ViperEkura 02625739fe perf: increase eval batch sizes and add max_seq_len ...

- humaneval/ifeval: default batch_size 64, add --max_seq_len=4096
- mmlu: batch 4 questions x 4 choices per forward, add --batch_size
- ppl: default batch_size 64

2026-07-30 23:55:37 +08:00

..

analyze_weights.py

feat: add --device flag for GPU-accelerated SVD, default to cuda

2026-07-27 08:53:40 +08:00

evaluate_humaneval.py

perf: increase eval batch sizes and add max_seq_len

2026-07-30 23:55:37 +08:00

evaluate_ifd.py

feat: append EOS to response in IFD evaluation

2026-07-28 22:22:59 +08:00

evaluate_ifeval.py

perf: increase eval batch sizes and add max_seq_len

2026-07-30 23:55:37 +08:00

evaluate_mmlu.py

perf: increase eval batch sizes and add max_seq_len

2026-07-30 23:55:37 +08:00

evaluate_ppl.py

perf: increase eval batch sizes and add max_seq_len

2026-07-30 23:55:37 +08:00

evaluate_rouge.py

refactor: map instruction/input/output to chat roles

2026-07-20 13:55:25 +08:00