Buckets:
| # Run inside the extracted remote bundle on a CUDA host. | |
| Set-Location /content/tinymind_remote/bundle | |
| python -m train.cli native-axiom-regenesis-train --dataset /content/tinymind_remote/bundle/data/omni_round_curriculum.jsonl --out-dir /content/tinymind_axiom_regenesis_8b_target --max-steps 50000 --eval-records 256 --limit-records 10000 --dim 2816 --layers 48 --lanes 64 --seq-len 1024 --vocab-size 4096 --tokenizer-mode char_v1 --virtual-dim 1048576 --basis-rank 256 --facets 64 --learning-rate 1.2e-05 --train-batch-size 1 --device cuda | |
| python -m train.cli native-baseline-probe --out-dir /content/tinymind_axiom_regenesis_8b_probe --native-checkpoint /content/tinymind_axiom_regenesis_8b_target/checkpoint.pt --baseline-report /content/tinymind_remote/bundle/baseline/deep_core_probe_report.json --max-new-tokens 160 --device cuda | |
| python -m train.cli native-broad-probe --out-dir /content/tinymind_axiom_regenesis_8b_broad_probe --native-checkpoint /content/tinymind_axiom_regenesis_8b_target/checkpoint.pt --max-new-tokens 192 --device cuda | |
Xet Storage Details
- Size:
- 1.05 kB
- Xet hash:
- 6fcfbaaee7153acf32b672a695ccb41f43d8031bc7c5dce73a65379cb656c80f
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.