mosaicml/llm-foundry

Pull Requests Commits

snarayan21 committed 1 year ago

aa2172ea

snarayan21 committed 1 year ago

688365d0

snarayan21 committed 1 year ago

1b678366

snarayan21 committed 1 year ago

e145adb3

snarayan21 committed 1 year ago

89e5a68e

snarayan21 committed 1 year ago

34820714

Dtensor oom (#1395)

dakinggg committed 1 year ago

Verified e8826585

ShashankMosaicML committed 1 year ago

Verified cfab70ed

Support rope scaling (#1391)

milocress committed 1 year ago

Verified 70586c40

bump transformers to 4.43.2 (#1393)

dakinggg committed 1 year ago

Verified af37ea04

Revert "Allow for multiple workers when autopacking (#1375)" (#1392)

dakinggg committed 1 year ago

Verified b636e5f4

Refactor loss function for ComposerMPTCausalLM (#1387)

irenedea committed 1 year ago

Verified ced63ee3

Avoid race condition in convert text to mds script (#1390)

dakinggg committed 1 year ago

Verified 7b160fcc

Revert "Use utils to get shared fs safe signal file name (#1381)" (#1389)

dakinggg committed 1 year ago

Verified 221d3e2b

Add convert_delta_to_json to CLI (#1355)

KuuCi committed 1 year ago

Verified 3d7d12e6

bump transformers (#1388)

dakinggg committed 1 year ago

Verified d49e6a27

[kushalkodnad/tokenizer-registry] Introduce new registry for tokenizers (#1386)

kushalkodn-db committed 1 year ago

Verified 51949c4c

Add transformation hooks to hf_causal_lm (#1383)

irenedea committed 1 year ago

Verified cefd6160

Use utils to get shared fs safe signal file name (#1381)

dakinggg committed 1 year ago

Verified d2d29ada

Do dtype conversion in torch hook to save memory (#1384)

irenedea committed 1 year ago

Verified 596dd9dc

fix load and save planner config logic (#1385)

irenedea committed 1 year ago

Verified 0bed4ffb

Allow train.py-like config for eval.py (#1351)

josejg committed 1 year ago

Verified eb41a6e7

Allow for multiple workers when autopacking (#1375)

b-chu committed 1 year ago

Verified d812f20c

update runners (#1360)

KevDevSha committed 1 year ago

Verified a7b4056a

Update accelerate requirement from <0.26,>=0.25 to >=0.32.1,<0.33 (#1341)

dependabot[bot] committed 1 year ago

Verified 48571959

Allow flash attention up to 3 (#1377)

dakinggg committed 1 year ago

Verified d50b0dbf

Allow for transforms on the model before MLFlow registration (#1372)

snarayan21 committed 1 year ago

Verified feb786cd

Add convert_finetuning_dataset to CLI (#1354)

KuuCi committed 1 year ago

Verified 59f1a0af

rank 0 first (#1371)

dakinggg committed 1 year ago

Verified 900c6a7f

Add LoadPlanner and SavePlanner registries (#1358)

irenedea committed 1 year ago

Verified acb55300

Older