Skip to content

fix: add revision to vLLM loader #427

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
majiayu000 wants to merge 2 commits into
base: main
Choose a base branch
from
Open

fix: add revision to vLLM loader #427

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e91bf16
fix: add revision to vLLM loader
majiayu000 Jan 2, 2026
a764d68
fix: add type annotation for revision parameter
majiayu000 Jan 2, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions unsloth_zoo/vllm_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1640,6 +1640,7 @@ def load_vllm(
gpu_memory_utilization : float = 0.8,
max_seq_length : int = 8192,
dtype : torch.dtype = None,
revision : str = None,
training : bool = True,
Comment on lines 1641 to 1644
Copy link

@chatgpt-codex-connector chatgpt-codex-connector bot Jan 2, 2026

Choose a reason for hiding this comment

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

P2 Badge Keep positional-arg compatibility for load_vllm

Inserting revision between dtype and training changes the positional argument order, so any external callers using positional arguments after dtype will now pass their training boolean into revision, shifting the rest of the parameters. That silently alters behavior (e.g., training flips back to default True, float8_kv_cache stays False, etc.) and can cause wrong runtime settings. To avoid a backward-compat regression, add revision at the end or make the remaining parameters keyword-only.

Useful? React with 👍 / 👎.

float8_kv_cache : bool = False,
random_state : int = 0,
Expand Down Expand Up @@ -2013,6 +2014,7 @@ def load_vllm(

engine_args = dict(
model = model_name,
revision = revision,
gpu_memory_utilization = actual_gpu_memory_utilization,
max_model_len = max_seq_length,
quantization = "bitsandbytes" if use_bitsandbytes else None,
Expand Down