Enable Llama2 70B to run with hqt on single card (#50)#762

Closed

HolyFalafel wants to merge 63 commits into

huggingface:mainfrom

HabanaAI:dev/dsemiat/llama_single_card_pre_1.10.0

Commits on Mar 5, 2024

Enable Llama2 70B to run with hqt on single card (#50)

Yantom1
authored and
HolyFalafel
committed

Commits on Mar 11, 2024

Add ControlNet Pipeline (#585)

nngokhale
authored and
HolyFalafel
committed
Fix tests (#669)

ssarkar2
authored and
HolyFalafel
committed
Add Flan T5 to model table (#677)

regisss
authored and
HolyFalafel
committed
Update dev version (#678)

regisss
authored and
HolyFalafel
committed
Change capture logic for HPU graphs in Diffusers pipelines (#679)

regisss
authored and
HolyFalafel
committed
Update llama-7b command to include eval (#665)

authored and
HolyFalafel
committed
Added check if reuse_cache supported by selected model_type (#675)

iermolae
authored and
HolyFalafel
committed
Update examples (#682)

regisss
authored and
HolyFalafel
committed
Add FSDP test to Gaudi2 CI (#683)

regisss
authored and
HolyFalafel
committed
Bring back workaround for Falcon with SynapseAI 1.13 (#685)

regisss
authored and
HolyFalafel
committed
Update examples (#686)

regisss
authored and
HolyFalafel
committed
Fix version check when console output is disabled (#688)

vidyasiv
authored and
HolyFalafel
committed
Patch for Gaudi Text-Generation Pipeline (#690)

sjagtap1803
authored and
HolyFalafel
committed
pt22 roberta fix (#701)

MohitIntel
authored and
HolyFalafel
committed
Add PPO and stack_llama support (#615)

sywangyi
authored and
HolyFalafel
committed
Updated requirements for image-classification samples: datasets>=2.14.0 (#699)

alekseyfa
authored and
HolyFalafel
committed
Update example diff file for image classification (#703)

regisss
authored and
HolyFalafel
committed
Upgrade to Transformers 4.37 (#651)

authored and
HolyFalafel
committed
Add instruction in README to checkout latest stable release (#705)

regisss
authored and
HolyFalafel
committed
Update diff file (#706)

regisss
authored and
HolyFalafel
committed
To fix LLAMA-V2-70B-FT-HF (8x) for eager mode (#709)

mkumargarg
authored and
HolyFalafel
committed
Fix Llama initialization

regisss
authored and
HolyFalafel
committed
Revert "Fix Llama initialization"

regisss
authored and
HolyFalafel
committed
Adding a flag whether to save checkpoint or not in run_clm.py (#711)

yeonsily
authored and
HolyFalafel
committed
Pin Accelerate (#714)

regisss
authored and
HolyFalafel
committed
Change for R1.10.2 (#719)

libinta
authored and
HolyFalafel
committed
Fix Llama initialization (#712)

regisss
authored and
HolyFalafel
committed
Enable torch_compile mode for distributed (#659)

kalyanjk
authored and
HolyFalafel
committed
Small `test_text_generation_example.py` refacto (#725)

regisss
authored and
HolyFalafel
committed
Add optimization for blip text model generation (#653)

sywangyi
authored and
HolyFalafel
committed
Update README, add PPO support (#721)

sywangyi
authored and
HolyFalafel
committed
Update the Mistral model naming (#726)

yafshar
authored and
HolyFalafel
committed
Fix beamsearch crash and incorrect output (#627)

sywangyi
authored and
HolyFalafel
committed
Changing backend name (#708)

authored and
HolyFalafel
committed
Update ppo_trainer.py (#718)

skaulintel
authored and
HolyFalafel
committed
Update example diff files (#730)

regisss
authored and
HolyFalafel
committed
Update latest stable release version in README (#734)

regisss
authored and
HolyFalafel
committed
Enable internal kv bucket in llama (#720)

xt574chen
authored and
HolyFalafel
committed
Add seed in sft example, make sft result reproducable (#735)

sywangyi
authored and
HolyFalafel
committed
Update example diff files

regisss
authored and
HolyFalafel
committed
Fix translation models (#710)

vidyasiv
authored and
HolyFalafel
committed
Adding a flag whether to save checkpoint or not in run_lora_clm.py (#736)

yeonsily
authored and
HolyFalafel
committed
Refactor and update CI for encoder-decoders (#742)

regisss
authored and
HolyFalafel
committed
SDXL fine tuning (#667)

authored and
HolyFalafel
committed
Make style for ruff v0.3

regisss
authored and
HolyFalafel
committed
Enable Mixtral-8x7B (#739)

authored and
HolyFalafel
committed
Support speech recognition with whisper models and seq2seq. (#704)

authored and
HolyFalafel
committed
Update example diff files

regisss
authored and
HolyFalafel
committed
Fix throughput calculation for diffusion models (#715)

skavulya
authored and
HolyFalafel
committed
Fix command in speech recognition example README

regisss
authored and
HolyFalafel
committed
Update CI baselines (#749)

regisss
authored and
HolyFalafel
committed
Fix crash in llama mode in llava image-to-text generation (#755)

sywangyi
authored and
HolyFalafel
committed
Update comment added in #507

regisss
authored and
HolyFalafel
committed
Update Mixtral-8x7B fp8 hqt example (#756)

Jinyan Chen
authored and
HolyFalafel
committed
Controlnet training (#650)

authored and
HolyFalafel
committed
Fix `get_dtype` and `convert_into_dtypes` (#769)

regisss
authored and
HolyFalafel
committed
Override sdpa option in Gaudi (#771)

jiminha
authored and
HolyFalafel
committed
Fix Llama-70B-FSDP model loading issue (#752)

hlahkar
authored and
HolyFalafel
committed
Expose Llama Fused OPs control from run_lora_clm.py (#751)

authored and
HolyFalafel
committed
Add image-to-text and visual question answering example (#738)

sywangyi
authored and
HolyFalafel
committed
Fixing tests by making static_shapes False (#778)

Bhargav
authored and
HolyFalafel
committed
stylize code
HolyFalafel
committed