Merge pull request #1 from deepseek-ai/main

sincronizando

README.md

@@ -233,8 +233,9 @@ DeepSeek-V3 can be deployed locally using the following hardware and open-source
 3. **LMDeploy**: Enables efficient FP8 and BF16 inference for local and cloud deployment.
 4. **TensorRT-LLM**: Currently supports BF16 inference and INT4/8 quantization, with FP8 support coming soon.
 5. **vLLM**: Support DeepSeek-V3 model with FP8 and BF16 modes for tensor parallelism and pipeline parallelism.
-6. **AMD GPU**: Enables running the DeepSeek-V3 model on AMD GPUs via SGLang in both BF16 and FP8 modes.
+6. **LightLLM**: Supports efficient single-node or multi-node deployment for FP8 and BF16.
-7. **Huawei Ascend NPU**: Supports running DeepSeek-V3 on Huawei Ascend devices.
+7. **AMD GPU**: Enables running the DeepSeek-V3 model on AMD GPUs via SGLang in both BF16 and FP8 modes.
+8. **Huawei Ascend NPU**: Supports running DeepSeek-V3 on Huawei Ascend devices.
 
 Since FP8 training is natively adopted in our framework, we only provide FP8 weights. If you require BF16 weights for experimentation, you can use the provided conversion script to perform the transformation.
 
@@ -321,18 +322,22 @@ For comprehensive step-by-step instructions on running DeepSeek-V3 with LMDeploy
 
 ### 6.4 Inference with TRT-LLM (recommended)
 
-[TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM) now supports the DeepSeek-V3 model, offering precision options such as BF16 and INT4/INT8 weight-only. Support for FP8 is currently in progress and will be released soon. You can access the custom branch of TRTLLM specifically for DeepSeek-V3 support through the following link to experience the new features directly: https://github.com/NVIDIA/TensorRT-LLM/tree/deepseek/examples/deepseek_v3. 
+[TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM) now supports the DeepSeek-V3 model, offering precision options such as BF16 and INT4/INT8 weight-only. Support for FP8 is currently in progress and will be released soon. You can access the custom branch of TRTLLM specifically for DeepSeek-V3 support through the following link to experience the new features directly: https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/deepseek_v3. 
 
 
 ### 6.5 Inference with vLLM (recommended)
 
 [vLLM](https://github.com/vllm-project/vllm) v0.6.6 supports DeepSeek-V3 inference for FP8 and BF16 modes on both NVIDIA and AMD GPUs. Aside from standard techniques, vLLM offers _pipeline parallelism_ allowing you to run this model on multiple machines connected by networks. For detailed guidance, please refer to the [vLLM instructions](https://docs.vllm.ai/en/latest/serving/distributed_serving.html). Please feel free to follow [the enhancement plan](https://github.com/vllm-project/vllm/issues/11539) as well.
 
-### 6.6 Recommended Inference Functionality with AMD GPUs
+### 6.6 Inference with LightLLM (recommended)
+
+[LightLLM](https://github.com/ModelTC/lightllm/tree/main) v1.0.1 supports single-machine and multi-machine tensor parallel deployment for DeepSeek-R1 (FP8/BF16) and provides mixed-precision deployment, with more quantization modes continuously integrated. For more details, please refer to [LightLLM instructions](https://lightllm-en.readthedocs.io/en/latest/getting_started/quickstart.html). Additionally, LightLLM offers PD-disaggregation deployment for DeepSeek-V2, and the implementation of PD-disaggregation for DeepSeek-V3 is in development.
+
+### 6.7 Recommended Inference Functionality with AMD GPUs
 
 In collaboration with the AMD team, we have achieved Day-One support for AMD GPUs using SGLang, with full compatibility for both FP8 and BF16 precision. For detailed guidance, please refer to the [SGLang instructions](#63-inference-with-lmdeploy-recommended).
 
-### 6.7 Recommended Inference Functionality with Huawei Ascend NPUs
+### 6.8 Recommended Inference Functionality with Huawei Ascend NPUs
 The [MindIE](https://www.hiascend.com/en/software/mindie) framework from the Huawei Ascend community has successfully adapted the BF16 version of DeepSeek-V3. For step-by-step guidance on Ascend NPUs, please follow the [instructions here](https://modelers.cn/models/MindIE/deepseekv3).
 
 

inference/kernel.py

@@ -87,7 +87,7 @@ def weight_dequant(x: torch.Tensor, s: torch.Tensor, block_size: int = 128) -> t
 
     Args:
         x (torch.Tensor): The quantized weight tensor of shape (M, N).
-        s (torch.Tensor): The scale tensor of shape (M, N).
+        s (torch.Tensor): The scale tensor of shape (M//block_size, N//block_size).
         block_size (int, optional): The block size to use for dequantization. Defaults to 128.
 
     Returns:

inference/model.py

@@ -392,7 +392,7 @@ def apply_rotary_emb(x: torch.Tensor, freqs_cis: torch.Tensor) -> torch.Tensor:
 
 class MLA(nn.Module):
     """
-    Multi-Headed Attention Layer (MLA).
+    Multi-Head Latent Attention (MLA) Layer.
 
     Attributes:
         dim (int): Dimensionality of the input features.
@@ -442,7 +442,7 @@ class MLA(nn.Module):
 
     def forward(self, x: torch.Tensor, start_pos: int, freqs_cis: torch.Tensor, mask: Optional[torch.Tensor]):
         """
-        Forward pass for the Multi-Headed Attention Layer (MLA).
+        Forward pass for the Multi-Head Latent Attention (MLA) Layer.
 
         Args:
             x (torch.Tensor): Input tensor of shape (batch_size, seq_len, dim).