README.md

---
license: mit
base_model:
- deepseek-ai/DeepSeek-V3.2
---

**Note that the MTP layers of this model are also PTPC-quantized.**

# Model Overview

- **Model Architecture:** DeepSeek-V3.2
  - **Input:** Text
  - **Output:** Text
- **Supported Hardware Microarchitecture:** AMD MI350/MI355
- **ROCm**: 7.0
- **Operating System(s):** Linux
- **Inference Engine:** [SGLang](https://docs.sglang.ai/)/[vLLM](https://docs.vllm.ai/en/latest/)
- **Model Optimizer:** [AMD-Quark](https://quark.docs.amd.com/latest/index.html) (V0.10)
  - **Weight quantization:** Perchannel, FP8E4M3, Static
  - **Activation quantization:** Pertoken, FP8E4M3, Dynamic
- **Calibration Dataset:** [Pile](https://huggingface.co/datasets/mit-han-lab/pile-val-backup)

This model was built with deepseek-ai/DeepSeek-V3.2 model by applying [AMD-Quark](https://quark.docs.amd.com/latest/index.html) for FP8E4M3 PTPC quantization.

# Model Quantization

The model was quantized from [deepseek-ai/DeepSeek-V3.2](https://huggingface.co/deepseek-ai/DeepSeek-V3.2) using [AMD-Quark](https://quark.docs.amd.com/latest/index.html). The weights are quantized to FP8 and activations are quantized to FP8.


### Accuracy

<table>
  <tr>
   <td><strong>Benchmark</strong>
   </td>
   <td><strong>DeepSeek-V3.2</strong>
   </td>
   <td><strong>DeepSeek-V3.2-ptpc(this model)</strong>
   </td>
  </tr>
  <tr>
   <td>gsm8k
   </td>
   <td>96.00
   </td>
   <td>95.75
   </td>
  </tr>
</table>

### Reproduction

Docker: rocm/vllm-private:rocm7.1_ubuntu22.04_vllm0.11.2_ptpc_fp8

vllm version: 0.11.2.dev521+gad32e3e19.rocm710

aiter version: 0.1.6.post2.dev55+g59bd8ff2c

lm_eval version: 0.4.9.2
```
export VLLM_USE_V1=1  
export SAFETENSORS_FAST_GPU=1
export VLLM_ROCM_USE_AITER=1
export VLLM_ROCM_USE_AITER_MOE=1
model_path="/model_path/deepseek-ai/DeepSeek-V3.2-ptpc"
vllm serve $model_path \
  --tensor-parallel-size 8 \
  --data-parallel-size 1 \
  --max-num-batched-tokens 32768 \
  --trust-remote-code \
  --no-enable-prefix-caching \
  --disable-log-requests \
  --kv-cache-dtype bfloat16 \
  --gpu_memory_utilization 0.85 \
  --compilation-config '{"cudagraph_mode": "FULL_AND_PIECEWISE"}' \
  --block-size 1

lm_eval \
  --model local-completions \
  --tasks gsm8k \
  --model_args model=/model_path/deepseek-ai/DeepSeek-V3.2-ptpc,base_url=http://127.0.0.1:8000/v1/completions \
  --batch_size auto \
  --limit 400

```

# Deployment

This model can be deployed efficiently using the [vLLM](https://docs.vllm.ai/en/latest/) backends.

# License
Modifications Copyright(c) 2025 Advanced Micro Devices, Inc. All rights reserved.