[ Misc ] Support Fp8 via llm-compressor (#6110)

Co-authored-by: Robert Shaw <rshaw@neuralmagic>

.buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-Instruct-FP8-compressed-tensors.yaml

@@ -0,0 +1,11 @@
+# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m nm-testing/Meta-Llama-3-8B-FP8-compressed-tensors-test -b 32 -l 250 -f 5 -t 1
+model_name: "nm-testing/Meta-Llama-3-8B-FP8-compressed-tensors-test"
+tasks:
+- name: "gsm8k"
+  metrics:
+  - name: "exact_match,strict-match"
+    value: 0.752
+  - name: "exact_match,flexible-extract"
+    value: 0.752
+limit: 250
+num_fewshot: 5

.buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-Instruct-FP8.yaml

@@ -1,4 +1,4 @@
-# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-hf-baseline.sh -m neuralmagic/Meta-Llama-3-8B-Instruct-FP8 -b 32 -l 250 -f 5 -t 1
+# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m neuralmagic/Meta-Llama-3-8B-Instruct-FP8 -b 32 -l 250 -f 5 -t 1
 model_name: "neuralmagic/Meta-Llama-3-8B-Instruct-FP8"
 tasks:
 - name: "gsm8k"

.buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-Instruct-INT8-compressed-tensors.yaml

@@ -0,0 +1,11 @@
+# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m nm-testing/Meta-Llama-3-8B-Instruct-W8-Channel-A8-Dynamic-Per-Token-Test -b "auto" -l 250 -f 5 -t 1
+model_name: "nm-testing/Meta-Llama-3-8B-Instruct-W8-Channel-A8-Dynamic-Per-Token-Test"
+tasks:
+- name: "gsm8k"
+  metrics:
+  - name: "exact_match,strict-match"
+    value: 0.728
+  - name: "exact_match,flexible-extract"
+    value: 0.728
+limit: 250
+num_fewshot: 5

.buildkite/lm-eval-harness/configs/models-small.txt

@@ -1,2 +1,4 @@
 Meta-Llama-3-8B-Instruct.yaml
 Meta-Llama-3-8B-Instruct-FP8.yaml
+Meta-Llama-3-8B-Instruct-FP8-compressed-tensors.yaml
+Meta-Llama-3-8B-Instruct-INT8-compressed-tensors.yaml

.buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh

@@ -46,6 +46,6 @@ while getopts "m:b:l:f:t:" OPT; do
 done
 
 lm_eval --model vllm \
-  --model_args pretrained=$MODEL,tensor_parallel_size=$TP_SIZE \
+  --model_args pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,add_bos_token=true \
   --tasks gsm8k --num_fewshot $FEWSHOT --limit $LIMIT \
   --batch_size $BATCH_SIZE

.buildkite/lm-eval-harness/test_lm_eval_correctness.py

@@ -24,7 +24,8 @@ TP_SIZE = os.environ.get("LM_EVAL_TP_SIZE", 1)
 
 def launch_lm_eval(eval_config):
     model_args = f"pretrained={eval_config['model_name']}," \
-                 f"tensor_parallel_size={TP_SIZE}"
+                 f"tensor_parallel_size={TP_SIZE}," \
+                 f"add_bos_token=true"
 
     results = lm_eval.simple_evaluate(
         model="vllm",

tests/quantization/test_compressed_tensors.py

@@ -9,7 +9,8 @@ import torch
 from vllm import SamplingParams
 from vllm.model_executor.layers.quantization.compressed_tensors.compressed_tensors import (  # noqa: E501
     CompressedTensorsLinearMethod, CompressedTensorsW4A16Sparse24,
-    CompressedTensorsW8A8, CompressedTensorsWNA16)
+    CompressedTensorsW8A8Fp8, CompressedTensorsW8A8Int8,
+    CompressedTensorsWNA16)
 from vllm.model_executor.layers.quantization.compressed_tensors.utils import (
     QuantizationType)
 
@@ -37,12 +38,11 @@ def test_compressed_tensors_w8a8_static_setup(vllm_runner, model_args):
                           CompressedTensorsLinearMethod)
         assert isinstance(down_proj.quant_method,
                           CompressedTensorsLinearMethod)
-        assert isinstance(qkv_proj.scheme, CompressedTensorsW8A8)
+        assert isinstance(qkv_proj.scheme, CompressedTensorsW8A8Int8)
 
         assert qkv_proj.scheme.strategy == strategy
         assert qkv_proj.scheme.is_static_input_scheme
-        expected_type = (torch.int8 if quant_type == QuantizationType.INT else
+        expected_type = torch.int8
-                         torch.float8_e4m3fn)
 
         assert qkv_proj.weight.dtype is expected_type
         assert o_proj.weight.dtype is expected_type
@@ -79,7 +79,7 @@ def test_compressed_tensors_w8a8_dynanmic_per_token(vllm_runner, model_args):
         qkv_proj = layer.self_attn.qkv_proj
 
         assert isinstance(qkv_proj.quant_method, CompressedTensorsLinearMethod)
-        assert isinstance(qkv_proj.scheme, CompressedTensorsW8A8)
+        assert isinstance(qkv_proj.scheme, CompressedTensorsW8A8Int8)
         assert not qkv_proj.scheme.is_static_input_scheme
         assert qkv_proj.scheme.strategy == strategy
         assert qkv_proj.weight.dtype is torch.int8
@@ -123,3 +123,25 @@ def test_compressed_tensors_w4a16_marlin24(vllm_runner):
         sampling_params = SamplingParams()
         output = llm.generate("Hello world!", sampling_params=sampling_params)
         assert output
+
+
+def test_compressed_tensors_fp8(vllm_runner):
+    model_path = "nm-testing/Meta-Llama-3-8B-FP8-compressed-tensors-test"
+    with vllm_runner(model_path) as llm:
+        model = llm.model.llm_engine.model_executor.driver_worker.model_runner.model  # noqa: E501
+        layer = model.model.layers[0]
+
+        qkv_proj = layer.self_attn.qkv_proj
+
+        assert isinstance(qkv_proj.quant_method, CompressedTensorsLinearMethod)
+        assert isinstance(qkv_proj.scheme, CompressedTensorsW8A8Fp8)
+        assert qkv_proj.weight.dtype is torch.float8_e4m3fn
+        assert qkv_proj.input_scale.dtype is torch.float32
+        assert qkv_proj.weight_scale.dtype is torch.float32
+        # should be scalars after processing
+        assert len(qkv_proj.input_scale.shape) == 0
+        assert len(qkv_proj.weight_scale.shape) == 0
+
+        sampling_params = SamplingParams()
+        output = llm.generate("Hello world!", sampling_params=sampling_params)
+        assert output

vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors.py

@@ -9,10 +9,11 @@ from vllm.model_executor.layers.quantization.base_config import (  # noqa: E501
 from vllm.model_executor.layers.quantization.compressed_tensors.schemes import (
     W4A16SPARSE24_SUPPORTED_BITS, WNA16_SUPPORTED_BITS,
     CompressedTensorsScheme, CompressedTensorsW4A16Sparse24,
-    CompressedTensorsW8A8, CompressedTensorsWNA16)
+    CompressedTensorsW8A8Fp8, CompressedTensorsW8A8Int8,
+    CompressedTensorsWNA16)
 from vllm.model_executor.layers.quantization.compressed_tensors.utils import (
     CompressionFormat, QuantizationArgs, QuantizationStrategy,
-    find_first_name_or_class_match)
+    QuantizationType, find_first_name_or_class_match)
 from vllm.platforms import current_platform
 
 
@@ -117,6 +118,40 @@ class CompressedTensorsConfig(QuantizationConfig):
 
         return is_8_bits and is_token and is_symmetric and is_dynamic
 
+    def _is_fp8_w8a8(self, weight_quant: BaseModel,
+                     input_quant: BaseModel) -> bool:
+        # Confirm weights and activations quantized.
+        if weight_quant is None or input_quant is None:
+            return False
+
+        # Confirm we have floating points.
+        if not (weight_quant.type == QuantizationType.FLOAT
+                and input_quant.type == QuantizationType.FLOAT):
+            return False
+
+        # Confirm weight scheme is supported.
+        is_symmetric_weight = weight_quant.symmetric
+        is_static_weight = not weight_quant.dynamic
+        is_per_tensor_weight = (
+            weight_quant.strategy == QuantizationStrategy.TENSOR)
+        if not (is_symmetric_weight and is_static_weight
+                and is_per_tensor_weight):
+            return False
+
+        # Dynamic quantization is always supported if weights supported.
+        if input_quant.dynamic:
+            return True
+
+        # Confirm activation scheme is supported.
+        is_symmetric_activation = input_quant.symmetric
+        is_per_tensor_activation = (
+            input_quant.strategy == QuantizationStrategy.TENSOR)
+        if not (is_symmetric_activation and is_per_tensor_activation):
+            return False
+
+        # All conditions satisfied.
+        return True
+
     def _is_wNa16_group_channel(self, weight_quant: BaseModel,
                                 input_quant: BaseModel) -> bool:
         input_quant_none = input_quant is None
@@ -147,14 +182,21 @@ class CompressedTensorsConfig(QuantizationConfig):
                     strategy=weight_quant.strategy,
                     group_size=weight_quant.group_size)
 
-        if self.quant_format == CompressionFormat.int_quantized.value:
+        if (self.quant_format == CompressionFormat.int_quantized.value or
+                self.quant_format == CompressionFormat.float_quantized.value):
+            if self._is_fp8_w8a8(weight_quant, input_quant):
+                return CompressedTensorsW8A8Fp8(
+                    input_dynamic=input_quant.dynamic)
+
             if self._is_static_tensor_w8a8(weight_quant, input_quant):
-                return CompressedTensorsW8A8(strategy=weight_quant.strategy,
+                return CompressedTensorsW8A8Int8(
-                                             is_static_input_scheme=True)
+                    strategy=weight_quant.strategy,
+                    is_static_input_scheme=True)
 
             if self._is_dynamic_token_w8a8(weight_quant, input_quant):
-                return CompressedTensorsW8A8(strategy=weight_quant.strategy,
+                return CompressedTensorsW8A8Int8(
-                                             is_static_input_scheme=False)
+                    strategy=weight_quant.strategy,
+                    is_static_input_scheme=False)
 
         raise NotImplementedError(
             "No compressed-tensors compatible scheme was found.")
@@ -187,7 +229,7 @@ class CompressedTensorsLinearMethod(LinearMethodBase):
         self.quantization_config = quantization_config
 
     def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
-        return layer.scheme.process_weights_after_loading(layer)
+        layer.scheme.process_weights_after_loading(layer)
 
     def create_weights(self, layer: torch.nn.Module,
                        input_size_per_partition: int,

vllm/model_executor/layers/quantization/compressed_tensors/schemes/__init__.py

@@ -1,8 +1,19 @@
-from .compressed_tensors_scheme import CompressedTensorsScheme  # noqa: F401
+from .compressed_tensors_scheme import CompressedTensorsScheme
-from .compressed_tensors_unquantized import (  # noqa: F401
+from .compressed_tensors_unquantized import CompressedTensorsUnquantized
-    CompressedTensorsUnquantized)
+from .compressed_tensors_w4a16_24 import (W4A16SPARSE24_SUPPORTED_BITS,
-from .compressed_tensors_w4a16_24 import (  # noqa: F401
+                                          CompressedTensorsW4A16Sparse24)
-    W4A16SPARSE24_SUPPORTED_BITS, CompressedTensorsW4A16Sparse24)
+from .compressed_tensors_w8a8_fp8 import CompressedTensorsW8A8Fp8
-from .compressed_tensors_w8a8 import CompressedTensorsW8A8  # noqa: F401
+from .compressed_tensors_w8a8_int8 import CompressedTensorsW8A8Int8
-from .compressed_tensors_wNa16 import WNA16_SUPPORTED_BITS  # noqa: F401
+from .compressed_tensors_wNa16 import (WNA16_SUPPORTED_BITS,
-from .compressed_tensors_wNa16 import CompressedTensorsWNA16  # noqa: F401
+                                       CompressedTensorsWNA16)
+
+__all__ = [
+    "CompressedTensorsScheme",
+    "CompressedTensorsUnquantized",
+    "CompressedTensorsWNA16",
+    "CompressedTensorsW4A16Sparse24",
+    "CompressedTensorsW8A8Int8",
+    "CompressedTensorsW8A8Fp8",
+    "WNA16_SUPPORTED_BITS",
+    "W4A16SPARSE24_SUPPORTED_BITS",
+]

vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8.py

@@ -1,109 +0,0 @@
-from typing import Callable, List, Tuple, Union
-
-import torch
-from torch.nn import Parameter
-
-from vllm import _custom_ops as ops
-from vllm.model_executor.layers.quantization.compressed_tensors.schemes import (
-    CompressedTensorsScheme)
-from vllm.model_executor.layers.quantization.compressed_tensors.utils import (
-    QuantizationStrategy)
-from vllm.model_executor.utils import set_weight_attrs
-
-
-class CompressedTensorsW8A8(CompressedTensorsScheme):
-
-    def __init__(self, strategy: str, is_static_input_scheme: bool):
-        self.strategy = strategy
-        self.is_static_input_scheme = is_static_input_scheme
-
-    # Cutlass kernels support only per-tensor and per-channel cases.
-    # So if we have a fused module (QKV, MLP) with per tensor scales (thus N
-    # scales being passed to the kernel), we convert to the per-channel case.
-    def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
-        if (self.strategy == QuantizationStrategy.TENSOR
-                and len(self.logical_widths) > 1):
-
-            # Load the N per-tensor scales into the channelwise buffer.
-            weight_scale_channel = torch.empty(
-                (sum(self.logical_widths), 1),
-                dtype=torch.float32,
-                device=layer.weight_scale.device)
-            start = 0
-            for idx, logical_width in enumerate(self.logical_widths):
-                end = start + logical_width
-                weight_scale_channel[start:end, :] = layer.weight_scale[idx]
-                start = end
-
-            layer.weight_scale = Parameter(weight_scale_channel,
-                                           requires_grad=False)
-
-        # transpose weights for cutlass.
-        weight = layer.weight
-        layer.weight = Parameter(weight.t(), requires_grad=False)
-
-    def create_weights(self, layer: torch.nn.Module,
-                       output_partition_sizes: List[int],
-                       input_size_per_partition: int,
-                       params_dtype: torch.dtype, weight_loader: Callable,
-                       **kwargs):
-        self.logical_widths = output_partition_sizes
-
-        # WEIGHT SCALE
-        shape: Union[Tuple[int], Tuple[int, int]]
-        if self.strategy == QuantizationStrategy.CHANNEL:
-            shape = (sum(self.logical_widths), 1)
-        else:
-            shape = (len(self.logical_widths), )
-
-        weight_scale = Parameter(torch.empty(*shape, dtype=torch.float32),
-                                 requires_grad=False)
-        layer.register_parameter("weight_scale", weight_scale)
-        if self.strategy == QuantizationStrategy.CHANNEL:
-            set_weight_attrs(weight_scale, {
-                "weight_loader": weight_loader,
-                "output_dim": 0,
-            })
-        else:
-            set_weight_attrs(weight_scale, {
-                "weight_loader": weight_loader,
-                "needs_scalar_to_array": True,
-            })
-
-        # WEIGHT
-        weight = Parameter(torch.empty(sum(output_partition_sizes),
-                                       input_size_per_partition,
-                                       dtype=torch.int8),
-                           requires_grad=False)
-        layer.register_parameter("weight", weight)
-        set_weight_attrs(weight, {
-            "input_dim": 1,
-            "output_dim": 0,
-            "weight_loader": weight_loader,
-        })
-
-        # INPUT SCALE
-        # Static quantization:  load from disk.
-        if self.is_static_input_scheme:
-            input_scale = Parameter(torch.empty(1, dtype=torch.float32),
-                                    requires_grad=False)
-            layer.register_parameter("input_scale", input_scale)
-            set_weight_attrs(input_scale, {
-                "weight_loader": weight_loader,
-                "ignore_warning": True,
-            })
-        # Dynamic quantization: set to None.
-        else:
-            layer.input_scale = None
-
-    def apply_weights(self, layer: torch.nn.Module, x: torch.Tensor):
-        # ops.scaled_int8_quant supports both dynamic and static quant.
-        # * dynamic, layer.input_scale is None and x_scale computed from x.
-        # * static, layer.input_scale is scalar and x_scale is input_scale.
-        x_q, x_scale = ops.scaled_int8_quant(x, layer.input_scale)
-
-        return ops.cutlass_scaled_mm(x_q,
-                                     layer.weight,
-                                     scale_a=x_scale,
-                                     scale_b=layer.weight_scale,
-                                     out_dtype=x.dtype)

vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_fp8.py

@@ -0,0 +1,87 @@
+from typing import Callable, List, Optional
+
+import torch
+
+from vllm.model_executor.layers.quantization.compressed_tensors.schemes import (
+    CompressedTensorsScheme)
+from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
+    apply_fp8_linear, create_per_tensor_scale_param, cutlass_fp8_supported,
+    requantize_with_max_scale)
+from vllm.model_executor.utils import set_weight_attrs
+
+__all__ = ["CompressedTensorsW8A8Fp8"]
+
+
+class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
+
+    def __init__(self, input_dynamic: bool):
+        self.input_dynamic = input_dynamic
+        self.cutlass_fp8_supported = cutlass_fp8_supported()
+
+    # W8A8-Fp8 kernels support only per-tensor and per-channel cases.
+    # So if we have a fused module (QKV, MLP) with per tensor scales (thus N
+    # scales being passed to the kernel), we requantize with a single scale.
+    def process_weights_after_loading(self, layer) -> None:
+        # Dequant -> Quant with max scale.
+        max_w_scale, weight = requantize_with_max_scale(
+            weight=layer.weight,
+            weight_scale=layer.weight_scale,
+            logical_widths=layer.logical_widths,
+        )
+
+        # Update layer with new values.
+        layer.weight = torch.nn.Parameter(weight.t(), requires_grad=False)
+        layer.weight_scale = torch.nn.Parameter(max_w_scale,
+                                                requires_grad=False)
+        if self.input_dynamic:
+            layer.input_scale = None
+        else:
+            layer.input_scale = torch.nn.Parameter(layer.input_scale.max(),
+                                                   requires_grad=False)
+
+    def create_weights(self, layer: torch.nn.Module,
+                       output_partition_sizes: List[int],
+                       input_size_per_partition: int,
+                       params_dtype: torch.dtype, weight_loader: Callable,
+                       **kwargs):
+
+        del params_dtype
+
+        output_size_per_partition = sum(output_partition_sizes)
+        layer.logical_widths = output_partition_sizes
+
+        # WEIGHT
+        weight = torch.nn.Parameter(torch.empty(output_size_per_partition,
+                                                input_size_per_partition,
+                                                dtype=torch.float8_e4m3fn),
+                                    requires_grad=False)
+        layer.register_parameter("weight", weight)
+        set_weight_attrs(weight, {
+            "input_dim": 1,
+            "output_dim": 0,
+            "weight_loader": weight_loader,
+        })
+
+        # WEIGHT SCALE
+        weight_scale = create_per_tensor_scale_param(
+            output_partition_sizes, weight_loader=weight_loader)
+        layer.register_parameter("weight_scale", weight_scale)
+
+        # INPUT SCALE
+        if not self.input_dynamic:
+            input_scale = create_per_tensor_scale_param(
+                output_partition_sizes, weight_loader=weight_loader)
+            layer.register_parameter("input_scale", input_scale)
+
+    def apply_weights(self,
+                      layer: torch.nn.Module,
+                      x: torch.Tensor,
+                      bias: Optional[torch.Tensor] = None) -> torch.Tensor:
+
+        return apply_fp8_linear(
+            input=x,
+            weight=layer.weight,
+            weight_scale=layer.weight_scale,
+            input_scale=layer.input_scale,
+            bias=bias,
+            cutlass_fp8_supported=self.cutlass_fp8_supported)

vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_int8.py

@@ -0,0 +1,85 @@
+from typing import Callable, List
+
+import torch
+from torch.nn import Parameter
+
+from vllm.model_executor.layers.quantization.compressed_tensors.schemes import (
+    CompressedTensorsScheme)
+from vllm.model_executor.layers.quantization.compressed_tensors.utils import (
+    QuantizationStrategy)
+from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
+    apply_int8_linear, convert_to_channelwise, create_per_channel_scale_param,
+    create_per_tensor_scale_param)
+from vllm.model_executor.utils import set_weight_attrs
+
+
+class CompressedTensorsW8A8Int8(CompressedTensorsScheme):
+
+    def __init__(self, strategy: str, is_static_input_scheme: bool):
+        self.strategy = strategy
+        self.is_static_input_scheme = is_static_input_scheme
+
+    def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
+        # WEIGHT
+        # Cutlass kernels need transposed weight.
+        weight = layer.weight
+        layer.weight = Parameter(weight.t(), requires_grad=False)
+
+        # WEIGHT SCALE
+        # Cutlass kernels support only per-tensor and per-channel.
+        # If we have a fused module (QKV, MLP) with per tensor scales (thus N
+        # scales being passed to the kernel), convert to the per-channel case.
+        is_fused_module = len(self.logical_widths) > 1
+        if is_fused_module and self.strategy == QuantizationStrategy.TENSOR:
+            ws_channelwise = convert_to_channelwise(layer.weight_scale,
+                                                    self.logical_widths)
+            layer.weight_scale = Parameter(ws_channelwise, requires_grad=False)
+
+        # INPUT SCALE
+        if self.is_static_input_scheme:
+            layer.input_scale = Parameter(layer.input_scale.max(),
+                                          requires_grad=False)
+        else:
+            layer.input_scale = None
+
+    def create_weights(self, layer: torch.nn.Module,
+                       output_partition_sizes: List[int],
+                       input_size_per_partition: int,
+                       params_dtype: torch.dtype, weight_loader: Callable,
+                       **kwargs):
+        self.logical_widths = output_partition_sizes
+
+        # WEIGHT
+        weight = Parameter(torch.empty(sum(output_partition_sizes),
+                                       input_size_per_partition,
+                                       dtype=torch.int8),
+                           requires_grad=False)
+        layer.register_parameter("weight", weight)
+        set_weight_attrs(weight, {
+            "input_dim": 1,
+            "output_dim": 0,
+            "weight_loader": weight_loader,
+        })
+
+        # WEIGHT SCALE
+        layer_kwargs = {"weight_loader": weight_loader}
+        if self.strategy == QuantizationStrategy.CHANNEL:
+            scale = create_per_channel_scale_param(output_partition_sizes,
+                                                   **layer_kwargs)
+        else:
+            assert self.strategy == QuantizationStrategy.TENSOR
+            scale = create_per_tensor_scale_param(output_partition_sizes,
+                                                  **layer_kwargs)
+        layer.register_parameter("weight_scale", scale)
+
+        # INPUT SCALE
+        if self.is_static_input_scheme:
+            scale = create_per_tensor_scale_param(output_partition_sizes,
+                                                  **layer_kwargs)
+            layer.register_parameter("input_scale", scale)
+
+    def apply_weights(self, layer: torch.nn.Module, x: torch.Tensor):
+        return apply_int8_linear(input=x,
+                                 weight=layer.weight,
+                                 weight_scale=layer.weight_scale,
+                                 input_scale=layer.input_scale)

vllm/model_executor/layers/quantization/compressed_tensors/utils.py

@@ -9,6 +9,7 @@ from torch.nn import Module
 class CompressionFormat(Enum):
     dense = "dense"
     sparse_bitmask = "sparse-bitmask"
+    float_quantized = "float-quantized"
     int_quantized = "int-quantized"
     pack_quantized = "pack-quantized"
     marlin_24 = "marlin-24"

vllm/model_executor/layers/quantization/fp8.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List, Optional, Union
+from typing import Any, Dict, List, Optional
 
 import torch
 from torch.nn import Module
@@ -11,11 +11,11 @@ from vllm.model_executor.layers.fused_moe import (FusedMoE, FusedMoEMethodBase,
 from vllm.model_executor.layers.linear import LinearBase, LinearMethodBase
 from vllm.model_executor.layers.quantization.base_config import (
     QuantizationConfig, QuantizeMethodBase)
-from vllm.model_executor.layers.quantization.gptq_marlin import (
-    GPTQ_MARLIN_MAX_PARALLEL, GPTQ_MARLIN_MIN_THREAD_N, GPTQMarlinState,
-    marlin_permute_scales)
 from vllm.model_executor.layers.quantization.utils.marlin_utils import (
-    pack_fp8_to_int32)
+    apply_fp8_marlin_linear, prepare_fp8_layer_for_marlin)
+from vllm.model_executor.layers.quantization.utils.w8a8_utils import (
+    all_close_1d, apply_fp8_linear, create_per_tensor_scale_param,
+    cutlass_fp8_supported, per_tensor_dequantize, requantize_with_max_scale)
 from vllm.model_executor.utils import set_weight_attrs
 from vllm.platforms import current_platform
 from vllm.utils import print_warning_once
@@ -25,13 +25,6 @@ ACTIVATION_SCHEMES = ["static", "dynamic"]
 logger = init_logger(__name__)
 
 
-def cutlass_fp8_supported() -> bool:
-    capability = current_platform.get_device_capability()
-    capability = capability[0] * 10 + capability[1]
-
-    return ops.cutlass_scaled_mm_supports_fp8(capability)
-
-
 class Fp8Config(QuantizationConfig):
     """Config class for FP8."""
 
@@ -117,23 +110,6 @@ class Fp8LinearMethod(LinearMethodBase):
         capability = capability[0] * 10 + capability[1]
         self.use_marlin = capability < 89
 
-    def _create_scale_param(
-        self,
-        scale_name: str,
-        layer: torch.nn.Module,
-        output_partition_sizes: List[int],
-        **extra_weight_attrs,
-    ) -> None:
-        scale = Parameter(torch.empty(len(output_partition_sizes),
-                                      dtype=torch.float32),
-                          requires_grad=False)
-        scale[:] = torch.finfo(torch.float8_e4m3fn).min
-        layer.register_parameter(scale_name, scale)
-        set_weight_attrs(scale, {
-            **extra_weight_attrs,
-            "needs_scalar_to_array": True,
-        })
-
     def create_weights(
         self,
         layer: torch.nn.Module,
@@ -147,7 +123,6 @@ class Fp8LinearMethod(LinearMethodBase):
         del input_size, output_size
         output_size_per_partition = sum(output_partition_sizes)
 
-        layer.process_after_load = True
         layer.logical_widths = output_partition_sizes
 
         layer.input_size_per_partition = input_size_per_partition
@@ -173,144 +148,50 @@ class Fp8LinearMethod(LinearMethodBase):
         # Otherwise, wait until process_weights_after_loading.
         if self.quant_config.is_checkpoint_fp8_serialized:
             # WEIGHT SCALE
-            self._create_scale_param(
+            scale = create_per_tensor_scale_param(output_partition_sizes,
-                scale_name="weight_scale",
+                                                  **extra_weight_attrs)
-                layer=layer,
+            layer.register_parameter("weight_scale", scale)
-                output_partition_sizes=output_partition_sizes,
-                **extra_weight_attrs)
 
             # INPUT ACTIVATION SCALE
             if self.quant_config.activation_scheme == "static":
-                self._create_scale_param(
+                scale = create_per_tensor_scale_param(output_partition_sizes,
-                    scale_name="input_scale",
+                                                      **extra_weight_attrs)
-                    layer=layer,
+                layer.register_parameter("input_scale", scale)
-                    output_partition_sizes=output_partition_sizes,
-                    **extra_weight_attrs)
-
-        # For GPUs without FP8 hardware support, we use Marlin for fast
-        # fused dequantization
-        if self.use_marlin:
-            layer.marlin_state = GPTQMarlinState.REPACK
-
-    def prepare_layer_for_marlin(self, layer: Module) -> None:
-        print_warning_once(
-            "Your GPU does not have native support for FP8 computation but "
-            "FP8 quantization is being used. Weight-only FP8 compression will "
-            "be used leveraging the Marlin kernel. This may degrade "
-            "performance for compute-heavy workloads.")
-
-        part_size_n = layer.output_size_per_partition
-        part_size_k = layer.input_size_per_partition
-
-        assert layer.marlin_state == GPTQMarlinState.REPACK
-        layer.marlin_state = GPTQMarlinState.READY
-
-        device = layer.weight.device
-
-        # WEIGHTS
-        # Repack weights to gptq format (packed int32 elements)
-        packed_gptq_qweight = pack_fp8_to_int32(layer.weight)
-
-        # Repack weights to marlin format
-        marlin_qweight = ops.gptq_marlin_repack(
-            b_q_weight=packed_gptq_qweight,
-            perm=torch.empty(0, dtype=torch.int, device=device),
-            size_k=part_size_k,
-            size_n=part_size_n,
-            num_bits=8,
-        )
-        layer.weight = Parameter(marlin_qweight, requires_grad=False)
-
-        # WEIGHT SCALES
-        # Currently Marlin doesn't support per-tensor scales, so we
-        # expand it to channelwise
-        scales = layer.weight_scale.repeat(1, part_size_n).to(
-            layer.orig_dtype).to(device)
-        # Permute scales
-        marlin_scales = marlin_permute_scales(
-            s=scales,
-            size_k=part_size_k,
-            size_n=part_size_n,
-            group_size=-1,
-            num_bits=8,
-        )
-        layer.weight_scale = Parameter(marlin_scales, requires_grad=False)
-
-        # Allocate marlin workspace
-        max_workspace_size = (
-            part_size_n // GPTQ_MARLIN_MIN_THREAD_N) * GPTQ_MARLIN_MAX_PARALLEL
-        workspace = torch.zeros(max_workspace_size,
-                                dtype=torch.int,
-                                device=device,
-                                requires_grad=False)
-
-        layer.workspace = workspace
 
     def process_weights_after_loading(self, layer: Module) -> None:
-        if (not hasattr(layer, "process_after_load")
+        # If checkpoint not serialized fp8, quantize the weights.
-                or not layer.process_after_load):
-            return
-
-        # If checkpoint is fp/bf16 (not serialized fp8), quantize the weights.
         if not self.quant_config.is_checkpoint_fp8_serialized:
             qweight, weight_scale = ops.scaled_fp8_quant(layer.weight,
                                                          scale=None)
+
+            # Update the layer with the new values.
             layer.weight = Parameter(qweight.t(), requires_grad=False)
             layer.weight_scale = Parameter(weight_scale, requires_grad=False)
-            layer.logical_widths = None
             layer.input_scale = None
-            if self.use_marlin:
-                self.prepare_layer_for_marlin(layer)
-            return
 
         # If checkpoint is fp8, requantize the separately quantized logical
         # weights into a single fp8 weight with a single weight scale.
         else:
-            # WEIGHT_SCALE / WEIGHT
+            # Dequant -> Quant with max scale.
-            #   Loop over logical weights, requantizing with single scale.
+            max_w_scale, weight = requantize_with_max_scale(
-            max_w_scale = layer.weight_scale.max()
+                weight=layer.weight,
+                weight_scale=layer.weight_scale,
+                logical_widths=layer.logical_widths,
+            )
 
-            # QKV / MLP is fused in the on disk checkpoint if any of the
+            # Update layer with new values.
-            # weight scales are still set to the default since we initialize
-            # N weight scales for N shards but we only load 1 weight scale
-            # from disk in this case. As a result, we skip dequant -> requant
-            # since we already have quantized QKV together.
-            # Sample Model with fused checkpoint:
-            #   * nm-testing/Phi-3-mini-128k-instruct-FP8
-            unfused_module_in_checkpoint = (
-                layer.weight_scale[-1] > torch.finfo(torch.float8_e4m3fn).min)
-
-            if unfused_module_in_checkpoint:
-                start = 0
-                for idx, logical_width in enumerate(layer.logical_widths):
-                    end = start + logical_width
-                    weight_dq = per_tensor_dequantize(
-                        layer.weight[start:end, :], layer.weight_scale[idx])
-
-                    layer.weight[start:end, :] = per_tensor_quantize(
-                        weight_dq, layer.weight_scale.max())
-                    start = end
-            layer.weight_scale = Parameter(max_w_scale, requires_grad=False)
-
-            # WEIGHT
-            #   Transpose weight for passing to torch._scaled_mm
-            weight = layer.weight
             layer.weight = Parameter(weight.t(), requires_grad=False)
-
+            layer.weight_scale = Parameter(max_w_scale, requires_grad=False)
-            # INPUT ACTIVATION SCALE
+            if self.quant_config.activation_scheme == "static":
-            #   Dynamic: set to None (required input to ops.scaled_fp8_quant).
-            #   Static:  set to max of the input_scales (since they are equal).
-            if self.quant_config.activation_scheme == "dynamic":
-                layer.input_scale = None
-            elif self.quant_config.activation_scheme == "static":
                 layer.input_scale = Parameter(layer.input_scale.max(),
                                               requires_grad=False)
             else:
-                raise ValueError(
+                layer.input_scale = None
-                    f"Unknown scheme {self.quant_config.activation_scheme}")
 
-            if self.use_marlin:
+        if self.use_marlin:
-                self.prepare_layer_for_marlin(layer)
+            prepare_fp8_layer_for_marlin(layer)
+            # Activations not quantized for marlin.
+            del layer.input_scale
 
     def apply(self,
               layer: torch.nn.Module,
@@ -318,65 +199,22 @@ class Fp8LinearMethod(LinearMethodBase):
               bias: Optional[torch.Tensor] = None) -> torch.Tensor:
 
         if self.use_marlin:
-            # For GPUs that lack FP8 hardware support, we can leverage the
+            return apply_fp8_marlin_linear(
-            # Marlin kernel for fast weight-only FP8 quantization
+                input=x,
-
+                weight=layer.weight,
-            reshaped_x = x.reshape(-1, x.shape[-1])
+                weight_scale=layer.weight_scale,
-            out_shape = x.shape[:-1] + (layer.output_size_per_partition, )
-
-            output = ops.fp8_marlin_gemm(
-                a=reshaped_x,
-                b_q_weight=layer.weight,
-                b_scales=layer.weight_scale,
                 workspace=layer.workspace,
-                num_bits=8,
-                size_m=reshaped_x.shape[0],
                 size_n=layer.output_size_per_partition,
                 size_k=layer.input_size_per_partition,
-            )
+                bias=bias)
 
-            if bias is not None:
+        return apply_fp8_linear(
-                output.add_(bias)  # In-place add
+            input=x,
-
+            weight=layer.weight,
-            return output.reshape(out_shape)
+            weight_scale=layer.weight_scale,
-
+            input_scale=layer.input_scale,
-        else:
+            bias=bias,
-
+            cutlass_fp8_supported=self.cutlass_fp8_supported)
-            # ops.scaled_fp8_quant supports both dynamic and static quant.
-            # If dynamic, layer.input_scale is None and x_scale computed from x
-            # If static, layer.input_scale is scalar and x_scale is input_scale
-
-            if bias is None and self.cutlass_fp8_supported:
-                qinput, x_scale = ops.scaled_fp8_quant(x, layer.input_scale)
-
-                # Fused GEMM_DQ
-                output = ops.cutlass_scaled_mm(
-                    qinput,
-                    layer.weight,
-                    out_dtype=x.dtype,
-                    scale_a=x_scale,
-                    scale_b=layer.weight_scale,
-                )
-
-            else:
-                qinput, x_scale = ops.scaled_fp8_quant(x,
-                                                       layer.input_scale,
-                                                       batch_dim_padding=17)
-
-                # Fused GEMM_DQ -- note we padded the input above because
-                # torch._scaled_mm is more performant for matrices with
-                # batch dimension > 16. Note that this could change
-                # in the future.
-                output, _ = torch._scaled_mm(
-                    qinput,
-                    layer.weight,
-                    out_dtype=x.dtype,
-                    scale_a=x_scale,
-                    scale_b=layer.weight_scale,
-                    bias=bias,
-                )
-
-        return torch.narrow(output, 0, 0, x.shape[0])
 
 
 class Fp8MoEMethod(FusedMoEMethodBase):
@@ -399,8 +237,6 @@ class Fp8MoEMethod(FusedMoEMethodBase):
                        intermediate_size: int, params_dtype: torch.dtype,
                        **extra_weight_attrs):
 
-        layer.process_after_load = True
-
         if self.quant_config.is_checkpoint_fp8_serialized:
             params_dtype = torch.float8_e4m3fn
 
@@ -465,9 +301,6 @@ class Fp8MoEMethod(FusedMoEMethodBase):
             layer.a2_scale = None
 
     def process_weights_after_loading(self, layer: Module) -> None:
-        if (not hasattr(layer, "process_after_load")
-                or not layer.process_after_load):
-            return
 
         # If checkpoint is fp16, quantize in place.
         if not self.quant_config.is_checkpoint_fp8_serialized:
@@ -531,7 +364,7 @@ class Fp8MoEMethod(FusedMoEMethodBase):
                                                     shard_size, :],
                         layer.w13_scale[expert_id][shard_id])
                     layer.w13_weight[expert_id][
-                        start:start + shard_size, :] = per_tensor_quantize(
+                        start:start + shard_size, :], _ = ops.scaled_fp8_quant(
                             dq_weight, max_w13_scales[expert_id])
                     start += shard_size
 
@@ -596,23 +429,3 @@ class Fp8KVCacheMethod(QuantizeMethodBase):
                     "cause accuracy issues. Please make sure kv-cache scaling "
                     "factor is available in the fp8 checkpoint.")
         del layer.kv_scale
-
-
-def per_tensor_quantize(tensor: torch.Tensor,
-                        inv_scale: Union[float, torch.Tensor]) -> torch.Tensor:
-    finfo = torch.finfo(torch.float8_e4m3fn)
-    qweight = (tensor / inv_scale).clamp(min=finfo.min, max=finfo.max)
-    return qweight.to(torch.float8_e4m3fn)
-
-
-def per_tensor_dequantize(
-        tensor: torch.Tensor, inv_scale: Union[float,
-                                               torch.Tensor]) -> torch.Tensor:
-    fake_qweight = tensor.to(torch.float16)
-    dq_weight = fake_qweight * inv_scale
-    return dq_weight
-
-
-def all_close_1d(x: torch.Tensor) -> bool:
-    assert len(x.shape) == 1
-    return all(torch.allclose(x[0], x[i]) for i in range(x.shape[0]))

vllm/model_executor/layers/quantization/gptq_marlin.py

@@ -11,20 +11,16 @@ from vllm.model_executor.layers.linear import (LinearBase, LinearMethodBase,
                                                set_weight_attrs)
 from vllm.model_executor.layers.quantization.base_config import (
     QuantizationConfig)
+from vllm.model_executor.layers.quantization.utils.marlin_utils import (
+    GPTQ_MARLIN_MAX_PARALLEL, GPTQ_MARLIN_MIN_THREAD_K,
+    GPTQ_MARLIN_MIN_THREAD_N, GPTQ_MARLIN_SUPPORTED_GROUP_SIZES,
+    GPTQ_MARLIN_SUPPORTED_NUM_BITS, GPTQ_MARLIN_SUPPORTED_SYM,
+    GPTQ_MARLIN_TILE)
 from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
 from vllm.platforms import current_platform
 
 logger = init_logger(__name__)
 
-GPTQ_MARLIN_TILE = 16
-GPTQ_MARLIN_MIN_THREAD_N = 64
-GPTQ_MARLIN_MIN_THREAD_K = 128
-GPTQ_MARLIN_MAX_PARALLEL = 16
-
-GPTQ_MARLIN_SUPPORTED_NUM_BITS = [4, 8]
-GPTQ_MARLIN_SUPPORTED_GROUP_SIZES = [-1, 32, 64, 128]
-GPTQ_MARLIN_SUPPORTED_SYM = [True]
-
 
 # Permutations for Marlin scale shuffling
 def get_scale_perms(num_bits: int):

vllm/model_executor/layers/quantization/utils/marlin_utils.py

@@ -1,9 +1,11 @@
 """This file is used for /tests and /benchmarks"""
 import random
+from typing import Optional
 
 import numpy
 import torch
 
+from vllm import _custom_ops as ops
 from vllm.model_executor.layers.quantization.utils.format_24 import (
     mask_creator, sparse_semi_structured_from_dense_cutlass)
 from vllm.model_executor.layers.quantization.utils.marlin_24_perms import (
@@ -13,8 +15,16 @@ from vllm.model_executor.layers.quantization.utils.marlin_perms import (
 from vllm.model_executor.layers.quantization.utils.quant_utils import (
     get_pack_factor, quantize_weights, sort_weights)
 from vllm.platforms import current_platform
+from vllm.utils import print_warning_once
 
-MARLIN_TILE = 16
+GPTQ_MARLIN_TILE = 16
+GPTQ_MARLIN_MIN_THREAD_N = 64
+GPTQ_MARLIN_MIN_THREAD_K = 128
+GPTQ_MARLIN_MAX_PARALLEL = 16
+
+GPTQ_MARLIN_SUPPORTED_NUM_BITS = [4, 8]
+GPTQ_MARLIN_SUPPORTED_GROUP_SIZES = [-1, 32, 64, 128]
+GPTQ_MARLIN_SUPPORTED_SYM = [True]
 
 
 def is_marlin_supported():
@@ -22,7 +32,92 @@ def is_marlin_supported():
     return capability[0] >= 8
 
 
-def marlin_permute_weights(q_w, size_k, size_n, perm, tile=MARLIN_TILE):
+def apply_fp8_marlin_linear(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    weight_scale: torch.Tensor,
+    workspace: torch.Tensor,
+    size_n: int,
+    size_k: int,
+    bias: Optional[torch.Tensor],
+) -> torch.Tensor:
+    # For GPUs that lack FP8 hardware support, we can leverage the
+    # Marlin kernel for fast weight-only FP8 quantization
+
+    reshaped_x = input.reshape(-1, input.shape[-1])
+    out_shape = input.shape[:-1] + (size_n, )
+
+    output = ops.fp8_marlin_gemm(
+        a=reshaped_x,
+        b_q_weight=weight,
+        b_scales=weight_scale,
+        workspace=workspace,
+        num_bits=8,
+        size_m=reshaped_x.shape[0],
+        size_n=size_n,
+        size_k=size_k,
+    )
+
+    if bias is not None:
+        output.add_(bias)  # In-place add
+
+    return output.reshape(out_shape)
+
+
+def prepare_fp8_layer_for_marlin(layer: torch.nn.Module) -> None:
+    print_warning_once(
+        "Your GPU does not have native support for FP8 computation but "
+        "FP8 quantization is being used. Weight-only FP8 compression will "
+        "be used leveraging the Marlin kernel. This may degrade "
+        "performance for compute-heavy workloads.")
+
+    part_size_n = layer.output_size_per_partition
+    part_size_k = layer.input_size_per_partition
+
+    device = layer.weight.device
+
+    # WEIGHTS
+    # Repack weights to gptq format (packed int32 elements)
+    packed_gptq_qweight = pack_fp8_to_int32(layer.weight)
+
+    # Repack weights to marlin format
+    marlin_qweight = ops.gptq_marlin_repack(
+        b_q_weight=packed_gptq_qweight,
+        perm=torch.empty(0, dtype=torch.int, device=device),
+        size_k=part_size_k,
+        size_n=part_size_n,
+        num_bits=8,
+    )
+    layer.weight = torch.nn.Parameter(marlin_qweight, requires_grad=False)
+
+    # WEIGHT SCALES
+    # Currently Marlin doesn't support per-tensor scales, so we
+    # expand it to channelwise
+    scales = layer.weight_scale.repeat(1, part_size_n).to(
+        layer.orig_dtype).to(device)
+    # Permute scales
+    num_bits = 8
+    marlin_scales = marlin_permute_scales(
+        s=scales,
+        size_k=part_size_k,
+        size_n=part_size_n,
+        group_size=-1,
+        scale_perm=marlin_scale_perm[num_bits],
+        scale_perm_single=marlin_scale_perm_single[num_bits])
+    layer.weight_scale = torch.nn.Parameter(marlin_scales, requires_grad=False)
+
+    # Allocate marlin workspace
+    max_workspace_size = (part_size_n //
+                          GPTQ_MARLIN_MIN_THREAD_N) * GPTQ_MARLIN_MAX_PARALLEL
+    workspace = torch.zeros(max_workspace_size,
+                            dtype=torch.int,
+                            device=device,
+                            requires_grad=False)
+
+    layer.workspace = workspace
+
+
+def marlin_permute_weights(q_w, size_k, size_n, perm, tile=GPTQ_MARLIN_TILE):
     assert q_w.shape == (size_k, size_n)
     assert size_k % tile == 0, f"size_k = {size_k}, tile = {tile}"
     assert size_n % tile == 0, f"size_k = {size_n}, tile = {tile}"

vllm/model_executor/layers/quantization/utils/w8a8_utils.py

@@ -0,0 +1,163 @@
+from typing import List, Optional, Tuple, Union
+
+import torch
+from torch.nn import Parameter
+
+from vllm import _custom_ops as ops
+from vllm.model_executor.utils import set_weight_attrs
+from vllm.platforms import current_platform
+
+
+def cutlass_fp8_supported() -> bool:
+    capability = current_platform.get_device_capability()
+    capability = capability[0] * 10 + capability[1]
+
+    return ops.cutlass_scaled_mm_supports_fp8(capability)
+
+
+def per_tensor_dequantize(
+        tensor: torch.Tensor, inv_scale: Union[float,
+                                               torch.Tensor]) -> torch.Tensor:
+    fake_qweight = tensor.to(torch.float16)
+    dq_weight = fake_qweight * inv_scale
+    return dq_weight
+
+
+def all_close_1d(x: torch.Tensor) -> bool:
+    assert len(x.shape) == 1
+    return all(torch.allclose(x[0], x[i]) for i in range(x.shape[0]))
+
+
+def create_per_tensor_scale_param(
+    output_partition_sizes: List[int],
+    **extra_weight_attrs,
+) -> Parameter:
+    scale = Parameter(torch.empty(len(output_partition_sizes),
+                                  dtype=torch.float32),
+                      requires_grad=False)
+    scale[:] = torch.finfo(torch.float32).min
+    set_weight_attrs(scale, {
+        "needs_scalar_to_array": True,
+        **extra_weight_attrs
+    })
+    return scale
+
+
+def create_per_channel_scale_param(output_partition_sizes: List[int],
+                                   **extra_weight_attrs) -> Parameter:
+    scale = Parameter(torch.empty((sum(output_partition_sizes), 1),
+                                  dtype=torch.float32),
+                      requires_grad=False)
+    scale[:] = torch.finfo(torch.float32).min
+    set_weight_attrs(scale, {"output_dim": 0, **extra_weight_attrs})
+    return scale
+
+
+def convert_to_channelwise(
+        weight_scale: torch.Tensor,
+        logical_widths: List[int]) -> Tuple[torch.Tensor, torch.Tensor]:
+    # Create channelwise buffer
+    weight_scale_channel = torch.empty((sum(logical_widths), 1),
+                                       dtype=torch.float32,
+                                       device=weight_scale.device)
+
+    # Expand each scale to match the size of each logical matrix.
+    start = 0
+    for idx, logical_width in enumerate(logical_widths):
+        end = start + logical_width
+        weight_scale_channel[start:end, :] = weight_scale[idx]
+        start = end
+
+    return weight_scale_channel
+
+
+def requantize_with_max_scale(
+        weight: torch.Tensor, weight_scale: torch.Tensor,
+        logical_widths: List[int]) -> Tuple[torch.Tensor, torch.Tensor]:
+    # Max scale to be used for requanitzation.
+    max_w_scale = weight_scale.max()
+
+    # QKV / MLP is fused in the on disk checkpoint if any of the
+    # weight scales are still set to the default since we initialize
+    # N weight scales for N shards but we only load 1 weight scale
+    # from disk in this case. Skip requantization in this case (since)
+    # we already are quantized with the single scale.
+    # * Sample Model: nm-testing/Phi-3-mini-128k-instruct-FP8
+    unfused_module_in_checkpoint = (weight_scale[-1] > torch.finfo(
+        torch.float8_e4m3fn).min)
+
+    # If unfused checkpoint, need requanize with the single scale.
+    if unfused_module_in_checkpoint:
+        start = 0
+        for idx, logical_width in enumerate(logical_widths):
+            end = start + logical_width
+            weight_dq = per_tensor_dequantize(weight[start:end, :],
+                                              weight_scale[idx])
+            weight[start:end, :], _ = ops.scaled_fp8_quant(
+                weight_dq, max_w_scale)
+            start = end
+
+    return max_w_scale, weight
+
+
+def apply_fp8_linear(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    weight_scale: torch.Tensor,
+    input_scale: torch.Tensor,
+    bias: Optional[torch.Tensor] = None,
+    cutlass_fp8_supported: bool = True,
+) -> torch.Tensor:
+    # ops.scaled_fp8_quant supports both dynamic and static quant.
+    #   If dynamic, layer.input_scale is None and x_scale computed from x.
+    #   If static, layer.input_scale is scalar and x_scale is input_scale.
+
+    if bias is None and cutlass_fp8_supported:
+        qinput, x_scale = ops.scaled_fp8_quant(input, input_scale)
+
+        # Fused GEMM_DQ
+        output = ops.cutlass_scaled_mm(qinput,
+                                       weight,
+                                       out_dtype=input.dtype,
+                                       scale_a=x_scale,
+                                       scale_b=weight_scale)
+
+    else:
+        qinput, x_scale = ops.scaled_fp8_quant(input,
+                                               input_scale,
+                                               batch_dim_padding=17)
+
+        # Fused GEMM_DQ -- note we padded the input above because
+        # torch._scaled_mm is more performant for matrices with
+        # batch dimension > 16. Note that this could change
+        # in the future.
+        output, _ = torch._scaled_mm(qinput,
+                                     weight,
+                                     out_dtype=input.dtype,
+                                     scale_a=x_scale,
+                                     scale_b=weight_scale,
+                                     bias=bias)
+
+    return torch.narrow(output, 0, 0, input.shape[0])
+
+
+def apply_int8_linear(
+    input: torch.Tensor,
+    weight: torch.Tensor,
+    weight_scale: torch.Tensor,
+    input_scale: torch.Tensor,
+    bias: Optional[torch.Tensor] = None,
+):
+    if bias is not None:
+        raise NotImplementedError("W8A8 with int8 does not yet support bias.")
+
+    # ops.scaled_int8_quant supports both dynamic and static quant.
+    # * dynamic, layer.input_scale is None and x_scale computed from x.
+    # * static, layer.input_scale is scalar and x_scale is input_scale.
+    x_q, x_scale = ops.scaled_int8_quant(input, input_scale)
+
+    return ops.cutlass_scaled_mm(x_q,
+                                 weight,
+                                 scale_a=x_scale,
+                                 scale_b=weight_scale,
+                                 out_dtype=input.dtype)