Hugging Face's logo

Spaces:
jester1177
/
mutant-hunter-env
Sleeping

App Files Community
mutant-hunter-env / scripts

Commit History

  1. Drop GRPO temp to 0.3, bump max_new_tokens to 2048, add inference smoke test⁠

    Krishna1107 +1: Claude Opus 4.7
    576dfc3

  2. Add inference-time retry loop with feedback for self-correction⁠

    Krishna1107 +1: Claude Opus 4.7
    10feced

  3. Add in-context demonstration learning support⁠

    Krishna1107 +1: Claude Opus 4.7
    968797f

  4. Fix make_plots.py to accept --wandb-run-id separately from --training-log-json⁠

    Krishna1107 +1: Claude Opus 4.7
    640263f

  5. pytorch fixes

    Krishna1107
    fe6913c

  6. no unsloth

    Krishna1107
    361a1fc

  7. Retry torch.cuda.is_available() in fresh python procs; NVML ready != cuInit ready

    Krishna1107
    27e2f37

  8. Rewrite run_hf_job_7b.sh end-to-end: nvidia-smi GPU poll, clean phase structure

    Krishna1107
    8e61ac4

  9. Drop venv approach; force-reinstall torch stack over base image to actually upgrade past 2.5.1

    Krishna1107
    1110d0a

  10. Phase 0: build fresh venv + restore CUDA poll + restore upfront huggingface_hub install

    Krishna1107
    c4df898

  11. Install torchaudio alongside torch/torchvision to clear pinned-base-image conflict

    Krishna1107
    7ab4e76

  12. Drop version pins on training extras and torch/torchvision install

    Krishna1107
    241a3bd

  13. Hard reset HF Job env: install torch+torchvision together at fixed versions

    Krishna1107
    bd78955

  14. Upgrade torchvision/torchaudio alongside torch to fix nms op registration

    Krishna1107
    94a6b52

  15. Upgrade torch to 2.5+ in Phase 0; pin trl>=0.14 for GRPOTrainer

    Krishna1107
    ce6481e

  16. Make 7B HF Job Phase 0 resilient: poll for CUDA readiness, install hf_hub up front for trap⁠

    Krishna1107 +1: Claude Opus 4.7
    27a9b93

  17. Use --env instead of --secrets KEY=VALUE for WANDB_API_KEY⁠

    Krishna1107 +1: Claude Opus 4.7
    571af06

  18. Fix wandb login in 7B HF Job: verify env var present, use CLI not Python API⁠

    Krishna1107 +1: Claude Opus 4.7
    cb93805

  19. Add 7B variant of HF Job for stronger cold-start.⁠

    Krishna1107 +1: Claude Opus 4.7
    f249cf5

  20. Add reduced-scope HF Job variant for time-constrained scenarios⁠

    Krishna1107 +1: Claude Opus 4.7
    51e8315

  1. Switch HF Job base image to pytorch:2.5.1-cuda12.4 to fix CUDA13 ABI bleed⁠

    Krishna1107 +1: Claude Opus 4.7
    12ecfb5

  2. Fix torch/torchvision ABI mismatch in HF Job Phase 0⁠

    Krishna1107 +1: Claude Opus 4.7
    efb52e5

  3. Fix GitHub username in HF Job scripts: jester1177 -> melohub-xbit⁠

    Krishna1107 +1: Claude Opus 4.7
    082986e

  4. Fix submit_hf_job.sh syntax: image as positional arg⁠

    Krishna1107 +1: Claude Opus 4.7
    900c737

  5. Pre-submission HF Job patches: Python version compat + skip torch reinstall⁠

    Krishna1107 +1: Claude Opus 4.7
    b5d30d0

  6. Add HF Job training pipeline: persistence-aware run script, judge-facing demo notebook, baseline JSON output⁠

    Krishna1107 +1: Claude Opus 4.7
    e01ee6d

  7. Re-apply training prep work on top of validated Layers 1-5 base⁠

    Krishna1107
    1b83cd4

  8. Initial commit: MutantHunter — RL env for mutation-score-rewarded test generation

    Krishna1107
    91487c9