Skip to content

update features/15_vllm_support notebook - #37

Open
yiksiu-chan wants to merge 2 commits into
ndif-team:docsfrom
yiksiu-chan:docs
Open

yiksiu-chan wants to merge 2 commits into
ndif-team:docsfrom
yiksiu-chan:docs

Conversation

@yiksiu-chan

Copy link
Copy Markdown

No description provided.

@khaiwang khaiwang left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The notebook looks good to me, scripts run well. a few obsolete links need fix, and the vllm-lens dependency needs confirm

" `.chunk(2, dim=-1)`; the separate projections don't exist.\n",
"\n",
" The full inventory with the fix for each lives in the repo's\n",
" [`intervention-gaps/REPORT.md`](https://github.com/ndif-team/nnsight/blob/main/src/nnsight/modeling/vllm/intervention-gaps/REPORT.md).\n",

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this doc will retire soon, you might want to point to docs/models/vllm.md

"async backend for transport compression.\n",
"\n",
"<pre><code>pip install nnsight \"vllm==0.18.0\" \"triton>=3.1\" zstandard\n",
"<pre><code>pip install \"nnsight[vllm]\" \"vllm-lens>=1.1.0\" zstandard\n",

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

is vllm-lens used anywhere in this notebook or nnsight (just curious)?

"\n",
"A full multi-node Docker example, including the Dockerfile and cluster\n",
"test harness, lives at\n",
"[`nnsight/src/nnsight/modeling/vllm/examples/multi_node_with_ray/`](https://github.com/ndif-team/nnsight/tree/main/src/nnsight/modeling/vllm/examples/multi_node_with_ray).\n",

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this doc is also relocated to examples/ray/ now

" distributed executor, and the async engine. Read this if you're modifying\n",
" the integration or trying to understand why a specific intervention\n",
" behaves the way it does inside vLLM.\n",
"- [`src/nnsight/modeling/vllm/examples/multi_node_with_ray/`](https://github.com/ndif-team/nnsight/tree/main/src/nnsight/modeling/vllm/examples/multi_node_with_ray) —\n",

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

same here

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants