[ROCm] gemm_a16w16 upstreaming

vllm/entrypoints/openai/tool_parsers/minimax_m2_tool_parser.py

@@ -2,11 +2,12 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 import json
-import re
 import uuid
 from collections.abc import Sequence
 from typing import Any
 
+import regex as re
+
 from vllm.entrypoints.openai.protocol import (
     ChatCompletionRequest,
     DeltaFunctionCall,

vllm/model_executor/layers/utils.py

@@ -100,12 +100,42 @@ def default_unquantized_gemm(
     return torch.nn.functional.linear(x, weight, bias)
 
 
+def use_aiter_triton_gemm(n, m, k, dtype):
+    if (
+        envs.VLLM_ROCM_USE_AITER == 0
+        # MI300's - fp8nuz=True
+        or current_platform.is_fp8_fnuz()
+        or dtype not in [torch.float16, torch.bfloat16]
+    ):
+        return False
+
+    # use hipblaslt for the larger GEMMs
+    if n > 2048 and m > 512:
+        return False
+    return (
+        (m == 5120 and k == 2880)
+        or (m == 2880 and k == 4096)
+        or (m == 128 and k == 2880)
+        or (m == 640 and k == 2880)
+        or (m == 2880 and k == 512)
+    )
+
+
 def rocm_unquantized_gemm_impl(
     x: torch.Tensor, weight: torch.Tensor, bias: torch.Tensor | None = None
 ) -> torch.Tensor:
     from vllm.platforms.rocm import on_gfx9
 
+    x_view = x.view(-1, x.size(-1))
+    n = x_view.shape[0]
+    m = weight.shape[0]
     k = weight.shape[1]
+
+    if use_aiter_triton_gemm(n, m, k, x.dtype):
+        from aiter.ops.triton.gemm_a16w16 import gemm_a16w16
+
+        return gemm_a16w16(x, weight, bias)
+
     use_skinny = (
         envs.VLLM_ROCM_USE_SKINNY_GEMM
         and on_gfx9()
@@ -116,12 +146,8 @@ def rocm_unquantized_gemm_impl(
     if use_skinny is not True:
         return torch.nn.functional.linear(x, weight, bias)
 
-    x_view = x.view(-1, x.size(-1))
-    n = x_view.shape[0]
-    m = weight.shape[0]
-    cu_count = current_platform.get_cu_count()
-
     if m > 8 and 0 < n <= 4:
+        cu_count = current_platform.get_cu_count()
         out = ops.wvSplitK(weight, x_view, cu_count, bias)
         return out.view(*x.shape[:-1], weight.shape[0])
     elif m % 4 == 0 and n == 1 and k <= 8192 and bias is None:
@@ -130,7 +156,7 @@ def rocm_unquantized_gemm_impl(
     return torch.nn.functional.linear(x, weight, bias)
 
 
-def rocm_unquantized_gemm_impl_fake(
+def rocm_unquantized_gemm_fake(
     x: torch.Tensor, weight: torch.Tensor, bias: torch.Tensor | None = None
 ) -> torch.Tensor:
     return x.new_empty((*x.shape[:-1], weight.shape[0]))
@@ -142,13 +168,13 @@ def rocm_unquantized_gemm(
     weight: torch.Tensor,
     bias: torch.Tensor | None = None,
 ) -> torch.Tensor:
-    return torch.ops.vllm.rocm_unquantized_gemm_impl(x, weight, bias)
+    return torch.ops.vllm.rocm_unquantized_gemm(x, weight, bias)
 
 
 direct_register_custom_op(
-    op_name="rocm_unquantized_gemm_impl",
+    op_name="rocm_unquantized_gemm",
     op_func=rocm_unquantized_gemm_impl,
-    fake_impl=rocm_unquantized_gemm_impl_fake,
+    fake_impl=rocm_unquantized_gemm_fake,
 )
 
 

vllm/model_executor/models/gpt_oss.py

@@ -25,12 +25,14 @@
 from vllm.model_executor.layers.logits_processor import LogitsProcessor
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.layers.rotary_embedding import get_rope
+from vllm.model_executor.layers.utils import rocm_unquantized_gemm
 from vllm.model_executor.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,
 )
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 from vllm.model_executor.models.utils import sequence_parallel_chunk
+from vllm.platforms import current_platform
 from vllm.sequence import IntermediateTensors
 from vllm.utils.math_utils import cdiv
 
@@ -153,6 +155,7 @@ def __init__(
 
         self.layer_idx = layer_idx
         self.num_experts = config.num_local_experts
+        self.hidden_size = config.hidden_size
         self.experts_per_token = config.num_experts_per_tok
         self.world_size = dist.get_world_size() if dist.is_initialized() else 1
         self.router = torch.nn.Linear(config.hidden_size, config.num_local_experts)
@@ -177,7 +180,12 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
         if self.is_sequence_parallel:
             x = sequence_parallel_chunk(x)
 
-        g = self.router(x)
+        if current_platform.is_rocm():
+            g = rocm_unquantized_gemm(
+                self, x[:, : self.hidden_size], self.router.weight, self.router.bias
+            )
+        else:
+            g = self.router(x)
         x = self.experts(hidden_states=x, router_logits=g)
 
         if self.is_sequence_parallel: