Code

🦙 Llama3 on TGI - Jetstream Pytorch#90

Code

Merged

tengomucho merged 9 commits intomainhuggingface/optimum-tpu:mainfrom llama3-on-jetstream-pt-tgihuggingface/optimum-tpu:llama3-on-jetstream-pt-tgiCopy head branch name to clipboardSep 10, 2024

+56-20

Commits

Commits on Sep 9, 2024

• fix(engine_loader): correct n_reps and cache_shape settings

  tengomucho

  committed
  26fdbf8

  Copy full SHA for 26fdbf8

• feat(tokenizer): donwload all json files when fetching model

  Show description for 1d63b07
  tengomucho

  committed
  1d63b07

  Copy full SHA for 1d63b07

• feat(jetstream pt): relax llama compatibility requirements

  Show description for d858713
  tengomucho

  committed
  d858713

  Copy full SHA for d858713

• test(jetstream pt): move Llama2-7b test to runslow/nightly

  tengomucho

  committed
  8ce7762

  Copy full SHA for 8ce7762

• fix(jetstream pt): clone weight before mapping

  Show description for 71ba46e
  tengomucho

  committed
  71ba46e

  Copy full SHA for 71ba46e

• test(jetstream pt): add test showing support of Llama3-8B

  tengomucho

  committed
  b323794

  Copy full SHA for b323794

Commits on Sep 10, 2024

• review: fix imports for type checking

  tengomucho

  committed
  6857b5f

  Copy full SHA for 6857b5f

• fix: correct type hint

  tengomucho

  committed
  c9f11da

  Copy full SHA for c9f11da

• fix(pyproject): correct jetstream git revision

  tengomucho

  committed
  525231a

  Copy full SHA for 525231a