dei-model / src /reasoning

Commit History

  1. Add HuggingFace Hub checkpoint persistence - upload and download checkpoints between jobs

    renpas22
    3b46388
  2. Add checkpoint resumption - automatically resume from latest checkpoint

    renpas22
    5419afd
  3. Fix ReasoningStep attribute - use description not content

    renpas22
    f941008
  4. Fix SPECIAL_TOKENS usage - import at module level and use string literals

    renpas22
    464ac9b
  5. Fix ReasoningChain dataclass - add image field and defaults, fix collate function

    renpas22
    3024a91
  6. Fix None image handling in collate function

    renpas22
    d29f3e7
  7. Convert learning_rate to float explicitly

    renpas22
    9e7779a
  8. Remove dead code with direct config access

    renpas22
    ccd696b
  9. Add getattr defaults for all config parameters

    renpas22
    cd76323
  10. Fix FineVision dataset loading with subset parameter

    renpas22
    e47ae2c
  11. Implement full SFT, PRM, and RL training with dataset loading

    renpas22
    84a183c
  12. Make train_prm and train_rl placeholders - dataset loading needs HF integration

    renpas22
    7bff7cb
  13. Add **kwargs to train_prm and train_rl to accept config parameters

    renpas22
    917e40e
  14. Fix train_prm and train_rl signatures to accept max_steps and learning_rate

    renpas22
    41bcc92
  15. Add placeholder train_sft method

    renpas22
    85ab8c2
  16. Fix quantized model device handling in inference_scaling

    renpas22
    a745b26
  17. Add type conversion for RLConfig parameters

    renpas22
    f15c2d7
  18. Add type conversion and debug logging for config values

    renpas22
    c74a578

  1. Skip .to(device) for quantized models with device_map

    renpas22
    fa9e543
  2. Fix gradient checkpointing for VLM models

    renpas22
    0326431
  3. Fix tokenize() calls to use actual tokenizer

    renpas22
    8268436

  1. Fix len() calls to use actual_tokenizer instead of processor

    renpas22
    4605c1b
  2. Fix tokenizer access for Processor objects

    renpas22
    b8bd3e8
  3. Add VLM support to trainer with auto-detection

    renpas22
    e20135f
  4. Add training scripts and configs

    renpas22
    2b8876a