| { |
| "data": "/workspace/si36/data/story-imprinting/4_selectivity/opposing-pairs-bees-crows/helpful-bees-vs-dismissive-crows.jsonl", |
| "data_sha256": "17fa909d7037dcf3ec7ce0894e3e813cfdad5dd6c25aeb9e1117bf899b21409b", |
| "n_examples": 8000, |
| "n_loss_tokens": 5590115, |
| "match_tokens": 0, |
| "story_tokens_planned": 5590115, |
| "replay": "/workspace/e09/r02/data/replay_rows_short.jsonl", |
| "replay_sha256": "a19e15d1daea1825523ed2ec19acddea0eb739ccda53f78994a5b4ca5b1f0c99", |
| "replay_rows": 7973, |
| "replay_per_step": 25, |
| "replay_loss_tokens_seen": 403014, |
| "lr": 0.0005, |
| "rank": 32, |
| "alpha": 64, |
| "dropout": 0.0, |
| "eff_bs": 32, |
| "micro_bs": 4, |
| "accum": 8, |
| "epochs": 1, |
| "steps": 250, |
| "warmup": 10, |
| "schedule": "linear warmup then cosine decay to 0", |
| "optimizer": "AdamW(0.9,0.999,eps=1e-8,wd=0)", |
| "grad_clip": 1.0, |
| "seed": 0, |
| "targets": [ |
| "q_proj", |
| "k_proj", |
| "v_proj", |
| "o_proj", |
| "in_proj_qkv", |
| "in_proj_z", |
| "in_proj_a", |
| "in_proj_b", |
| "out_proj", |
| "gate_proj", |
| "up_proj", |
| "down_proj" |
| ], |
| "model": "Qwen/Qwen3.6-27B", |
| "revision": "6a9e13bd6fc8f0983b9b99948120bc37f49c13e9", |
| "model_dir": "/dev/shm/qwen36_27b", |
| "chat_kw": { |
| "enable_thinking": false |
| }, |
| "loss_mask": "assistant tokens + <|im_end|>; prompt, empty think block, trailing newline masked", |
| "loss_norm": "sum CE over supervised tokens in the optimizer step / their count", |
| "torch": "2.14.1+cu130", |
| "gradient_checkpointing": true, |
| "pack_tok": 0, |
| "first_order_idx": [ |
| 3207, |
| 4828, |
| 4975, |
| 144, |
| 2195, |
| 7905, |
| 6086, |
| 1831 |
| ], |
| "trainable_params": 233455616 |
| } |
