Add the support of AMD MI300X/MI325X/MI355X of Ernie 4.5 recipe

Ernie/Ernie4.5.md

@@ -4,15 +4,21 @@ This guide describes how to run [ERNIE-4.5-21B-A3B-PT](https://huggingface.co/ba
 
 ## Installing vLLM
 Note: transformers >= 4.54.0 and vllm >= 0.10.1
-
+### CUDA
 ```bash
 uv venv --python 3.12 --seed
 source .venv/bin/activate
 uv pip install vllm --torch-backend=auto
 ```
 
-## Running Ernie4.5
+### AMD ROCm: MI300x/MI325x/MI355x
+```bash
+uv pip install vllm --extra-index-url https://wheels.vllm.ai/rocm/0.16.0/rocm700
+```
+⚠️ The vLLM wheel for ROCm is compatible with Python 3.12, ROCm 7.0, and glibc >= 2.35. If your environment is incompatible, please use docker flow in [vLLM](https://vllm.ai/) 
 
+## Running Ernie4.5
+### Serving Ernie4.5 Model on H100 GPUs
 ```bash
 # 21B model 80G*1 GPU
 vllm serve baidu/ERNIE-4.5-21B-A3B-PT
@@ -33,8 +39,19 @@ vllm serve baidu/ERNIE-4.5-300B-A47B-PT \
   --tensor-parallel-size 16
 ```
 
-## Running Ernie4.5 MTP
+### Serving Ernie4.5 Model on MI300X/MI325X/MI355X GPUs
+Run the vLLM online serving on AMD GPUs using the command below:
+```bash
+VLLM_ROCM_USE_AITER=1 \
+SAFETENSORS_FAST_GPU=1 \
+vllm serve baidu/ERNIE-4.5-21B-A3B-PT \
+    --tensor-parallel-size 4 \
+    --gpu-memory-utilization 0.9 \
+    --disable-log-requests \
+    --trust-remote-code
+```
 
+## Running Ernie4.5 MTP
 ```bash
 # 21B MTP model 80G*1 GPU
 vllm serve baidu/ERNIE-4.5-21B-A3B-PT \
@@ -58,12 +75,8 @@ vllm serve baidu/ERNIE-4.5-300B-A47B-PT \
   --speculative-config '{"method": "ernie_mtp","model": "baidu/ERNIE-4.5-300B-A47B-PT","num_speculative_tokens": 1}'
 ```
 
-
 ## Benchmarking
-
 For benchmarking, only the first `vllm bench serve` after service startup to ensure it is not affected by prefix cache
-
-
 ```bash
 # Prompt-heavy benchmark (8k/1k)
 vllm bench serve \
@@ -79,18 +92,14 @@ vllm bench serve \
 ```
 
 ### Benchmark Configurations
-
 Test different workloads by adjusting input/output lengths:
-
 - **Prompt-heavy**: 8000 input / 1000 output
 - **Decode-heavy**: 1000 input / 8000 output
 - **Balanced**: 1000 input / 1000 output
 
 Test different batch sizes by changing `--num-prompts`, e.g., 1, 16, 32, 64, 128, 256, 512
 
 ### Expected Output
-
-
 ```shell
 ============ Serving Benchmark Result ============
 Successful requests:                     16        
@@ -114,5 +123,4 @@ Mean ITL (ms):                           16.84
 Median ITL (ms):                         15.49     
 P99 ITL (ms):                            20.69     
 ==================================================
-```
-
+```