Llamas 3.1 405B fp4 changes upstreaming from 355_wip (vllm-project#25135)

maleksan85 · Aleksandr Malyshev · Doug Lehr · xuebwang-amd · commit 04bcde1a1b16 · 2025-10-24T09:19:21.000Z
Signed-off-by: Aleksandr Malyshev &lt;maleksan@amd.com&gt;
Co-authored-by: Aleksandr Malyshev &lt;maleksan@amd.com&gt;
Co-authored-by: Doug Lehr &lt;douglehr@amd.com&gt;
Signed-off-by: xuebwang-amd &lt;xuebwang@amd.com&gt;
diff --git a/vllm/envs.py b/vllm/envs.py
@@ -106,6 +106,8 @@
     VLLM_ROCM_USE_AITER_RMSNORM: bool = True
     VLLM_ROCM_USE_AITER_MLA: bool = True
     VLLM_ROCM_USE_AITER_MHA: bool = True
+    VLLM_ROCM_USE_AITER_FP4_ASM_GEMM: bool = False
+    VLLM_ROCM_USE_TRITON_ROPE: bool = False
     VLLM_ROCM_USE_AITER_FP8BMM: bool = True
     VLLM_ROCM_USE_SKINNY_GEMM: bool = True
     VLLM_ROCM_FP8_PADDING: bool = True
@@ -934,6 +936,18 @@ def get_vllm_port() -> Optional[int]:
     lambda: (os.getenv("VLLM_ROCM_USE_AITER_MHA", "True").lower() in
              ("true", "1")),
 
+    # Whether to use aiter fp4 gemm asm.
+    # By default is disabled.
+    "VLLM_ROCM_USE_AITER_FP4_ASM_GEMM":
+    lambda: (os.getenv("VLLM_ROCM_USE_AITER_FP4_ASM_GEMM", "False").lower() in
+             ("true", "1")),
+
+    # Whether to use aiter rope.
+    # By default is disabled.
+    "VLLM_ROCM_USE_TRITON_ROPE":
+    lambda: (os.getenv("VLLM_ROCM_USE_TRITON_ROPE", "False").lower() in
+             ("true", "1")),
+
     # Whether to use aiter triton fp8 bmm kernel
     # By default is enabled.
     "VLLM_ROCM_USE_AITER_FP8BMM":
@@ -1539,6 +1553,8 @@ def compute_hash() -> str:
         "VLLM_ROCM_USE_AITER_RMSNORM",
         "VLLM_ROCM_USE_AITER_MLA",
         "VLLM_ROCM_USE_AITER_MHA",
+        "VLLM_ROCM_USE_AITER_FP4_ASM_GEMM",
+        "VLLM_ROCM_USE_TRITON_ROPE",
         "VLLM_ROCM_USE_AITER_FP8BMM",
         "VLLM_ROCM_USE_SKINNY_GEMM",
         "VLLM_ROCM_FP8_PADDING",
diff --git a/vllm/model_executor/layers/linear.py b/vllm/model_executor/layers/linear.py
@@ -323,6 +323,12 @@ def __init__(
         return_bias: bool = True,
         disable_tp: bool = False,
     ):
+        # If MergedReplicatedLinear, use output size of each partition.
+        if hasattr(self, "output_sizes"):
+            self.output_partition_sizes = self.output_sizes
+        else:
+            self.output_partition_sizes = [output_size]
+
         super().__init__(input_size,
                          output_size,
                          skip_bias_add,
@@ -335,7 +341,8 @@ def __init__(
         # All the linear layer supports quant method.
         assert self.quant_method is not None
         self.quant_method.create_weights(self,
-                                         self.input_size, [self.output_size],
+                                         self.input_size,
+                                         self.output_partition_sizes,
                                          self.input_size,
                                          self.output_size,
                                          self.params_dtype,
@@ -374,12 +381,15 @@ def weight_loader(self, param: Parameter, loaded_weight: torch.Tensor):
         param.data.copy_(loaded_weight)
 
     def forward(
-        self, x: torch.Tensor
+        self,
+        x: torch.Tensor,
     ) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
         bias = self.bias if not self.skip_bias_add else None
         assert self.quant_method is not None
+
         output = self.quant_method.apply(self, x, bias)
         output_bias = self.bias if self.skip_bias_add else None
+
         if not self.return_bias:
             return output
         return output, output_bias
@@ -413,7 +423,7 @@ class ColumnParallelLinear(LinearBase):
         output_sizes: list of output sizes packed into one output, like for QKV
                        the list would be size 3.
         prefix: The name of the layer in the state dict, including all parents
-                        (e.g. model.layers.0.qkv_proj)
+                        (e.g. model.layers.0.qkv_proj) 
         return_bias: If true, return bias together with outputs in forward pass.
         disable_tp: If true, weights matrix won't be sharded through tp rank.
     """
@@ -535,13 +545,15 @@ def weight_loader_v2(self, param: BasevLLMParameter,
         param.load_column_parallel_weight(loaded_weight=loaded_weight)
 
     def forward(
-        self, input_
+        self,
+        input_,
     ) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
         bias = self.bias if not self.skip_bias_add else None
 
         # Matrix multiply.
         assert self.quant_method is not None
         output_parallel = self.quant_method.apply(self, input_, bias)
+
         if self.gather_output and self.tp_size > 1:
             # All-gather across the partitions.
             output = tensor_model_parallel_all_gather(output_parallel)
@@ -1326,7 +1338,8 @@ def weight_loader_v2(self, param: BasevLLMParameter,
         param.load_row_parallel_weight(loaded_weight=loaded_weight)
 
     def forward(
-        self, input_
+        self,
+        input_,
     ) -> Union[torch.Tensor, tuple[torch.Tensor, Optional[Parameter]]]:
         if self.input_is_parallel:
             input_parallel = input_
@@ -1340,9 +1353,8 @@ def forward(
         # Only fuse bias add into GEMM for rank 0 (this ensures that
         # bias will not get added more than once in TP>1 case)
         bias_ = None if (self.tp_rank > 0 or self.skip_bias_add) else self.bias
-        output_parallel = self.quant_method.apply(self,
-                                                  input_parallel,
-                                                  bias=bias_)
+        output_parallel = self.quant_method.apply(self, input_parallel, bias_)
+
         if self.reduce_results and self.tp_size > 1:
             output = tensor_model_parallel_all_reduce(output_parallel)
         else:
diff --git a/vllm/model_executor/layers/quantization/quark/quark.py b/vllm/model_executor/layers/quantization/quark/quark.py
@@ -402,6 +402,7 @@ def apply(self,
         scheme = layer.scheme
         if scheme is None:
             raise ValueError("A scheme must be defined for each layer")
+
         return scheme.apply_weights(layer, x, bias=bias)
 
 
diff --git a/vllm/model_executor/layers/quantization/quark/schemes/quark_w4a4_mxfp4.py b/vllm/model_executor/layers/quantization/quark/schemes/quark_w4a4_mxfp4.py
@@ -1,20 +1,104 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
+from functools import cache
 from typing import Any, Callable, Optional
 
 import torch
 import torch.nn.functional as F
 
-from vllm.logger import init_logger
+from vllm import envs
 from vllm.model_executor.layers.quantization.quark.schemes import QuarkScheme
 from vllm.model_executor.layers.quantization.utils.mxfp4_utils import (
     OCP_MX_BLOCK_SIZE, dequant_mxfp4, quant_dequant_mxfp4)
 from vllm.model_executor.parameter import (GroupQuantScaleParameter,
                                            PackedvLLMParameter)
 from vllm.platforms import current_platform
 
-logger = init_logger(__name__)
+
+@cache
+def is_rocm_aiter_fp4_asm_gemm_enabled() -> bool:
+    return current_platform.is_rocm() \
+        and envs.VLLM_ROCM_USE_AITER_FP4_ASM_GEMM \
+        and envs.VLLM_ROCM_USE_AITER
+
+
+try:
+    from aiter.ops.shuffle import shuffle_weight
+    from aiter.ops.triton.gemm_afp4wfp4 import gemm_afp4wfp4
+    from aiter.ops.triton.quant import dynamic_mxfp4_quant
+
+    from vllm.utils import direct_register_custom_op
+    if is_rocm_aiter_fp4_asm_gemm_enabled():
+        from aiter import gemm_a4w4, per_1x32_f4_quant_hip
+
+    def gemm_with_dynamic_quant(
+        x: torch.Tensor,
+        weight: torch.Tensor,
+        weight_scale: torch.Tensor,
+        rocm_use_aiter_fp4_asm_gemm: bool = False,
+        out_dtype: Optional[torch.dtype] = torch.bfloat16,
+        x_scales: Optional[torch.Tensor] = None,
+    ) -> torch.Tensor:
+        M = x.shape[0]
+        if rocm_use_aiter_fp4_asm_gemm:
+            if x_scales is None:
+                # use hip quant kernel for performance
+                x_q, x_s = per_1x32_f4_quant_hip(x, shuffle=True)
+            else:
+                x_q = x
+                x_s = x_scales
+
+            # 32 alignment is enough for dim0 padding of output for
+            # gemm_a4w4 kernel
+            y = torch.empty((M + 31) // 32 * 32,
+                            weight.shape[0],
+                            device=x_q.device,
+                            dtype=out_dtype)
+
+            gemm_a4w4(x_q,
+                      weight,
+                      x_s,
+                      weight_scale.view(x_s.dtype),
+                      y,
+                      bpreshuffle=True)
+            return y[:M]
+        else:
+            if x_scales is None:
+                x_q, x_s = dynamic_mxfp4_quant(x)
+            else:
+                x_q = x
+                x_s = x_scales
+            y = torch.empty(x_q.shape[0],
+                            weight.shape[0],
+                            device=x_q.device,
+                            dtype=out_dtype)
+
+            gemm_afp4wfp4(x_q, weight, x_s, weight_scale.T, out_dtype, y)
+            return y
+
+    def gemm_with_dynamic_quant_fake(
+        x: torch.Tensor,
+        weight: torch.Tensor,
+        weight_scale: torch.Tensor,
+        x_scales: torch.Tensor = None,
+        rocm_use_aiter_fp4_asm_gemm: bool = False,
+        out_dtype: Optional[torch.dtype] = torch.bfloat16,
+    ) -> torch.Tensor:
+        return torch.empty((*x.shape[:-1], weight.shape[0]),
+                           dtype=out_dtype,
+                           device=x.device)
+
+    direct_register_custom_op(
+        op_name="gemm_with_dynamic_quant",
+        op_func=gemm_with_dynamic_quant,
+        mutates_args=[],
+        fake_impl=gemm_with_dynamic_quant_fake,
+        dispatch_key=current_platform.dispatch_key,
+    )
+
+except ImportError:
+    dynamic_mxfp4_quant = gemm_afp4wfp4 = None
 
 __all__ = ["QuarkW4A4MXFP4"]
 
@@ -27,29 +111,15 @@ def __init__(self, weight_quant_spec: dict[str, Any],
         self.qscheme = "per_group"
         self.weight_quant_spec = weight_quant_spec
         self.input_quant_spec = input_quant_spec
-
-        self.static_input_scales = not input_quant_spec.get("is_dynamic")
-
-        if self.static_input_scales:
+        self.emulate = not current_platform.supports_mx()
+        self.rocm_use_aiter_fp4_asm_gemm = is_rocm_aiter_fp4_asm_gemm_enabled()
+        if not self.emulate and (dynamic_mxfp4_quant is None
+                                 or gemm_afp4wfp4 is None):
+            # Currently need these kernels if not emulating
             raise NotImplementedError(
-                "QuarkW4A4MXFP4 with static input scales is currently not "
-                "implemented. Please open an issue.")
-
-        if not current_platform.supports_mx():
-            self.emulate = True
-            logger.warning_once(
-                "The current platform does not support native MXFP4 "
-                "computation. Simulated weight dequantization and activation "
-                "QDQ (quantize and dequantize) will be used, with the linear "
-                "layers computed in high precision.")
-        else:
-            self.emulate = True
-            logger.warning_once(
-                "The current platform supports native MXFP4 "
-                "computation, but kernels are not yet integrated in vLLM. "
-                "Simulated weight dequantization and activation "
-                "QDQ (quantize and dequantize) will be used, with the linear "
-                "layers computed in high precision.")
+                f"{self.__class__.__name__} requires AITER to be installed "
+                "for non-emulation mode! Please refer to "
+                "https://github.com/ROCm/aiter for installation details.")
 
     @classmethod
     def get_min_capability(cls) -> int:
@@ -58,8 +128,65 @@ def get_min_capability(cls) -> int:
     def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
         layer.weight = torch.nn.Parameter(layer.weight.data,
                                           requires_grad=False)
-        layer.weight_scale = torch.nn.Parameter(layer.weight_scale.data,
-                                                requires_grad=False)
+
+        if self.emulate:
+            layer.weight_scale = torch.nn.Parameter(layer.weight_scale.data,
+                                                    requires_grad=False)
+            try:
+                from quark.torch.export.nn.modules import realquantizer
+                from quark.torch.quantization.config.config import (
+                    QuantizationSpec)
+            except ImportError as err:
+                raise ImportError(
+                    "The package `amd-quark` is required to use AMD Quark "
+                    "MX-FP4 models. Please install it with `pip install "
+                    "amd-quark`.") from err
+
+            weight_quant_spec = QuantizationSpec.from_dict(
+                self.weight_quant_spec)
+
+            weight_quantizer = realquantizer.get_real_quantizer(
+                qspec=weight_quant_spec,
+                quantizer=None,
+                real_quantized=True,
+                reorder=False,
+                float_dtype=self.out_dtype,
+                scale_shape=layer.weight_scale.shape,
+                zero_point_shape=None,
+            )
+            weight_quantizer.scale.data = layer.weight_scale.data
+
+            layer.weight = torch.nn.Parameter(
+                weight_quantizer(layer.weight.data).to(self.out_dtype),
+                requires_grad=False,
+            )
+            layer.weight_scale = None
+
+            # This call is necessary to release the scales memory.
+            torch.cuda.empty_cache()
+        else:
+            if self.rocm_use_aiter_fp4_asm_gemm:
+                # shuffle weight scale
+                weight_scale_shuffle = layer.weight_scale.data
+                sm, sn = weight_scale_shuffle.shape
+                weight_scale_shuffle = weight_scale_shuffle.view(
+                    sm // 32, 2, 16, sn // 8, 2, 4, 1)
+                weight_scale_shuffle = weight_scale_shuffle.permute(
+                    0, 3, 5, 2, 4, 1, 6).contiguous()
+                weight_scale_shuffle = weight_scale_shuffle.view(sm, sn)
+                layer.weight_scale = torch.nn.Parameter(weight_scale_shuffle,
+                                                        requires_grad=False)
+
+                # shuffle weight
+                weight_shuffle = layer.weight.data
+                weight_shuffle = shuffle_weight(weight_shuffle,
+                                                layout=(16, 16))
+                layer.weight = torch.nn.Parameter(weight_shuffle,
+                                                  requires_grad=False)
+            else:
+                layer.weight_scale = torch.nn.Parameter(
+                    layer.weight_scale.data.T.contiguous(),
+                    requires_grad=False)
 
     def create_weights(self, layer: torch.nn.Module,
                        output_partition_sizes: list[int],
@@ -104,9 +231,9 @@ def apply_weights(self,
 
         if self.emulate:
             dq_w = dequant_mxfp4(layer.weight, layer.weight_scale, x.dtype)
-
             x = quant_dequant_mxfp4(x)
-
             return F.linear(x, dq_w, bias)
         else:
-            raise NotImplementedError()
+            return torch.ops.vllm.gemm_with_dynamic_quant(
+                x, layer.weight, layer.weight_scale,
+                self.rocm_use_aiter_fp4_asm_gemm, self.out_dtype)
diff --git a/vllm/model_executor/layers/rotary_embedding/base.py b/vllm/model_executor/layers/rotary_embedding/base.py
diff --git a/vllm/model_executor/layers/rotary_embedding/rocm_aiter_rope_ops.py b/vllm/model_executor/layers/rotary_embedding/rocm_aiter_rope_ops.py
