Nicolas Patry

:rotating_light: :rotating_light: :rotating_light: Fixing BPE spm converter.

2

# What does this PR do? The spm BPE converter seemed to have been wrong (for quite a while if true). The merges are recreated from the vocab, but where...

Reverting Deta cloning mecanism.

11

# What does this PR do? This one is quite odd. With the revert the slow test will work (I guess what we care most about): ```python from transformers import...

[WIP] Testing safetensors==0.3.1rc1

1

# What does this PR do? Testing Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other...

Upgrading sentencepiece modeling file (for proto > 4 support).

2

# What does this PR do? Upgrades the file. Taken from `google/sentencepiece` directly. Should prevent "Downgrade the protobuf package". Fixes # (issue) ## Before submitting - [ ] This PR...

# What does this PR do? Reworked the loading logic. Idea is to use cleaner loading code: - Remove need for `no_init_weights` - Remove all weird `bnb_linear` and `load_weights` and...

[WIP] Adding GPTQ support for llama

3

# What does this PR do? Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the other checks...

[Meta] Road to 1.0 checklist

11

### CI - [x] Add custom multi GPU runner to CI - [x] test docker image on MR - [x] load test on daily cron (low prio) ### server -...

Inference support for GPTQ (llama + falcon tested) + Quantization script

37

Let's start discussing implementation. - Need to expose the quantization scripts (either included here or add doc on how to use https://github.com/qwopqwop200/GPTQ-for-LLaMa) - Make sure GPTQ works for multiple models...

Adding some help for the options in `text-generation-benchmark`.

5

# What does this PR do? Fixes https://github.com/huggingface/text-generation-inference/issues/420 Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs (you can dismiss the...

# What does this PR do? Adds an integration test on llama-7b-gptq Fixes # (issue) ## Before submitting - [ ] This PR fixes a typo or improves the docs...