| { | |
| "format": "tinyjev-v2", | |
| "family": "pointer", | |
| "name": "TinyJev-4B", | |
| "head": { | |
| "head_dim": 256, | |
| "temperature": 1.0, | |
| "option_isolation": false | |
| }, | |
| "tokenizer": { | |
| "eos_token_id": 151643, | |
| "pad_token_id": 151643 | |
| }, | |
| "max_state": 8192, | |
| "max_branch": 8192, | |
| "dtypes": { | |
| "backbone": "float16", | |
| "head": "float32" | |
| }, | |
| "upstream": { | |
| "repo": "AnkitAI/tinyjev-4b", | |
| "trained_with": "Kev study runner (jaredpalmer/kev @ 2855ba2), trial tinyjev-e11-4b: LoRA r16 lr 5e-5 seed 2, 2 epochs on decision-v7 train (12,576 records), one H100, 45 min; LoRA merged into the base here", | |
| "base_model": "Qwen/Qwen3-4B-Base", | |
| "base_revision": "906bfd4b4dc7f14ee4320094d8b41684abff8539", | |
| "lora_rank": 16, | |
| "lora_alpha": 32, | |
| "merged_tensors": 252, | |
| "full_finetune": false | |
| } | |
| } |