| { | |
| "failed_groups": [], | |
| "groups": { | |
| "attn": { | |
| "bit_identical_tensors": 0, | |
| "relative_l2_drift": 0.011973360644893525, | |
| "tensors": 192 | |
| }, | |
| "embed_tokens": { | |
| "bit_identical_tensors": 0, | |
| "relative_l2_drift": 0.0009845347353876215, | |
| "tensors": 1 | |
| }, | |
| "lm_head": { | |
| "bit_identical_tensors": 0, | |
| "relative_l2_drift": 0.009277163298582905, | |
| "tensors": 1 | |
| }, | |
| "mlp": { | |
| "bit_identical_tensors": 0, | |
| "relative_l2_drift": 0.012908234501502311, | |
| "tensors": 96 | |
| }, | |
| "norm": { | |
| "bit_identical_tensors": 1, | |
| "relative_l2_drift": 0.0009931557005228024, | |
| "tensors": 65 | |
| } | |
| }, | |
| "init": "allenai/OLMo-2-1124-7B-Instruct", | |
| "min_drift": 1e-05, | |
| "ok": true, | |
| "trained": "/home/fs01/smw362/mars/distill/SOLD/outputs/alpaca_smollm_olmo/instruct/inv_reference_mixed_ce_exposure" | |
| } | |