v9-fix: oracle training routing, calibrated inference, update docs 5a14212 natmin322 commited on Mar 23
v7-fix: enable C5 for task 2+ (add --run_single False, --n_batches_c5 100) 060b67d natmin322 commited on Mar 22
v7: C5 Data-Informed Subspace Init + restructure contributions to 2 core claims aeb2d78 natmin322 commited on Mar 21
fix: reduce CPU RAM to prevent OOM SIGKILL - gc.collect, del temps, eval_accumulation_steps, cache cleanup c03ffe2 natmin322 commited on Mar 18
Revert "fix: reduce OOM in V2 script - fp16+gradient_ckpt for T4, EVAL_BSZ 128->32" 46044f0 natmin322 commited on Mar 18
fix: reduce OOM in V2 script - fp16+gradient_ckpt for T4, EVAL_BSZ 128->32 33d5e3a natmin322 commited on Mar 18
fix: add trust_remote_code=True to second load_dataset call in run_llama.py e9acf77 natmin322 commited on Mar 18
dataset: allow custom dataset code by setting trust_remote_code=True in run_llama.py 55f7d25 natmin322 commited on Mar 18
C4: Spectrally-Conditioned LoRA Training — preconditioned gradient + spectral entropy regularization 2d42b51 natmin322 commited on Mar 17
SpecRoute V3: adaptive bias, symmetric inference, threshold 0.995, batch size optimization 9ea634d natmin322 commited on Mar 17
fix: eval_strategy epoch (was steps=10, causing 437 evals/task → Kaggle session kill) dd291ef natmin322 commited on Mar 16
natmin322 commited on 