davanstrien HF Staff commited on
Commit
b27caef
·
verified ·
1 Parent(s): a927975

Revert nightly index, add note: vLLM nightly x86_64 builds broken as of 2026-02-13

Browse files
Files changed (1) hide show
  1. glm-ocr.py +7 -2
glm-ocr.py CHANGED
@@ -5,11 +5,14 @@
5
  # "pyarrow>=17.0.0,<18.0.0",
6
  # "huggingface-hub",
7
  # "pillow",
8
- # "vllm>=0.16.0rc",
9
  # "toolz",
10
  # "torch",
11
  # ]
12
  #
 
 
 
13
  # [tool.uv]
14
  # prerelease = "allow"
15
  # override-dependencies = ["transformers>=5.1.0"]
@@ -22,8 +25,10 @@ GLM-OCR is a compact 0.9B parameter OCR model achieving 94.62% on OmniDocBench V
22
  Uses CogViT visual encoder with GLM-0.5B language decoder and Multi-Token Prediction
23
  (MTP) loss for fast, accurate document parsing.
24
 
25
- NOTE: Requires vLLM>=0.16.0rc (GLM-OCR added in v0.16.0, PR #33005) and
26
  transformers>=5.1.0 (GLM-OCR support landed in stable release).
 
 
27
  First run may take a few minutes to download and install dependencies.
28
 
29
  Features:
 
5
  # "pyarrow>=17.0.0,<18.0.0",
6
  # "huggingface-hub",
7
  # "pillow",
8
+ # "vllm",
9
  # "toolz",
10
  # "torch",
11
  # ]
12
  #
13
+ # [[tool.uv.index]]
14
+ # url = "https://wheels.vllm.ai/nightly"
15
+ #
16
  # [tool.uv]
17
  # prerelease = "allow"
18
  # override-dependencies = ["transformers>=5.1.0"]
 
25
  Uses CogViT visual encoder with GLM-0.5B language decoder and Multi-Token Prediction
26
  (MTP) loss for fast, accurate document parsing.
27
 
28
+ NOTE: Requires vLLM nightly wheels (GLM-OCR added in v0.16.0, PR #33005) and
29
  transformers>=5.1.0 (GLM-OCR support landed in stable release).
30
+ As of 2026-02-13, vLLM nightly only has ARM wheels — x86_64 builds broken.
31
+ Check https://wheels.vllm.ai/nightly for updated builds before running.
32
  First run may take a few minutes to download and install dependencies.
33
 
34
  Features: