change AMD GPU and dcu to hcu

CMakeLists.txt

@@ -40,7 +40,7 @@ set(PYTHON_SUPPORTED_VERSIONS "3.8" "3.9" "3.10" "3.11" "3.12")
 # Supported NVIDIA architectures.
 set(CUDA_SUPPORTED_ARCHS "7.0;7.5;8.0;8.6;8.9;9.0")
 
-# Supported AMD GPU architectures.
+# Supported hcu architectures.
 set(HIP_SUPPORTED_ARCHS "gfx906;gfx908;gfx90a;gfx940;gfx941;gfx942;gfx1030;gfx1100;gfx926;gfx928;gfx936")
 
 #

benchmarks/benchmark_latency.py

@@ -184,7 +184,7 @@ if __name__ == '__main__':
         default="auto",
         help='Data type for kv cache storage. If "auto", will use model '
         'data type. CUDA 11.8+ supports fp8 (=fp8_e4m3) and fp8_e5m2. '
-        'ROCm (AMD GPU) supports fp8 (=fp8_e4m3)')
+        'ROCm (hcu) supports fp8 (=fp8_e4m3)')
     parser.add_argument(
         '--quantization-param-path',
         type=str,
@@ -193,7 +193,7 @@ if __name__ == '__main__':
         'This should generally be supplied, when KV cache dtype is FP8. '
         'Otherwise, KV cache scaling factors default to 1.0, which may cause '
         'accuracy issues. FP8_E5M2 (without scaling) is only supported on '
-        'cuda version greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is '
+        'cuda version greater than 11.8. On ROCm (hcu), FP8_E4M3 is '
         'instead supported for common inference criteria.')
     parser.add_argument(
         '--profile',

benchmarks/benchmark_prioritization.py

@@ -243,7 +243,7 @@ if __name__ == "__main__":
         default="auto",
         help='Data type for kv cache storage. If "auto", will use model '
         'data type. CUDA 11.8+ supports fp8 (=fp8_e4m3) and fp8_e5m2. '
-        'ROCm (AMD GPU) supports fp8 (=fp8_e4m3)')
+        'ROCm (hcu) supports fp8 (=fp8_e4m3)')
     parser.add_argument(
         '--quantization-param-path',
         type=str,
@@ -252,7 +252,7 @@ if __name__ == "__main__":
         'This should generally be supplied, when KV cache dtype is FP8. '
         'Otherwise, KV cache scaling factors default to 1.0, which may cause '
         'accuracy issues. FP8_E5M2 (without scaling) is only supported on '
-        'cuda version greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is '
+        'cuda version greater than 11.8. On ROCm (hcu), FP8_E4M3 is '
         'instead supported for common inference criteria.')
     parser.add_argument(
         "--device",

benchmarks/benchmark_throughput.py

@@ -522,7 +522,7 @@ if __name__ == "__main__":
         default="auto",
         help='Data type for kv cache storage. If "auto", will use model '
         'data type. CUDA 11.8+ supports fp8 (=fp8_e4m3) and fp8_e5m2. '
-        'ROCm (AMD GPU) supports fp8 (=fp8_e4m3)')
+        'ROCm (hcu) supports fp8 (=fp8_e4m3)')
     parser.add_argument(
         '--quantization-param-path',
         type=str,
@@ -531,7 +531,7 @@ if __name__ == "__main__":
         'This should generally be supplied, when KV cache dtype is FP8. '
         'Otherwise, KV cache scaling factors default to 1.0, which may cause '
         'accuracy issues. FP8_E5M2 (without scaling) is only supported on '
-        'cuda version greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is '
+        'cuda version greater than 11.8. On ROCm (hcu), FP8_E4M3 is '
         'instead supported for common inference criteria.')
     parser.add_argument("--device",
                         type=str,

benchmarks/kernels/benchmark_paged_attention.py

@@ -268,7 +268,7 @@ if __name__ == '__main__':
         default="auto",
         help="Data type for kv cache storage. If 'auto', will use model "
         "data type. CUDA 11.8+ supports fp8 (=fp8_e4m3) and fp8_e5m2. "
-        "ROCm (AMD GPU) supports fp8 (=fp8_e4m3)")
+        "ROCm (hcu) supports fp8 (=fp8_e4m3)")
     args = parser.parse_args()
     print(args)
 

csrc/mamba/causal_conv1d/causal_conv1d.cu

@@ -402,7 +402,7 @@ void causal_conv1d_fwd_launch(ConvParamsBase &params, cudaStream_t stream) {
             // There is a slight signature discrepancy in HIP and CUDA "FuncSetAttribute" function.
             C10_CUDA_CHECK(cudaFuncSetAttribute(
                 (void *) kernel, cudaFuncAttributeMaxDynamicSharedMemorySize, kSmemSize));
-            std::cerr << "Warning (causal_conv1d fwd launch): attempting to set maxDynamicSharedMemorySize on an AMD GPU which is currently a non-op (in ROCm versions <= 6.1). This might lead to undefined behavior. \n" << std::endl;
+            std::cerr << "Warning (causal_conv1d fwd launch): attempting to set maxDynamicSharedMemorySize on an hcu which is currently a non-op (in ROCm versions <= 6.1). This might lead to undefined behavior. \n" << std::endl;
             #endif
         }
         kernel<<<grid, Ktraits::kNThreads, kSmemSize, stream>>>(params);

docs/source/getting_started/amd-installation.rst

@@ -3,7 +3,7 @@
 Installation with ROCm
 ======================
 
-vLLM supports AMD GPUs with ROCm 6.2.
+vLLM supports hcus with ROCm 6.2.
 
 Requirements
 ------------

docs/source/quantization/supported_hardware.rst

@@ -15,7 +15,7 @@ The table below shows the compatibility of various quantization implementations
      - Ampere
      - Ada
      - Hopper
-     - AMD GPU
+     - hcu
      - Intel GPU
      - x86 CPU
      - AWS Inferentia

examples/fp8/README.md

 # FP8 KV Cache 
 
-This utility extracts the KV cache scaling factors from a quantized HF (Hugging Face) model. The extracted scaling factors are saved to a JSON file, which can later be used by vLLM (variable-length language model) during runtime. This tool is particularly useful when the KV cache data type is FP8 and is intended for use on ROCm (AMD GPU) platforms.
+This utility extracts the KV cache scaling factors from a quantized HF (Hugging Face) model. The extracted scaling factors are saved to a JSON file, which can later be used by vLLM (variable-length language model) during runtime. This tool is particularly useful when the KV cache data type is FP8 and is intended for use on ROCm (hcu) platforms.
 
 ## Prerequisites
 
@@ -41,7 +41,7 @@ Usage: extract_scales.py [-h] --quantized_model QUANTIZED_MODEL [--load_format {
 KV Scale Extraction Example
 
 optional arguments:
---quantized_model: Specify either the local path to, or name of, a quantized HF model. It is expected that the quantization format is FP8_E4M3, for use on ROCm (AMD GPU).
+--quantized_model: Specify either the local path to, or name of, a quantized HF model. It is expected that the quantization format is FP8_E4M3, for use on ROCm (hcu).
 Optional arguments:
 --cache_dir: Specify a cache directory to use in the event of a HF model download. (Default: None)
 --load_format: Specify the format of the model's tensor files containing the KV cache scaling factors. (Choices: auto, safetensors, npz, pt; Default: auto)
@@ -87,8 +87,8 @@ optional arguments:
   --max-model-len MAX_MODEL_LEN  Maximum length of a sequence (including prompt and output). If None, will be derived from the model.
   --dtype {auto,half,float16,bfloat16,float,float32}  data type for model weights and activations. The "auto" option will use FP16 precision for FP32 and FP16 models, and BF16 precision for BF16 models.
   --enforce-eager  enforce eager execution
-  --kv-cache-dtype {auto,fp8} Data type for kv cache storage. If "auto", will use model data type. FP8_E5M2 (without scaling) is only supported on cuda version greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is instead supported ```for common inference criteria.
-  --quantization-param-path QUANT_PARAM_JSON Path to the JSON file containing the KV cache scaling factors. This should generally be supplied, when KV cache dtype is FP8. Otherwise, KV cache scaling factors default to 1.0, which may cause accuracy issues. FP8_E5M2 (without scaling) is only supported on cuda version greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is instead supported for common inference criteria.
+  --kv-cache-dtype {auto,fp8} Data type for kv cache storage. If "auto", will use model data type. FP8_E5M2 (without scaling) is only supported on cuda version greater than 11.8. On ROCm (hcu), FP8_E4M3 is instead supported ```for common inference criteria.
+  --quantization-param-path QUANT_PARAM_JSON Path to the JSON file containing the KV cache scaling factors. This should generally be supplied, when KV cache dtype is FP8. Otherwise, KV cache scaling factors default to 1.0, which may cause accuracy issues. FP8_E5M2 (without scaling) is only supported on cuda version greater than 11.8. On ROCm (hcu), FP8_E4M3 is instead supported for common inference criteria.
 ```
 ```
 Example:

examples/fp8/extract_scales.py

@@ -325,12 +325,12 @@ if __name__ == "__main__":
         "use by vLLM (pass this file to the appropriate "
         "runtime typically using the argument "
         "--quantization-param-path <filename>). This is only used "
-        "if the KV cache dtype is FP8 and on ROCm (AMD GPU).")
+        "if the KV cache dtype is FP8 and on ROCm (hcu).")
     parser.add_argument(
         "--quantized-model",
         help="Specify the directory containing a single quantized HF model. "
         "It is expected that the quantization format is FP8_E4M3, for use "
-        "on ROCm (AMD GPU).",
+        "on ROCm (hcu).",
         required=True)
     parser.add_argument(
         "--load_format",

examples/medusa/medusa_benchmark_throughput.py

@@ -470,7 +470,7 @@ if __name__ == "__main__":
         default="auto",
         help='Data type for kv cache storage. If "auto", will use model '
         'data type. CUDA 11.8+ supports fp8 (=fp8_e4m3) and fp8_e5m2. '
-        'ROCm (AMD GPU) supports fp8 (=fp8_e4m3)')
+        'ROCm (hcu) supports fp8 (=fp8_e4m3)')
     parser.add_argument(
         '--quantization-param-path',
         type=str,
@@ -479,7 +479,7 @@ if __name__ == "__main__":
         'This should generally be supplied, when KV cache dtype is FP8. '
         'Otherwise, KV cache scaling factors default to 1.0, which may cause '
         'accuracy issues. FP8_E5M2 (without scaling) is only supported on '
-        'cuda version greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is '
+        'cuda version greater than 11.8. On ROCm (hcu), FP8_E4M3 is '
         'instead supported for common inference criteria.')
     parser.add_argument("--device",
                         type=str,

requirements-rocm.txt

 # Common dependencies
 -r requirements-common.txt
 
-# Dependencies for AMD GPUs
+# Dependencies for hcus
 awscli
 boto3
 botocore

setup.py

@@ -399,9 +399,9 @@ def get_version_add(sha: Optional[str] = None) -> str:
 try:
     __version__ = "0.6.2"
     __version_tuple__ = (0, 6, 2)
-    __dcu_version__ = f'0.6.2+{version}' 
+    __hcu_version__ = f'0.6.2+{version}' 
     
-    from vllm.version import __version__, __version_tuple__, __dcu_version__
+    from vllm.version import __version__, __version_tuple__, __hcu_version__
 except Exception as e:
     import warnings
 
@@ -420,7 +420,7 @@ def get_version():
     version_file = 'vllm/version.py'
     with open(version_file, encoding='utf-8') as f:
         exec(compile(f.read(), version_file, 'exec'))
-    return locals()['__dcu_version__']
+    return locals()['__hcu_version__']
 
 
 def get_vllm_version() -> str:

tests/kernels/test_encoder_decoder_attn.py

@@ -751,7 +751,7 @@ def test_encoder_only(
     No KV cache is required for encoder-only attention.
 
     Note on ROCm/HIP: currently encoder/decoder models are not supported on
-    AMD GPUs, therefore this test simply is skipped if is_hip(). 
+    hcus, therefore this test simply is skipped if is_hip(). 
 
     This test globally forces an override of the usual backend
     auto-selection process, forcing the specific backend-under-test
@@ -860,7 +860,7 @@ def test_e2e_enc_dec_attn(
     to be utilized.
 
     Note on ROCm/HIP: currently encoder/decoder models are not supported on
-    AMD GPUs, therefore this test simply is skipped if is_hip(). 
+    hcus, therefore this test simply is skipped if is_hip(). 
 
     Note on metadata: there is a single attention metadata structure shared by
     all prefill-phase attention operations (encoder, decoder, enc/dec cross), 

vllm/__init__.py

@@ -11,7 +11,7 @@ from vllm.outputs import (CompletionOutput, EmbeddingOutput,
                           EmbeddingRequestOutput, RequestOutput)
 from vllm.pooling_params import PoolingParams
 from vllm.sampling_params import SamplingParams
-from vllm.version import __version__, __version_tuple__, __dcu_version__
+from vllm.version import __version__, __version_tuple__, __hcu_version__
 
 
 __all__ = [

vllm/attention/selector.py

@@ -207,7 +207,7 @@ def which_attn_to_use(
                 # not Instinct series GPUs.
                 logger.info("flash_attn is not supported on NAVI GPUs.")
         else:
-            logger.info("%s is not supported in AMD GPUs.", selected_backend)
+            logger.info("%s is not supported in hcus.", selected_backend)
         return _Backend.ROCM_FLASH
 
     # FlashAttn in NVIDIA GPUs.

vllm/benchmarks/benchmark_throughput.py

@@ -522,7 +522,7 @@ if __name__ == "__main__":
         default="auto",
         help='Data type for kv cache storage. If "auto", will use model '
         'data type. CUDA 11.8+ supports fp8 (=fp8_e4m3) and fp8_e5m2. '
-        'ROCm (AMD GPU) supports fp8 (=fp8_e4m3)')
+        'ROCm (hcu) supports fp8 (=fp8_e4m3)')
     parser.add_argument(
         '--quantization-param-path',
         type=str,
@@ -531,7 +531,7 @@ if __name__ == "__main__":
         'This should generally be supplied, when KV cache dtype is FP8. '
         'Otherwise, KV cache scaling factors default to 1.0, which may cause '
         'accuracy issues. FP8_E5M2 (without scaling) is only supported on '
-        'cuda version greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is '
+        'cuda version greater than 11.8. On ROCm (hcu), FP8_E4M3 is '
         'instead supported for common inference criteria.')
     parser.add_argument("--device",
                         type=str,

vllm/config.py

@@ -936,7 +936,7 @@ class ParallelConfig:
             self.disable_custom_all_reduce = True
             logger.info(
                 "Disabled the custom all-reduce kernel because it is not "
-                "supported on AMD GPUs.")
+                "supported on hcus.")
         if self.ray_workers_use_nsight and not self.use_ray:
             raise ValueError("Unable to use nsight profiling unless workers "
                              "run with Ray.")

vllm/distributed/device_communicators/pynccl_wrapper.py

@@ -195,7 +195,7 @@ class NCCLLibrary:
         except Exception as e:
             logger.error(
                 "Failed to load NCCL library from %s ."
-                "It is expected if you are not running on NVIDIA/AMD GPUs."
+                "It is expected if you are not running on NVIDIA/hcus."
                 "Otherwise, the nccl library might not exist, be corrupted "
                 "or it does not support the current platform %s."
                 "If you already have the library, please set the "

vllm/engine/arg_utils.py

@@ -294,7 +294,7 @@ class EngineArgs:
             default=EngineArgs.kv_cache_dtype,
             help='Data type for kv cache storage. If "auto", will use model '
             'data type. CUDA 11.8+ supports fp8 (=fp8_e4m3) and fp8_e5m2. '
-            'ROCm (AMD GPU) supports fp8 (=fp8_e4m3)')
+            'ROCm (hcu) supports fp8 (=fp8_e4m3)')
         parser.add_argument(
             '--quantization-param-path',
             type=nullable_str,
@@ -304,7 +304,7 @@ class EngineArgs:
             'KV cache dtype is FP8. Otherwise, KV cache scaling factors '
             'default to 1.0, which may cause accuracy issues. '
             'FP8_E5M2 (without scaling) is only supported on cuda version'
-            'greater than 11.8. On ROCm (AMD GPU), FP8_E4M3 is instead '
+            'greater than 11.8. On ROCm (hcu), FP8_E4M3 is instead '
             'supported for common inference criteria.')
         parser.add_argument('--max-model-len',
                             type=int,