[Quantization][FP8] Adding support for fp8 gemm layer input in fp8 (#14578)

Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>

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

@@ -23,6 +23,7 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
 
     def __init__(self, strategy: str, is_static_input_scheme: bool):
         self.strategy = strategy
+        self.out_dtype = torch.get_default_dtype()
         self.is_static_input_scheme = is_static_input_scheme
         self.fp8_linear = Fp8LinearOp(use_per_token_if_dynamic=True)
 
@@ -143,5 +144,6 @@ class CompressedTensorsW8A8Fp8(CompressedTensorsScheme):
         return self.fp8_linear.apply(input=x,
                                      weight=layer.weight,
                                      weight_scale=layer.weight_scale,
+                                     out_dtype=self.out_dtype,
                                      input_scale=layer.input_scale,
                                      bias=bias)

vllm/model_executor/layers/quantization/fbgemm_fp8.py

@@ -73,6 +73,7 @@ class FBGEMMFp8LinearMethod(LinearMethodBase):
     def __init__(self, quant_config: FBGEMMFp8Config):
         self.quant_config = quant_config
         self.fp8_linear = Fp8LinearOp(use_per_token_if_dynamic=True)
+        self.out_dtype = torch.get_default_dtype()
 
     def create_weights(
         self,
@@ -161,6 +162,7 @@ class FBGEMMFp8LinearMethod(LinearMethodBase):
         return self.fp8_linear.apply(input=x,
                                      weight=layer.weight,
                                      weight_scale=layer.weight_scale,
+                                     out_dtype=self.out_dtype,
                                      input_scale=None,
                                      input_scale_ub=layer.input_scale_ub,
                                      bias=bias)

vllm/model_executor/layers/quantization/fp8.py

@@ -116,6 +116,21 @@ class Fp8Config(QuantizationConfig):
             return Fp8KVCacheMethod(self)
         return None
 
+    def get_cache_scale(self, name: str) -> Optional[str]:
+        """
+        Check whether the param name matches the format for k/v cache scales
+        in compressed-tensors. If this is the case, return its equivalent
+        param name expected by vLLM
+
+        :param name: param name
+        :return: matching param name for KV cache scale in vLLM
+        """
+        if name.endswith(".output_scale") and ".k_proj" in name:
+            return name.replace(".k_proj.output_scale", ".attn.k_scale")
+        if name.endswith(".output_scale") and ".v_proj" in name:
+            return name.replace(".v_proj.output_scale", ".attn.v_scale")
+        return None
+
 
 class Fp8LinearMethod(LinearMethodBase):
     """Linear method for FP8.
@@ -138,6 +153,7 @@ class Fp8LinearMethod(LinearMethodBase):
     def __init__(self, quant_config: Fp8Config):
         self.quant_config = quant_config
         self.cutlass_block_fp8_supported = cutlass_block_fp8_supported()
+        self.out_dtype = torch.get_default_dtype()
 
         # For GPUs that lack FP8 hardware support, we can leverage the Marlin
         # kernel for fast weight-only FP8 quantization
@@ -386,6 +402,7 @@ class Fp8LinearMethod(LinearMethodBase):
         return self.fp8_linear.apply(input=x,
                                      weight=layer.weight,
                                      weight_scale=layer.weight_scale,
+                                     out_dtype=self.out_dtype,
                                      input_scale=layer.input_scale,
                                      bias=bias)
 

vllm/model_executor/layers/quantization/quark/schemes/quark_w8a8_fp8.py

@@ -22,6 +22,7 @@ class QuarkW8A8Fp8(QuarkScheme):
         self.qscheme = qscheme
         self.is_static_input_scheme = is_static_input_scheme
         self.fp8_linear = Fp8LinearOp(use_per_token_if_dynamic=True)
+        self.out_dtype = torch.get_default_dtype()
 
     @classmethod
     def get_min_capability(cls) -> int:
@@ -134,5 +135,6 @@ class QuarkW8A8Fp8(QuarkScheme):
         return self.fp8_linear.apply(input=x,
                                      weight=layer.weight,
                                      weight_scale=layer.weight_scale,
+                                     out_dtype=self.out_dtype,
                                      input_scale=layer.input_scale,
                                      bias=bias)

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

@@ -163,6 +163,7 @@ class Fp8LinearOp:
         input: torch.Tensor,
         weight: torch.Tensor,
         weight_scale: torch.Tensor,
+        out_dtype: Optional[torch.dtype] = None,
         input_scale: Optional[torch.Tensor] = None,
         input_scale_ub: Optional[torch.Tensor] = None,
         bias: Optional[torch.Tensor] = None,
@@ -182,8 +183,13 @@ class Fp8LinearOp:
         if use_per_token_if_dynamic is None:
             use_per_token_if_dynamic = self.use_per_token_if_dynamic
 
+        if out_dtype is None:
+            out_dtype = input.dtype
+
         # cutlass_scaled_mm supports per tensor/channel W and per tensor/token A
         if self.cutlass_fp8_supported:
+            assert input.dtype != current_platform.fp8_dtype(
+            ), "FP8 input to cutlass is not currently implemented"
             qinput, x_scale = ops.scaled_fp8_quant(
                 input_2d,
                 input_scale,
@@ -193,7 +199,7 @@ class Fp8LinearOp:
             # Fused GEMM_DQ
             output = ops.cutlass_scaled_mm(qinput,
                                            weight,
-                                           out_dtype=input.dtype,
+                                           out_dtype=out_dtype,
                                            scale_a=x_scale,
                                            scale_b=weight_scale,
                                            bias=bias)
@@ -202,12 +208,15 @@ class Fp8LinearOp:
         # torch.scaled_mm supports per tensor weights + activations only
         # so fallback to naive if per channel or per token
         else:
-            # Maybe apply padding to output, see comment in __init__
-            qinput, x_scale = ops.scaled_fp8_quant(
-                input_2d,
-                input_scale,
-                num_token_padding=self.output_padding,
-                use_per_token_if_dynamic=use_per_token_if_dynamic)
+            if input.dtype != current_platform.fp8_dtype():
+                # Maybe apply padding to output, see comment in __init__
+                qinput, x_scale = ops.scaled_fp8_quant(
+                    input_2d,
+                    input_scale,
+                    num_token_padding=self.output_padding,
+                    use_per_token_if_dynamic=use_per_token_if_dynamic)
+            else:
+                qinput, x_scale = input_2d, input_scale
 
             per_tensor_weights = (weight_scale.numel() == 1)
             per_tensor_activations = (x_scale.numel() == 1)
@@ -216,7 +225,7 @@ class Fp8LinearOp:
                 # Fused GEMM_DQ
                 output = torch._scaled_mm(qinput,
                                           weight,
-                                          out_dtype=input.dtype,
+                                          out_dtype=out_dtype,
                                           scale_a=x_scale,
                                           scale_b=weight_scale,
                                           bias=bias)
@@ -240,7 +249,7 @@ class Fp8LinearOp:
                 # Fused GEMM_DQ Rowwise GEMM
                 output = torch._scaled_mm(qinput,
                                           weight,
-                                          out_dtype=input.dtype,
+                                          out_dtype=out_dtype,
                                           scale_a=x_scale,
                                           scale_b=weight_scale.t(),
                                           bias=bias)