mosaicml/llm-foundry

Pull Requests Commits

jerrychen109 committed 1 year ago

4695a627

Merge remote-tracking branch 'origin/main' into jerry/mlflow-objectstore-exp

jerrychen109 committed 1 year ago

6ff214b0

Read UC delta table (#773)

XiaohanZhangCMU committed 1 year ago

Verified 6de8c37f

Minor cleanups (#858)

mvpatel2000 committed 1 year ago

Verified a7c36bcc

Remove foundry time wrangling (#855)

aspfohl committed 1 year ago

Verified c694121e

Fix typo kwargs=>hf_kwargs (#853)

irenedea committed 1 year ago

Verified d822630b

Bump pre-commit version (#847)

b-chu committed 1 year ago

Verified ddba5c8d

Pin feature-complete mlflow object store branch from composer

jerrychen109 committed 1 year ago

9f014f1c

Merge remote-tracking branch 'origin/main' into jerry/mlflow-objectstore-exp

jerrychen109 committed 1 year ago

4a6d4e6b

Disable omegaconf environment variables (#845)

irenedea committed 1 year ago

Verified 4772ba29

Add use_tokenizer_eos option to convert text to mds script (#843)

irenedea committed 1 year ago

Verified c03ca1a5

Fix error when decoding a token in the id gap (or out of range) in a tiktoken tokenizer (#841)

dakinggg committed 1 year ago

Verified 5b994884

Shashank/new benchmarks (#838)

ShashankMosaicML committed 1 year ago

Verified 5e85bd62

Adding support for alibi when using flash attention (#820)

ShashankMosaicML committed 1 year ago

Verified d991f37d

Add safe_load option to restrict HF dataset downloads to allowed file types (#798)

irenedea committed 1 year ago

Verified f0fd749c

Add links for ML HW

abhi-mosaic committed 1 year ago

Verified 083b4b24

Bump einops version, which has improved support for torch compile (#832)

sashaDoubov committed 1 year ago

Verified 63323e59

zero bias conversion corrected (#624)

megha95 committed 1 year ago

Verified 44fe631b

Update composer to 0.17.2 and streaming to 0.7.2 (#822)

irenedea committed 1 year ago

Verified 1c4ebbf4

Use `sync_module_states: True` when using HSDP (#830)

abhi-mosaic committed 1 year ago

Verified 29b41ba9

Align GLU implementation with LLaMa (#829)

vchiley committed 1 year ago

Verified 1ac5c4bd

Link to new composer commit

jerrychen109 committed 1 year ago

8222f59c

Bump turbo version to 0.0.7 (#827)

mvpatel2000 committed 2 years ago

Verified 2b1fa79d

Revert "Default to using tokenizer eos and bos in convert_text_to_mds.py (#823)" (#825)

irenedea committed 2 years ago

Verified f2b7bef4

Default to using tokenizer eos and bos in convert_text_to_mds.py (#823)

irenedea committed 2 years ago

Verified 5eb96e92

Fix token counting to allow there to be no attention mask (#818)

dakinggg committed 2 years ago

Verified 836ab953

Add generic flatten imports to HF checkpointer (#814)

b-chu committed 2 years ago

Verified bbf5cc7e

Remove redundant transposes for rope rotation (#807)

ShashankMosaicML committed 2 years ago

Verified 289536bd

Updating the Flash Attention version to fix cross entropy loss (#812)

ShashankMosaicML committed 2 years ago

Verified 2ba9224f

Async eval callback (#702)

aspfohl committed 2 years ago

Verified a7e916b3

Older