Commit Verlauf

Autor SHA1 Nachricht Datum
  Abhilash Majumder 977d68a7ce fix memory bug vor 1 Jahr
  Abhilash Majumder a2ca4a7710 fix bug on indent vor 1 Jahr
  Abhilash Majumder d5f39914e8 Merge branch 'main' into ipex_feature vor 1 Jahr
  abhilash1910 82d3ca6e06 Fix bugs in data loading vor 1 Jahr
  abhilash1910 ed7ba999a9 enable xpu finetuning and inference vor 1 Jahr
  lchu feaa344af3 resolve conflicts vor 1 Jahr
  Hamid Shojanazeri 75f291fe1c resolved conflicts vor 1 Jahr
  Hamid Shojanazeri 44ef280d31 adding flash attention and xformer memory efficient through PT SDPA vor 1 Jahr
  luoyifan 79d0d4fc4e fix some typos. vor 1 Jahr
  lchu 80a4c36707 further fix #90 vor 1 Jahr
  Hamid Shojanazeri 88d3e1febc fix the save_train_param condition vor 1 Jahr
  Hamid Shojanazeri 62be60355a resolving conflicts vor 1 Jahr
  Hamid Shojanazeri 017cadd04b Merge branch 'checkpoint_handler_path_fix' of https://github.com/facebookresearch/llama-recipes into checkpoint_handler_path_fix vor 1 Jahr
  Hamid Shojanazeri 4f70348b94 remove the redundant lr step vor 1 Jahr
  Hamid Shojanazeri 5b916114eb merge main branch vor 1 Jahr
  Hamid Shojanazeri 668c364f6b add rank to save_train_params vor 1 Jahr
  Hamid Shojanazeri 231c9e7da9 adding train_param.yaml saving for fsdp checkpoint loading for inference vor 1 Jahr
  Hamid Shojanazeri 41dd7ff1cb Merge branch 'main' into checkpoint_handler_path_fix vor 1 Jahr
  Hamid Shojanazeri a955ed1999 added checks for dist barrier and commented cuda exapnadable segements and dist_dbug vor 1 Jahr
  Hamid Shojanazeri a2403c7c1a clean up vor 1 Jahr
  Hamid Shojanazeri e9559d2669 fixing the train/eval_loss calcualtion vor 1 Jahr
  Hamid Shojanazeri 4ba4400a75 adding dist barrier before and after checkpointing vor 1 Jahr
  Hamid Shojanazeri a49a2c2804 adding PT cuda allocation expand flag vor 1 Jahr
  Hamid Shojanazeri 442c1ccf7c adding barrier to end of trainer loop vor 1 Jahr
  Hamid Shojanazeri f74d57dc08 printing scores based on fsdp usage or single gpu vor 1 Jahr
  Hamid Shojanazeri 3d887ea483 update with active memory and removing rank0 for eval score vor 1 Jahr
  Hamid Shojanazeri bedb96b78a fixing the full state path in checkpoint handler vor 1 Jahr
  Hamid Shojanazeri 563e572f7c adding active mem stat vor 1 Jahr
  Hamid Shojanazeri bd01f64cbd Merge branch 'main' into fix-cuda_id vor 1 Jahr
  Andrew Gu 71fdc4920a Save memory and fix typos vor 1 Jahr