fix: hf to int4 script

docs/en/advanced/low-precision.md

@@ -87,6 +87,7 @@ First, download the PTQ (Post-Training Quantization) calibration dataset from Hu
 Next, use the `tools/convert_hf_to_int4.py` script to convert BF16 weights to INT4 format. Ensure that the `--hf-checkpoint` parameter points to a directory where `config.json` contains the correct `quantization_config`. slime will automatically utilize INT4 quantization during weight updates.
 
 ```bash
+pip install llmcompressor
 python tools/convert_hf_to_int4.py \
   --input-dir /path/to/your/original/models \
   --output-dir /path/to/your/save/models \

docs/zh/advanced/low-precision.md

@@ -75,6 +75,7 @@ bash scripts/low_precision/run-qwen3-30b-a3b-fp8.sh
 [wikitext-2-raw-v1](https://huggingface.co/datasets/Salesforce/wikitext/tree/main/wikitext-2-raw-v1)
 接着，使用 `tools/convert_hf_to_int4.py` 脚本进行转换。确保 `--hf-checkpoint` 指向的目录中 `config.json` 包含正确的 `quantization_config`。
 ```bash
+pip install llmcompressor
 python tools/convert_hf_to_int4.py \
   --input-dir /path/to/your/original/models \
   --output-dir /path/to/your/save/models \

tools/convert_hf_to_int4.py

@@ -77,6 +77,7 @@ def main():
         "re:.*embed.*",
         "re:.*self_attn.*",
         "re:.*shared_experts.*",
+        "re:.*mlp\\.gate.*",
         "re:.*mlp\\.(gate|up|gate_up|down)_proj.*",
     ]