| { |
| "failed_groups": [], |
| "groups": { |
| "attn": { |
| "bit_identical_tensors": 0, |
| "relative_l2_drift": 0.010709750470233547, |
| "tensors": 192 |
| }, |
| "embed_tokens": { |
| "bit_identical_tensors": 0, |
| "relative_l2_drift": 0.0009333307752371594, |
| "tensors": 1 |
| }, |
| "lm_head": { |
| "bit_identical_tensors": 1, |
| "relative_l2_drift": 0.0, |
| "tensors": 1 |
| }, |
| "mlp": { |
| "bit_identical_tensors": 0, |
| "relative_l2_drift": 0.011819219220247947, |
| "tensors": 96 |
| }, |
| "norm": { |
| "bit_identical_tensors": 0, |
| "relative_l2_drift": 0.004231760185087731, |
| "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_mse_ref_final" |
| } |
|
|
