adding download api to launch client, updating example #196
Conversation
|
can you share screenshot of newly added page? |
i.e. the .ipynb example? @yunfeng-scale not sure what you mean by newly added page? |
|
yixu34
left a comment
yixu34
left a comment
There was a problem hiding this comment.
Normally I'd suggest adding a helper function like Model.download instead of relying on these code snippets, but given that we don't actually want to make this too easy, I think that's fine.
clients/python/llmengine/model.py
Outdated
| def download( | ||
| cls, | ||
| model_name: str, | ||
| download_format: str = "huggingface", |
There was a problem hiding this comment.
It seems like we've used hugging_face elsewhere in LLM Engine. I think the motivation for this was due to Hugging Face being two words?
There was a problem hiding this comment.
oh okay, that's a relatively easy fix (currently we're not actually doing anything with the format token anyway)
| download format requested (default=huggingface) | ||
| Returns: | ||
| DownloadModelResponse: an object that contains a dictionary of filenames, urls from which to download the model weights. | ||
| The urls are presigned urls that grant temporary access and expire after an hour. |
There was a problem hiding this comment.
Think this is OK for now, but not sure if this is technically true in all contexts, e.g. self-hosting.
There was a problem hiding this comment.
This make sense, I'm actually not exactly sure what the behavior would be in a self-hosted context. We can think about this more as it becomes relevant in the future?
No description provided.