[Kernel] Support Microsoft Runtime Kernel Lib for our Low Precision Computation - BitBLAS (#6036)

Signed-off-by: xinyuxiao <xinyuxiao2024@gmail.com>
Co-authored-by: xinyuxiao <xinyuxiao2024@gmail.com>

docs/source/features/quantization/bitblas.md

@@ -0,0 +1,40 @@
+# BitBLAS
+
+vLLM now supports [BitBLAS](https://github.com/microsoft/BitBLAS) for more efficient and flexible model inference. Compared to other quantization frameworks, BitBLAS provides more precision combinations.
+
+Below are the steps to utilize BitBLAS with vLLM.
+
+```console
+pip install bitblas>=0.1.0
+```
+
+vLLM reads the model's config file and supports pre-quantized checkpoints.
+
+You can find pre-quantized models on:
+
+- [Hugging Face (BitBLAS)](https://huggingface.co/models?other=bitblas)
+- [Hugging Face (GPTQ)](https://huggingface.co/models?other=gptq)
+
+Usually, these repositories have a `quantize_config.json` file that includes a `quantization_config` section.
+
+## Read bitblas format checkpoint
+
+```python
+from vllm import LLM
+import torch
+
+# "hxbgsyxh/llama-13b-4bit-g-1-bitblas" is a pre-quantized checkpoint.
+model_id = "hxbgsyxh/llama-13b-4bit-g-1-bitblas"
+llm = LLM(model=model_id, dtype=torch.bfloat16, trust_remote_code=True, quantization="bitblas")
+```
+
+## Read gptq format checkpoint
+
+```python
+from vllm import LLM
+import torch
+
+# "hxbgsyxh/llama-13b-4bit-g-1" is a pre-quantized checkpoint.
+model_id = "hxbgsyxh/llama-13b-4bit-g-1"
+llm = LLM(model=model_id, dtype=torch.float16, trust_remote_code=True, quantization="bitblas", max_model_len=1024)
+```

docs/source/features/quantization/index.md

@@ -11,6 +11,7 @@ Quantization trades off model precision for smaller memory footprint, allowing l
 supported_hardware
 auto_awq
 bnb
+bitblas
 gguf
 gptqmodel
 int4

docs/source/features/quantization/supported_hardware.md

@@ -74,6 +74,17 @@ The table below shows the compatibility of various quantization implementations
   * ❌
   * ❌
   * ❌
+- * BitBLAS (GPTQ)
+  * ✅︎
+  * ✅︎
+  * ✅︎
+  * ✅︎
+  * ✅︎
+  * ✅︎
+  * ❌
+  * ❌
+  * ❌
+  * ❌
 - * AQLM
   * ✅︎
   * ✅︎