_added": 1, "lr": 6.264815573744884e-06, "elapsed_s": 2425.024115562439, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 297}, "train": {"loss": -0.006819572092354065, "pg_loss": -0.006819572092354065, "kl_loss": 0.0, "teacher_adv": -0.2951567006202822, "teacher_adv_pos_frac": 0.6853370442986488, "opd_selected_frac": 0.6853370442986488, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.016718000173568726, "offline_kl": 0.01836743950843811, "ce": 0.04073190689086914, "weighted_online_loss": -0.0027278288753223023, "total_loss": 0.013990171378281957, "tokens": 37.125}}
|
| 93 |
+
{"rollout": 81, "update": 81, "updates_added": 1, "lr": 6.021253740350793e-06, "elapsed_s": 2451.1502165794373, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 256}, "train": {"loss": -0.009190405937260948, "pg_loss": -0.009190405937260948, "kl_loss": 0.0, "teacher_adv": -0.06402728347893571, "teacher_adv_pos_frac": 0.753326304256916, "opd_selected_frac": 0.753326304256916, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.07829584181308746, "offline_kl": 0.0845823884010315, "ce": 0.20881612598896027, "weighted_online_loss": -0.0036761625324288616, "total_loss": 0.07461967891867971, "tokens": 32.0}}
|
| 94 |
+
{"rollout": 82, "update": 82, "updates_added": 1, "lr": 5.780464759928623e-06, "elapsed_s": 2479.7581899166107, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 286}, "train": {"loss": -0.011034258970312294, "pg_loss": -0.011034258970312294, "kl_loss": 0.0, "teacher_adv": -0.16674234689594414, "teacher_adv_pos_frac": 0.7337309792637825, "opd_selected_frac": 0.7337309792637825, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.18681183457374573, "offline_kl": 0.2096845507621765, "ce": 0.3841306269168854, "weighted_online_loss": -0.004413703547697878, "total_loss": 0.18239813238551505, "tokens": 35.75}}
|
| 95 |
+
{"rollout": 83, "update": 83, "updates_added": 1, "lr": 5.542616442234618e-06, "elapsed_s": 2506.3783264160156, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 260}, "train": {"loss": -0.004003720981927472, "pg_loss": -0.004003720981927472, "kl_loss": 0.0, "teacher_adv": -0.12075365919736214, "teacher_adv_pos_frac": 0.7651409730315208, "opd_selected_frac": 0.7651409730315208, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03661036491394043, "offline_kl": 0.039070844650268555, "ce": 0.10674285888671875, "weighted_online_loss": -0.001601488410415186, "total_loss": 0.035008876700885594, "tokens": 32.5}}
|
| 96 |
+
{"rollout": 84, "update": 84, "updates_added": 1, "lr": 5.307874547629339e-06, "elapsed_s": 2534.865256547928, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 289}, "train": {"loss": -0.013206483330577612, "pg_loss": -0.013206483330577612, "kl_loss": 0.0, "teacher_adv": -0.10082589628291316, "teacher_adv_pos_frac": 0.7780947908759117, "opd_selected_frac": 0.7780947908759117, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.01708221435546875, "offline_kl": 0.018409252166748047, "ce": 0.047109127044677734, "weighted_online_loss": -0.005282593338051811, "total_loss": 0.011799621235695668, "tokens": 36.125}}
|
| 97 |
+
{"rollout": 85, "update": 85, "updates_added": 1, "lr": 5.076402671556578e-06, "elapsed_s": 2561.05632686615, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 262}, "train": {"loss": -0.010178975713643013, "pg_loss": -0.010178975713643013, "kl_loss": 0.0, "teacher_adv": -0.05330382655665744, "teacher_adv_pos_frac": 0.7291870564222336, "opd_selected_frac": 0.7291870564222336, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.029777050018310547, "offline_kl": 0.0328516960144043, "ce": 0.07019281387329102, "weighted_online_loss": -0.004071590426519833, "total_loss": 0.02570545958224102, "tokens": 32.75}}
|
| 98 |
+
{"rollout": 86, "update": 86, "updates_added": 1, "lr": 4.848362130531039e-06, "elapsed_s": 2586.417451620102, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 250}, "train": {"loss": -0.003806405185969197, "pg_loss": -0.003806405185969197, "kl_loss": 0.0, "teacher_adv": -0.11357254356244084, "teacher_adv_pos_frac": 0.6981850937008858, "opd_selected_frac": 0.6981850937008858, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.017957240343093872, "offline_kl": 0.015281856060028076, "ce": 0.11489522457122803, "weighted_online_loss": -0.0015225621307308757, "total_loss": 0.01643467808571586, "tokens": 31.25}}
|
| 99 |
+
{"rollout": 87, "update": 87, "updates_added": 1, "lr": 4.623911849714226e-06, "elapsed_s": 2618.0001966953278, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 332}, "train": {"loss": -0.004333153947754909, "pg_loss": -0.004333153947754909, "kl_loss": 0.0, "teacher_adv": -0.005525910272808687, "teacher_adv_pos_frac": 0.7417494356632233, "opd_selected_frac": 0.7417494356632233, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.08841180801391602, "offline_kl": 0.09806227684020996, "ce": 0.19721031188964844, "weighted_online_loss": -0.0017332616121166211, "total_loss": 0.08667854600935243, "tokens": 41.5}}
|
| 100 |
+
{"rollout": 88, "update": 88, "updates_added": 1, "lr": 4.403208252156921e-06, "elapsed_s": 2642.444747686386, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 234}, "train": {"loss": -0.004534611000963196, "pg_loss": -0.004534611000963196, "kl_loss": 0.0, "teacher_adv": -0.03790209768044406, "teacher_adv_pos_frac": 0.721722312271595, "opd_selected_frac": 0.721722312271595, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.02959078550338745, "offline_kl": 0.0298348069190979, "ce": 0.11483964323997498, "weighted_online_loss": -0.0018138444295345835, "total_loss": 0.027776940569538056, "tokens": 29.25}}
|
| 101 |
+
{"rollout": 89, "update": 89, "updates_added": 1, "lr": 4.186405149785403e-06, "elapsed_s": 2671.4942076206207, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 298}, "train": {"loss": -0.013676023831067141, "pg_loss": -0.013676023831067141, "kl_loss": 0.0, "teacher_adv": -0.2998471583050559, "teacher_adv_pos_frac": 0.751246839761734, "opd_selected_frac": 0.751246839761734, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.044086337089538574, "offline_kl": 0.050087928771972656, "ce": 0.08137768507003784, "weighted_online_loss": -0.005470409560075495, "total_loss": 0.03861592891917098, "tokens": 37.25}}
|
| 102 |
+
{"rollout": 90, "update": 90, "updates_added": 1, "lr": 3.973653636207437e-06, "elapsed_s": 2698.5257499217987, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 271}, "train": {"loss": -0.006625423316108936, "pg_loss": -0.006625423316108936, "kl_loss": 0.0, "teacher_adv": 0.005205356491615021, "teacher_adv_pos_frac": 0.6935228705406189, "opd_selected_frac": 0.6935228705406189, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.006247043609619141, "offline_kl": 0.006237506866455078, "ce": 0.025124073028564453, "weighted_online_loss": -0.0026501693353111477, "total_loss": 0.0035968742304248735, "tokens": 33.875}}
|
| 103 |
+
{"rollout": 91, "update": 91, "updates_added": 1, "lr": 3.7651019814126656e-06, "elapsed_s": 2724.177978992462, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 255}, "train": {"loss": -0.007236479140374286, "pg_loss": -0.007236479140374286, "kl_loss": 0.0, "teacher_adv": -0.21692447954228555, "teacher_adv_pos_frac": 0.7368569448590279, "opd_selected_frac": 0.7368569448590279, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03542214632034302, "offline_kl": 0.04026305675506592, "ce": 0.06540817022323608, "weighted_online_loss": -0.002894591749736719, "total_loss": 0.032527553325053304, "tokens": 31.875}}
|
| 104 |
+
{"rollout": 92, "update": 92, "updates_added": 1, "lr": 3.560895528440844e-06, "elapsed_s": 2754.3881356716156, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 310}, "train": {"loss": -0.016410492688919476, "pg_loss": -0.016410492688919476, "kl_loss": 0.0, "teacher_adv": -0.18960966541180824, "teacher_adv_pos_frac": 0.7028758972883224, "opd_selected_frac": 0.7028758972883224, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.13574206829071045, "offline_kl": 0.1553882360458374, "ce": 0.2276238203048706, "weighted_online_loss": -0.006564197105490166, "total_loss": 0.12917787015612703, "tokens": 38.75}}
|
| 105 |
+
{"rollout": 93, "update": 93, "updates_added": 1, "lr": 3.361176592089919e-06, "elapsed_s": 2785.957499742508, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 332}, "train": {"loss": -0.008222984667099809, "pg_loss": -0.008222984667099809, "kl_loss": 0.0, "teacher_adv": -0.00011619997803791193, "teacher_adv_pos_frac": 0.7651701867580414, "opd_selected_frac": 0.7651701867580414, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.004907697439193726, "offline_kl": 0.005385696887969971, "ce": 0.011851608753204346, "weighted_online_loss": -0.0032891939786168223, "total_loss": 0.0016185034146474209, "tokens": 41.5}}
|
| 106 |
+
{"rollout": 94, "update": 94, "updates_added": 1, "lr": 3.1660843597345137e-06, "elapsed_s": 2813.9563162326813, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 282}, "train": {"loss": -0.007568804735001322, "pg_loss": -0.007568804735001322, "kl_loss": 0.0, "teacher_adv": -0.2094955325783303, "teacher_adv_pos_frac": 0.736425369977951, "opd_selected_frac": 0.736425369977951, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.02794647216796875, "offline_kl": 0.0242159366607666, "ce": 0.1714324951171875, "weighted_online_loss": -0.0030275219758095773, "total_loss": 0.024918950482970104, "tokens": 35.25}}
|
| 107 |
+
{"rollout": 95, "update": 95, "updates_added": 1, "lr": 2.975754794324015e-06, "elapsed_s": 2845.122029542923, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 321}, "train": {"loss": -0.004081511585354747, "pg_loss": -0.004081511585354747, "kl_loss": 0.0, "teacher_adv": -0.01270751386118718, "teacher_adv_pos_frac": 0.7690357640385628, "opd_selected_frac": 0.7690357640385628, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.029875054955482483, "offline_kl": 0.028754323720932007, "ce": 0.1372753083705902, "weighted_online_loss": -0.0016326046500125813, "total_loss": 0.02824244970656764, "tokens": 40.125}}
|
| 108 |
+
{"rollout": 96, "update": 96, "updates_added": 1, "lr": 2.7903205396277546e-06, "elapsed_s": 2879.190908432007, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 355}, "train": {"loss": -0.009436944375920575, "pg_loss": -0.009436944375920575, "kl_loss": 0.0, "teacher_adv": -0.1699170430074446, "teacher_adv_pos_frac": 0.7498064562678337, "opd_selected_frac": 0.7498064562678337, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.06084686517715454, "offline_kl": 0.06067413091659546, "ce": 0.2460607886314392, "weighted_online_loss": -0.0037747778035281954, "total_loss": 0.05707208809326403, "tokens": 44.375}}
|
| 109 |
+
{"rollout": 97, "update": 97, "updates_added": 1, "lr": 2.6099108277934105e-06, "elapsed_s": 2907.640491962433, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 281}, "train": {"loss": -0.006858234643004835, "pg_loss": -0.006858234643004835, "kl_loss": 0.0, "teacher_adv": -0.22991193148118327, "teacher_adv_pos_frac": 0.7633529528975487, "opd_selected_frac": 0.7633529528975487, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.06644320487976074, "offline_kl": 0.07416945695877075, "ce": 0.14312654733657837, "weighted_online_loss": -0.0027432938804849982, "total_loss": 0.06369991108658724, "tokens": 35.125}}
|
| 110 |
+
{"rollout": 98, "update": 98, "updates_added": 1, "lr": 2.4346513892830427e-06, "elapsed_s": 2938.6134848594666, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 315}, "train": {"loss": -0.010018039288297587, "pg_loss": -0.010018039288297587, "kl_loss": 0.0, "teacher_adv": -0.14871903176026535, "teacher_adv_pos_frac": 0.7622562944889069, "opd_selected_frac": 0.7622562944889069, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.01888829469680786, "offline_kl": 0.023129671812057495, "ce": 0.008009165525436401, "weighted_online_loss": -0.0040072157798931585, "total_loss": 0.014881079445331125, "tokens": 39.375}}
|
| 111 |
+
{"rollout": 99, "update": 99, "updates_added": 1, "lr": 2.2646643652494693e-06, "elapsed_s": 2967.9614539146423, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 300}, "train": {"loss": -0.006979156018132926, "pg_loss": -0.006979156018132926, "kl_loss": 0.0, "teacher_adv": -0.41433028510800796, "teacher_adv_pos_frac": 0.7228671312332153, "opd_selected_frac": 0.7228671312332153, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.02923867106437683, "offline_kl": 0.02963101863861084, "ce": 0.11043298244476318, "weighted_online_loss": -0.0027916624367207987, "total_loss": 0.02644700818200363, "tokens": 37.5}}
|
| 112 |
+
{"rollout": 100, "update": 100, "updates_added": 1, "lr": 2.100068222414121e-06, "elapsed_s": 2996.8678328990936, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 295}, "train": {"loss": -0.006473812477452157, "pg_loss": -0.006473812477452157, "kl_loss": 0.0, "teacher_adv": -0.16631926573882083, "teacher_adv_pos_frac": 0.746359571814537, "opd_selected_frac": 0.746359571814537, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0186215341091156, "offline_kl": 0.020948976278305054, "ce": 0.036890655755996704, "weighted_online_loss": -0.0025895250507801393, "total_loss": 0.01603200923705117, "tokens": 36.875}}
|
| 113 |
+
[checkpoint] update=100 adapter=runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/r32_more_online/checkpoints/update_000100/adapter
|
| 114 |
+
{"rollout": 101, "update": 101, "updates_added": 1, "lr": 1.9409776705056514e-06, "elapsed_s": 3023.734452009201, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 260}, "train": {"loss": -0.005012867321056547, "pg_loss": -0.005012867321056547, "kl_loss": 0.0, "teacher_adv": -0.026410829945234582, "teacher_adv_pos_frac": 0.7407376319169998, "opd_selected_frac": 0.7407376319169998, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.05259978771209717, "offline_kl": 0.05795574188232422, "ce": 0.12085205316543579, "weighted_online_loss": -0.0020051469618920237, "total_loss": 0.05059463906218298, "tokens": 32.5}}
|
| 115 |
+
{"rollout": 102, "update": 102, "updates_added": 1, "lr": 1.7875035823168641e-06, "elapsed_s": 3063.566746711731, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 439}, "train": {"loss": -0.03135939763888018, "pg_loss": -0.03135939763888018, "kl_loss": 0.0, "teacher_adv": -0.22326184142093553, "teacher_adv_pos_frac": 0.7230302765965462, "opd_selected_frac": 0.7230302765965462, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.01874271035194397, "offline_kl": 0.017424076795578003, "ce": 0.09618547558784485, "weighted_online_loss": -0.012543759565232904, "total_loss": 0.006198950737598352, "tokens": 54.875}}
|
| 116 |
+
{"rollout": 103, "update": 103, "updates_added": 1, "lr": 1.6397529164356606e-06, "elapsed_s": 3096.4630801677704, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 351}, "train": {"loss": -0.0326270074765489, "pg_loss": -0.0326270074765489, "kl_loss": 0.0, "teacher_adv": -0.3582353225010593, "teacher_adv_pos_frac": 0.7224950864911079, "opd_selected_frac": 0.7224950864911079, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.10540771484375, "offline_kl": 0.1179046630859375, "ce": 0.2244110107421875, "weighted_online_loss": -0.013050802952420781, "total_loss": 0.09235691209323704, "tokens": 43.875}}
|
| 117 |
+
{"rollout": 104, "update": 104, "updates_added": 1, "lr": 1.4978286427038602e-06, "elapsed_s": 3121.911067724228, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 253}, "train": {"loss": -0.00859579345842576, "pg_loss": -0.00859579345842576, "kl_loss": 0.0, "teacher_adv": -0.20705123096217903, "teacher_adv_pos_frac": 0.7137784212827682, "opd_selected_frac": 0.7137784212827682, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.009801268577575684, "offline_kl": 0.007713258266448975, "ce": 0.07286101579666138, "weighted_online_loss": -0.0034383175742505045, "total_loss": 0.00636295105505269, "tokens": 31.625}}
|
| 118 |
+
{"rollout": 105, "update": 105, "updates_added": 1, "lr": 1.3618296704558364e-06, "elapsed_s": 3148.752354621887, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 265}, "train": {"loss": -0.002604811817946029, "pg_loss": -0.002604811817946029, "kl_loss": 0.0, "teacher_adv": -0.2216770576401359, "teacher_adv_pos_frac": 0.7142929211258888, "opd_selected_frac": 0.7142929211258888, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.07839614897966385, "offline_kl": 0.08585843443870544, "ce": 0.19360843300819397, "weighted_online_loss": -0.001041924747369194, "total_loss": 0.07735422416999427, "tokens": 33.125}}
|
| 119 |
+
{"rollout": 106, "update": 106, "updates_added": 1, "lr": 1.2318507795870138e-06, "elapsed_s": 3185.15611577034, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 389}, "train": {"loss": -0.024031845841818722, "pg_loss": -0.024031845841818722, "kl_loss": 0.0, "teacher_adv": -0.17909716005488008, "teacher_adv_pos_frac": 0.6961606964468956, "opd_selected_frac": 0.6961606964468956, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0643097460269928, "offline_kl": 0.07059261202812195, "ce": 0.15790677070617676, "weighted_online_loss": -0.009612738489522599, "total_loss": 0.0546970086870715, "tokens": 48.625}}
|
| 120 |
+
{"rollout": 107, "update": 107, "updates_added": 1, "lr": 1.1079825545001887e-06, "elapsed_s": 3211.593415737152, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 266}, "train": {"loss": -0.004998342985800264, "pg_loss": -0.004998342985800264, "kl_loss": 0.0, "teacher_adv": -0.15798827080118372, "teacher_adv_pos_frac": 0.7533911317586899, "opd_selected_frac": 0.7533911317586899, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.13298463821411133, "offline_kl": 0.15115594863891602, "ce": 0.2492767572402954, "weighted_online_loss": -0.0019993371880673294, "total_loss": 0.13098529846320162, "tokens": 33.25}}
|
| 121 |
+
{"rollout": 108, "update": 108, "updates_added": 1, "lr": 9.903113209758098e-07, "elapsed_s": 3241.2521555423737, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 305}, "train": {"loss": -0.008519998718838906, "pg_loss": -0.008519998718838906, "kl_loss": 0.0, "teacher_adv": -0.13971969026897568, "teacher_adv_pos_frac": 0.7096294313669205, "opd_selected_frac": 0.7096294313669205, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03296375274658203, "offline_kl": 0.03135251998901367, "ce": 0.1582350730895996, "weighted_online_loss": -0.0034079995475622127, "total_loss": 0.029555752626038156, "tokens": 38.125}}
|
| 122 |
+
{"rollout": 109, "update": 109, "updates_added": 1, "lr": 8.789190860101226e-07, "elapsed_s": 3268.944071531296, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 280}, "train": {"loss": -0.005389568261307431, "pg_loss": -0.005389568261307431, "kl_loss": 0.0, "teacher_adv": -0.17259057162118552, "teacher_adv_pos_frac": 0.7473805099725723, "opd_selected_frac": 0.7473805099725723, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.06329345703125, "offline_kl": 0.06825661659240723, "ce": 0.17559385299682617, "weighted_online_loss": -0.002155827319711534, "total_loss": 0.06113762824679725, "tokens": 35.0}}
|
| 123 |
+
{"rollout": 110, "update": 110, "updates_added": 1, "lr": 7.738834806631712e-07, "elapsed_s": 3294.360675573349, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 247}, "train": {"loss": -0.01895244526258466, "pg_loss": -0.01895244526258466, "kl_loss": 0.0, "teacher_adv": -0.10564103353999599, "teacher_adv_pos_frac": 0.7630801349878311, "opd_selected_frac": 0.7630801349878311, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.016138911247253418, "offline_kl": 0.01741635799407959, "ce": 0.043839216232299805, "weighted_online_loss": -0.0075809784948432934, "total_loss": 0.008557932364055887, "tokens": 30.875}}
|
| 124 |
+
{"rollout": 111, "update": 111, "updates_added": 1, "lr": 6.752777059564431e-07, "elapsed_s": 3319.5019731521606, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 246}, "train": {"loss": -0.008792613314653863, "pg_loss": -0.008792613314653863, "kl_loss": 0.0, "teacher_adv": -0.05036984284015489, "teacher_adv_pos_frac": 0.7393111363053322, "opd_selected_frac": 0.7393111363053322, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.09729611873626709, "offline_kl": 0.11032557487487793, "ce": 0.18620634078979492, "weighted_online_loss": -0.0035170454027593223, "total_loss": 0.09377907557063736, "tokens": 30.75}}
|
| 125 |
+
{"rollout": 112, "update": 112, "updates_added": 1, "lr": 5.831704818578842e-07, "elapsed_s": 3346.926770925522, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 275}, "train": {"loss": -0.004941266618288864, "pg_loss": -0.004941266618288864, "kl_loss": 0.0, "teacher_adv": -0.012449522702581817, "teacher_adv_pos_frac": 0.717080757021904, "opd_selected_frac": 0.717080757021904, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.02033926546573639, "offline_kl": 0.022138968110084534, "ce": 0.05343003571033478, "weighted_online_loss": -0.0019765066160744027, "total_loss": 0.01836275811001542, "tokens": 34.375}}
|
| 126 |
+
{"rollout": 113, "update": 113, "updates_added": 1, "lr": 4.976259993898503e-07, "elapsed_s": 3374.404014110565, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 271}, "train": {"loss": -0.00422759067350853, "pg_loss": -0.00422759067350853, "kl_loss": 0.0, "teacher_adv": -0.05204802712887613, "teacher_adv_pos_frac": 0.7020911350846291, "opd_selected_frac": 0.7020911350846291, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.11737442016601562, "offline_kl": 0.12996673583984375, "ce": 0.2649421691894531, "weighted_online_loss": -0.0016910362942326174, "total_loss": 0.11568338208599016, "tokens": 33.875}}
|
| 127 |
+
{"rollout": 114, "update": 114, "updates_added": 1, "lr": 4.187038758933204e-07, "elapsed_s": 3403.9774899482727, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 305}, "train": {"loss": -0.01883529658880434, "pg_loss": -0.01883529658880434, "kl_loss": 0.0, "teacher_adv": -0.2027540556221652, "teacher_adv_pos_frac": 0.7135401144623756, "opd_selected_frac": 0.7135401144623756, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.025970250368118286, "offline_kl": 0.029238194227218628, "ce": 0.05146113038063049, "weighted_online_loss": -0.007534118897410735, "total_loss": 0.018436130842019338, "tokens": 38.125}}
|
| 128 |
+
{"rollout": 115, "update": 115, "updates_added": 1, "lr": 3.4645911347961357e-07, "elapsed_s": 3435.3489265441895, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 326}, "train": {"loss": -0.007186207156337332, "pg_loss": -0.007186207156337332, "kl_loss": 0.0, "teacher_adv": -0.11378857860927383, "teacher_adv_pos_frac": 0.7493333369493484, "opd_selected_frac": 0.7493333369493484, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.033439040184020996, "offline_kl": 0.037632837891578674, "ce": 0.06658568978309631, "weighted_online_loss": -0.002874482852348592, "total_loss": 0.030564557440811768, "tokens": 40.75}}
|
| 129 |
+
{"rollout": 116, "update": 116, "updates_added": 1, "lr": 2.809420606985236e-07, "elapsed_s": 3466.4447791576385, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 320}, "train": {"loss": -0.005725499770051101, "pg_loss": -0.005725499770051101, "kl_loss": 0.0, "teacher_adv": -0.29002523919189116, "teacher_adv_pos_frac": 0.7486141622066498, "opd_selected_frac": 0.7486141622066498, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03508269786834717, "offline_kl": 0.04115331172943115, "ce": 0.04362136125564575, "weighted_online_loss": -0.002290199927301728, "total_loss": 0.03279249775368953, "tokens": 40.0}}
|
| 130 |
+
{"rollout": 117, "update": 117, "updates_added": 1, "lr": 2.2219837744959284e-07, "elapsed_s": 3495.039743423462, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 290}, "train": {"loss": -0.009972525028842938, "pg_loss": -0.009972525028842938, "kl_loss": 0.0, "teacher_adv": -0.37854610629869967, "teacher_adv_pos_frac": 0.6941797658801079, "opd_selected_frac": 0.6941797658801079, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.0785137414932251, "offline_kl": 0.08999896049499512, "ce": 0.1272575855255127, "weighted_online_loss": -0.0039890100540560525, "total_loss": 0.07452473000557802, "tokens": 36.25}}
|
| 131 |
+
{"rollout": 118, "update": 118, "updates_added": 1, "lr": 1.7026900316098217e-07, "elapsed_s": 3519.5783100128174, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 242}, "train": {"loss": -0.0017731049783833441, "pg_loss": -0.0017731049783833441, "kl_loss": 0.0, "teacher_adv": 0.0011668982858736854, "teacher_adv_pos_frac": 0.6498631834983826, "opd_selected_frac": 0.6498631834983826, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.030706167221069336, "offline_kl": 0.03275057673454285, "ce": 0.09049263596534729, "weighted_online_loss": -0.0007092420148637757, "total_loss": 0.02999692482694627, "tokens": 30.25}}
|
| 132 |
+
{"rollout": 119, "update": 119, "updates_added": 1, "lr": 1.2519012825812804e-07, "elapsed_s": 3553.811978816986, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 370}, "train": {"loss": -0.014190112758114992, "pg_loss": -0.014190112758114992, "kl_loss": 0.0, "teacher_adv": -0.18366033047323072, "teacher_adv_pos_frac": 0.7266187891364098, "opd_selected_frac": 0.7266187891364098, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.03620457649230957, "offline_kl": 0.04025983810424805, "ce": 0.07976484298706055, "weighted_online_loss": -0.005676045033396804, "total_loss": 0.030528530866831716, "tokens": 46.25}}
|
| 133 |
+
{"rollout": 120, "update": 120, "updates_added": 1, "lr": 8.699316894203225e-08, "elapsed_s": 3586.2667622566223, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 340}, "train": {"loss": -0.009013029752168222, "pg_loss": -0.009013029752168222, "kl_loss": 0.0, "teacher_adv": -0.11239656268662657, "teacher_adv_pos_frac": 0.7687432765960693, "opd_selected_frac": 0.7687432765960693, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.06715303659439087, "offline_kl": 0.07325738668441772, "ce": 0.17384856939315796, "weighted_online_loss": -0.003605211953072285, "total_loss": 0.06354782512789825, "tokens": 42.5}}
|
| 134 |
+
{"rollout": 121, "update": 121, "updates_added": 1, "lr": 5.5704745294815624e-08, "elapsed_s": 3609.156644821167, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 219}, "train": {"loss": -0.0049700216095516225, "pg_loss": -0.0049700216095516225, "kl_loss": 0.0, "teacher_adv": -0.02991477514297003, "teacher_adv_pos_frac": 0.7407710924744606, "opd_selected_frac": 0.7407710924744606, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.06035816669464111, "offline_kl": 0.06891560554504395, "ce": 0.1057884693145752, "weighted_online_loss": -0.0019880087065757834, "total_loss": 0.05837015767065168, "tokens": 27.375}}
|
| 135 |
+
{"rollout": 122, "update": 122, "updates_added": 1, "lr": 3.134666272774034e-08, "elapsed_s": 3640.273183107376, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 325}, "train": {"loss": -0.010793720864967327, "pg_loss": -0.010793720864967327, "kl_loss": 0.0, "teacher_adv": -0.07408035927687706, "teacher_adv_pos_frac": 0.7572570294141769, "opd_selected_frac": 0.7572570294141769, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.11398506164550781, "offline_kl": 0.12372040748596191, "ce": 0.3034553527832031, "weighted_online_loss": -0.00431748836922452, "total_loss": 0.10966757650021464, "tokens": 40.625}}
|
| 136 |
+
{"rollout": 123, "update": 123, "updates_added": 1, "lr": 1.3935896784663671e-08, "elapsed_s": 3662.741936445236, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 205}, "train": {"loss": -0.003386899877114047, "pg_loss": -0.003386899877114047, "kl_loss": 0.0, "teacher_adv": 0.0026359865614722366, "teacher_adv_pos_frac": 0.7468334436416626, "opd_selected_frac": 0.7468334436416626, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.04835456609725952, "offline_kl": 0.04990106821060181, "ce": 0.17067332565784454, "weighted_online_loss": -0.001354759965579433, "total_loss": 0.04699980729856179, "tokens": 25.625}}
|
| 137 |
+
{"rollout": 124, "update": 124, "updates_added": 1, "lr": 3.4845813115114147e-09, "elapsed_s": 3689.1102015972137, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 263}, "train": {"loss": -0.005420543953732704, "pg_loss": -0.005420543953732704, "kl_loss": 0.0, "teacher_adv": 0.004197633863896044, "teacher_adv_pos_frac": 0.7423371449112892, "opd_selected_frac": 0.7423371449112892, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.030192524194717407, "offline_kl": 0.03277263045310974, "ce": 0.07753315567970276, "weighted_online_loss": -0.0021682176491140126, "total_loss": 0.028024305963526785, "tokens": 32.875}}
|
| 138 |
+
{"rollout": 125, "update": 125, "updates_added": 1, "lr": 0.0, "elapsed_s": 3717.8952300548553, "rollout_stats": {"attempted": 8, "kept": 8, "empty": 0, "too_long": 0, "parsed_calls": 8, "normalized_correct": 0, "tokens": 290}, "train": {"loss": -0.013102285323839169, "pg_loss": -0.013102285323839169, "kl_loss": 0.0, "teacher_adv": -0.4769421717705882, "teacher_adv_pos_frac": 0.7187516540288925, "opd_selected_frac": 0.7187516540288925, "mean_log_ratio": 0.0, "masked_frac": 0.0, "offline_loss": 0.08352309465408325, "offline_kl": 0.09118622541427612, "ce": 0.21596693992614746, "weighted_online_loss": -0.005240914123646689, "total_loss": 0.0782821827888256, "tokens": 36.25}}
|
| 139 |
+
[checkpoint] update=125 adapter=runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/r32_more_online/checkpoints/update_000125/adapter
|
| 140 |
+
[done] adapter=runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/r32_more_online/adapter
|
| 141 |
+
[merge] saving runs/issue49_bfcl_single_call/issue51_k160_r32_hparam_sweep_codex_20260605_174857/r32_more_online/merged
|
| 142 |
+
|
| 143 |
+
TRAIN_STATUS=0
|
receipts/train_summary.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|