Lazy import torchao (#17626)

docs/get_started/install.md

@@ -1,7 +1,6 @@
 # Install SGLang
 
 You can install SGLang using one of the methods below.
-
 This page primarily applies to common NVIDIA GPU platforms.
 For other or newer platforms, please refer to the dedicated pages for [AMD GPUs](../platforms/amd_gpu.md), [Intel Xeon CPUs](../platforms/cpu_server.md), [TPU](../platforms/tpu.md), [NVIDIA DGX Spark](https://lmsys.org/blog/2025-11-03-gpt-oss-on-nvidia-dgx-spark/), [NVIDIA Jetson](../platforms/nvidia_jetson.md), [Ascend NPUs](../platforms/ascend_npu.md), and [Intel XPU](../platforms/xpu.md).
 
@@ -18,7 +17,7 @@ uv pip install "sglang"
 **Quick fixes to common problems**
 - In some cases (e.g., GB200), the above command might install a wrong torch version (e.g., the CPU version) due to dependency resolution. To fix this, you can first run the above command and then force-reinstall the correct [PyTorch](https://pytorch.org/get-started/locally/) with the following:
   ```
-  uv pip install "torch" --extra-index-url https://download.pytorch.org/whl/cu129 --force-reinstall
+  uv pip install "torch==2.9.1" "torchvision" --extra-index-url https://download.pytorch.org/whl/cu129 --force-reinstall
   ```
 - For CUDA 13, Docker is recommended (see the Method 3 note on B300/GB300/CUDA 13). If you do not have Docker access, installing the matching `sgl_kernel` wheel from [the sgl-project whl releases](https://github.com/sgl-project/whl/releases) after installing SGLang also works. Replace `X.Y.Z` with the `sgl_kernel` version required by your SGLang (you can find this by running `uv pip show sgl_kernel`). Examples:
   ```bash

python/sglang/srt/configs/lfm2.py

@@ -99,4 +99,3 @@ class Lfm2Config(HFLfm2Config):
 # Cannot use .register() because lfm2 is already registered by transformers
 # Directly modify the internal _extra_content dict instead
 CONFIG_MAPPING._extra_content["lfm2"] = Lfm2Config
-logger.info("Registered SGLang Lfm2Config to override HuggingFace's version")

python/sglang/srt/layers/torchao_utils.py

@@ -42,6 +42,9 @@ def apply_torchao_config_to_model(
         quantize the model, e.g. int4wo-128 means int4 weight only quantization with group_size
         128
     """
+    if torchao_config == "" or torchao_config is None:
+        return model
+
     # Lazy import to suppress some warnings
     from torchao.quantization import (
         float8_dynamic_activation_float8_weight,
@@ -53,9 +56,7 @@ def apply_torchao_config_to_model(
     )
     from torchao.quantization.observer import PerRow, PerTensor
 
-    if torchao_config == "" or torchao_config is None:
-        return model
-    elif "int8wo" in torchao_config:
+    if "int8wo" in torchao_config:
         quantize_(model, int8_weight_only(), filter_fn=proj_filter_conv3d)
     elif "int8dq" in torchao_config:
         quantize_(model, int8_dynamic_activation_int8_weight(), filter_fn=filter_fn)