phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

Pin torchtune pkg version (#791)

Browse source 
## context
This is the follow up of
https://github.com/meta-llama/llama-stack/pull/674. Since torchtune is
still in alpha stage and the apis are not guarantee backward compatible.
Pin the torchtune and torchao pkg version to avoid the latest torchtune
release breaks llama stack post training.

We will bump the version number manually after with the new pkg release
some testing

## test 
ping an old torchtune pkg version (0.4.0) and the 0.4.0 was installed 
<img width="1016" alt="Screenshot 2025-01-16 at 3 06 47 PM"
src="https://github.com/user-attachments/assets/630b05d0-8d0d-4e2f-8b48-22e578a62659"
/>
This commit is contained in:
Botao Chen 2025-01-16 16:31:13 -08:00 • committed by GitHub
parent d1f3b032c9
commit 35bf6ea75a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
llama_stack/providers/registry/post_training.py
View file

@ -14,7 +14,7 @@ def available_providers() -> List[ProviderSpec]:
InlineProviderSpec(
api=Api.post_training,
provider_type="inline::torchtune",
pip_packages=["torch", "torchtune", "torchao", "numpy"],
pip_packages=["torch", "torchtune==0.5.0", "torchao==0.8.0", "numpy"],
module="llama_stack.providers.inline.post_training.torchtune",
config_class="llama_stack.providers.inline.post_training.torchtune.TorchtunePostTrainingConfig",
api_dependencies=[