mosaicml/llm-foundry

Pull Requests Commits

cody committed 2 years ago

5fe47329

cody committed 2 years ago

bbed727a

Update extension file not found error to be less confusing (#579)

irenedea committed 2 years ago

Verified e891c30c

StreamingTextDataset takes correct device batch size (#580)

snarayan21 committed 2 years ago

Verified 8687a165

Add regression tests (#574)

dakinggg committed 2 years ago

Verified b8854e16

Add handling for various types of malformed finetuning data (#576)

dakinggg committed 2 years ago

Verified 2799918a

pin version of kernels more precisely (#577)

dblalock committed 2 years ago

Verified 1c75fda6

added sampling_method to StreamingTextDataset (#573)

snarayan21 committed 2 years ago

Verified d8b871d8

Add ValueError in evaluation script if load_path checkpoint is not specified in config for mpt_causal_lm's (#535)

j316chuck committed 2 years ago

Verified 3be3d24f

Refactor build_tokenizer to use kwargs syntax and specify name (#532)

j316chuck committed 2 years ago

Verified 186dd198

Adjust print statements in conversion script (#569)

dakinggg committed 2 years ago

Verified 68448b27

Fix conversion script to work with checkpoints from composer dev (#568)

dakinggg committed 2 years ago

Verified c836cb9c

make 8bit flag optional (#567)

bmosaicml committed 2 years ago

Verified 755354c6

Enable gauntlet training (#501)

bmosaicml committed 2 years ago

Verified aabdb3c7

MPT: Change order of operands to enable PT2 compile for inference (#559)

tdoublep committed 2 years ago

Verified a8c7dc47

change docstring (#563)

bmosaicml committed 2 years ago

Verified aad9f64b

Enable eval script for HuggingFace 8bit models (#516)

es94129 committed 2 years ago

Verified 7f373e39

Fix propagation of `drop_last` and add error message when it would produce no batches (#549)

dakinggg committed 2 years ago

Verified 3eac52e6

Move the model creation to the last step before trainer creation (#547)

dakinggg committed 2 years ago

Verified 3e3c7d37

8-bit LION, take 2 (#514)

dblalock committed 2 years ago

Verified 795ab4a0

Revert "Add Programming to Foundry (#441)" (#557)

rishab-partha committed 2 years ago

Verified db2a8d95

Add Programming to Foundry (#441)

rishab-partha committed 2 years ago

Verified 52a3500f

Fix init device in conversion script and add tests (#556)

dakinggg committed 2 years ago

Verified 2f30418e

Add Composer MPT to FasterTransformer Conversion Script (#519)

nik-mosaic committed 2 years ago

Verified a5b39b67

Add python log level in llm foundry eval script (#546)

j316chuck committed 2 years ago

Verified 78b796bd

add missing mixed init (#539)

dakinggg committed 2 years ago

Verified 66700688

remove : (#537)

mvpatel2000 committed 2 years ago

Verified 1a99f097

Refactor build optimizer and peft models to use kwargs syntax (#525)

j316chuck committed 2 years ago

Verified 36cee64b

Adapt composer -> HF conversion script to support all causal lms (#526)

dakinggg committed 2 years ago

Verified 9946fc6f

Add an mcli yaml for running llama2 models (#533)

dakinggg committed 2 years ago

Verified 0b8c81d6

Older