Update README.md

README.md

@@ -1,14 +1,85 @@
 ---
 license: mit
+library_name: transformers
+pipeline_tag: image-text-to-text
+tags:
+- glm4v_moe
+- AWQ
+- vLLM
 language:
 - zh
 - en
 base_model:
-- zai-org/GLM-4.5-Air-Base
-pipeline_tag: image-text-to-text
-library_name: transformers
+  - zai-org/GLM-4.5V
+base_model_relation: quantized
+
 ---
+# GLM-4.5V-AWQ
+Base model: [zai-org/GLM-4.5V](https://huggingface.co/zai-org/GLM-4.5V)
+
+### 【vLLM Single Node with 4 GPUs — Startup Command】
+<i>❗Required: Use `--enable-expert-parallel` when launching this model.  
+Without it, the `expert tensors` cannot be evenly partitioned.</i>
+
+<i>❗Required for 8 GPUs: Use `prefill/decode disaggregated serving`  
+([reference](https://docs.vllm.ai/en/latest/examples/online_serving/disaggregated_serving.html)),  
+otherwise the `vision attention heads` cannot be evenly partitioned.</i>
+```
+CONTEXT_LENGTH=32768
+
+vllm serve \
+    QuantTrio/GLM-4.5V-AWQ \
+    --served-model-name GLM-4.5V-AWQ \
+    --tool-call-parser glm45 \
+    --reasoning-parser glm45 \
+    --enable-auto-tool-choice \
+    --allowed-local-media-path / \
+    --media-io-kwargs '{"video": {"num_frames": -1}}' \ 
+    --enable-expert-parallel \
+    --swap-space 16 \
+    --max-num-seqs 512 \
+    --max-model-len $CONTEXT_LENGTH \
+    --max-seq-len-to-capture $CONTEXT_LENGTH \
+    --gpu-memory-utilization 0.9 \
+    --tensor-parallel-size 4 \
+    --trust-remote-code \
+    --disable-log-requests \
+    --host 0.0.0.0 \
+    --port 8000
+```
+
+### 【Dependencies / Installation】
+
+As of **2025-08-12**, create a fresh Python environment and run:
 
+```bash
+# Patched vLLM (see: https://github.com/vllm-project/vllm/pull/22716)
+git clone -b glm-45 https://github.com/zRzRzRzRzRzRzR/vllm.git
+cd vllm
+VLLM_USE_PRECOMPILED=1 pip install .
+
+# Install preview build of Transformers with GLM-4.5V support
+pip install transformers-v4.55.0-GLM-4.5V-preview
+```
+
+### 【Logs】
+```
+2025-08-12
+1. Initial commit
+```
+
+### 【Model Files】
+| File Size | Last Updated |
+|-----------|--------------|
+| `57GB`    | `2025-08-12` |
+
+### 【Model Download】
+```python
+from huggingface_hub import snapshot_download
+snapshot_download('QuantTrio/GLM-4.5V-AWQ', cache_dir="your_local_path")
+```
+
+### 【Overview】
 # GLM-4.5V
 
 <div align="center">
@@ -109,4 +180,3 @@ If you use this model, please cite the following paper:
       url={https://arxiv.org/abs/2507.01006},
 }
 ```
-