Cronologia Commit

Autore SHA1 Messaggio Data
  Matthias Reso 113ea18bf1 Replace LlamaTokenizer with AutoTokenizer 11 mesi fa
  Hamid Shojanazeri 11f51db28c adding the kbit prep in the code 11 mesi fa
  Hamid Shojanazeri f058ff6ccd update due to peft new release 11 mesi fa
  Hamid Shojanazeri ffdc93f00a Merge branch 'main' into wandb_logging 1 anno fa
  Matthias Reso c5a382e509 Make tests run on cpu only machines 1 anno fa
  Hamid Shojanazeri 43ea6bfa71 Merge branch 'main' into ssdp 1 anno fa
  Hamid Shojanazeri d51d2cce9c adding sdpa for flash attn 1 anno fa
  Hamid Shojanazeri 9a2434f408 update the fine-tuning script 1 anno fa
  Hamid Shojanazeri 162be4c045 Revert "Flop counter, profiling and GC (#357)" 1 anno fa
  Hamid Shojanazeri 71d137c722 Merge branch 'main' into flop_counter_gc 1 anno fa
  Less Wright 3f2c33e4f8 Update finetuning.py - remove nightly check 1 anno fa
  Hamid Shojanazeri b15ffeeaf4 clean up 1 anno fa
  Hamid Shojanazeri 19089269d3 add gc 1 anno fa
  kldarek 989b6ee812 wandb logging feedback 1 anno fa
  Abhilash Majumder 4793f0fdf3 Merge branch 'main' into ipex_feature 1 anno fa
  kldarek fc5485d916 fixing wandb for fsdp 1 anno fa
  kldarek f2406cac07 cleanup spaces 1 anno fa
  kldarek cf373529f7 basic wandb logging instrumentation 1 anno fa
  Hamid Shojanazeri aa24e8f57e fix typos 1 anno fa
  Hamid Shojanazeri 38c8cf08c4 adding hsdp 1 anno fa
  Abhilash Majumder 6a78b96764 Merge branch 'main' into ipex_feature 1 anno fa
  Matthias Reso 4c225c65eb Fix order of concat vs sampler 1 anno fa
  Matthias Reso 5da84b2913 Fix usage of dataclass for train_config and fsdp_config 1 anno fa
  Matthias Reso 653a79e3dd Invalidate context in labels for samsum + grammar 1 anno fa
  Matthias Reso a647955fc8 Make packing/padding a training setting 1 anno fa
  Matthias Reso 2e4bd2a665 Resize vocab size to fix idx error 1 anno fa
  Matthias Reso ca41c1c697 Adjust tests to len based batch sampling 1 anno fa
  Shijie Wu 91e2573aa8 pass weight_decay into optimizer 1 anno fa
  Howard Liberty cc356b6017 Add FSDP CPU offloading option 1 anno fa
  abhilash1910 ad6b27d316 merge conflicts 1 anno fa