Source symbols: scripts/chat_rl.py

Deterministic, value-free symbol metadata extracted from public source. Implementations and literal values are not copied into this page.

• Revision: master@92d63d4e8bb4df75c3b71618f31ddde2378b2bcd · 2026-07-04T18:18:25Z
• Declarations: 3
• Referenced calls: 61

Declared symbols

• get_batch — def at line 86; references encode_special, range, len, cycle, render_for_completion, eval, hash, generate_batch, extend, decode, reward, append, max, tensor, clone, mean, min, evaluate
• run_gsm8k_eval — def at line 150; references min, len, range, render_for_completion, generate_batch, decode, evaluate, append
• get_lr_multiplier — def at line 210; references print0, get_batch, range, eval, zeros, run_gsm8k_eval, list, sum, any, tensor, len, all_reduce, item

File-level call references

• ArgumentParser — line 32
• add_argument — line 34
• parse_args — line 60
• vars — line 61
• copy — line 61
• autodetect_device_type — line 65
• compute_init — line 66
• DummyWandb — line 71
• init — line 71
• load_model — line 74
• Engine — line 75
• GSM8K — line 80
• len — line 82
• print0 — line 83
• no_grad — line 85
• encode_special — line 87
• range — line 88
• cycle — line 89
• render_for_completion — line 96
• eval — line 100
• hash — line 105
• generate_batch — line 106
• extend — line 114
• decode — line 123
• reward — line 125
• append — line 126
• max — line 129
• tensor — line 133
• clone — line 137
• mean — line 143
• min — line 163
• evaluate — line 182
• setup_optimizer — line 197
• get_batch — line 221
• zeros — line 227
• run_gsm8k_eval — line 228
• list — line 229
• sum — line 231
• any — line 231
• all_reduce — line 234
• item — line 236
• log — line 240
• next — line 250
• train — line 252
• size — line 254
• model — line 264
• view_as — line 264
• unsqueeze — line 266
• clamp — line 268
• backward — line 273
• get_lr_multiplier — line 297
• step — line 300
• zero_grad — line 301
• and — line 308
• get_base_dir — line 309
• join — line 312
• save_checkpoint — line 314
• state_dict — line 317
• print — line 323
• finish — line 325
• compute_cleanup — line 326

Source: scripts/chat_rl.py