[Misc] clean up vllm in sgl-kernel test (#5189)

sgl-kernel/tests/test_awq_dequant.py

@@ -4,7 +4,6 @@ from typing import Optional, Tuple
 import pytest
 import torch
 from sgl_kernel import awq_dequantize
-from vllm import _custom_ops as ops
 
 
 def reverse_awq_order(t: torch.Tensor):
@@ -58,12 +57,6 @@ def awq_dequantize_torch(
     return (iweights - zeros) * scales
 
 
-def vllm_awq_dequantize(
-    qweight: torch.Tensor, scales: torch.Tensor, qzeros: torch.Tensor
-) -> torch.Tensor:
-    return ops.awq_dequantize(qweight, scales, qzeros, 0, 0, 0)
-
-
 def sglang_awq_dequantize(
     qweight: torch.Tensor, scales: torch.Tensor, qzeros: torch.Tensor
 ) -> torch.Tensor:
@@ -110,7 +103,6 @@ def test_awq_dequant_compare_implementations(
     )
 
     # Run both implementations
-    vllm_out = vllm_awq_dequantize(qweight, scales.to(torch.float16), qzeros)
     torch_out = awq_dequantize_torch(qweight, scales, qzeros, group_size)
     sglang_out = sglang_awq_dequantize(qweight, scales, qzeros)
 
@@ -118,13 +110,6 @@ def test_awq_dequant_compare_implementations(
     torch.testing.assert_close(
         torch_out.to(torch.float32), sglang_out.to(torch.float32), rtol=1e-3, atol=1e-5
     )
-    if not is_bf16_act:
-        torch.testing.assert_close(
-            vllm_out.to(torch.float32),
-            sglang_out.to(torch.float32),
-            rtol=1e-3,
-            atol=1e-5,
-        )
 
 
 if __name__ == "__main__":

sgl-kernel/tests/test_int8_gemm.py

@@ -1,7 +1,6 @@
 import pytest
 import torch
 from sgl_kernel import int8_scaled_mm
-from vllm._custom_ops import cutlass_scaled_mm as vllm_scaled_mm
 
 
 def to_int8(tensor: torch.Tensor) -> torch.Tensor:
@@ -28,9 +27,7 @@ def _test_accuracy_once(M, N, K, with_bias, out_dtype, device):
         bias = None
     o = int8_scaled_mm(a, b, scale_a, scale_b, out_dtype, bias)
     o1 = torch_scaled_mm(a, b, scale_a, scale_b, out_dtype, bias)
-    o2 = vllm_scaled_mm(a, b, scale_a, scale_b, out_dtype, bias)
     torch.testing.assert_close(o, o1)
-    torch.testing.assert_close(o, o2)
     print(f"M={M}, N={N}, K={K}, with_bias={with_bias}, out_dtype={out_dtype}: OK")
 
 

sgl-kernel/tests/test_per_tensor_quant_fp8.py

@@ -4,7 +4,6 @@ from typing import Optional, Tuple
 import pytest
 import torch
 from sgl_kernel import sgl_per_tensor_quant_fp8
-from vllm import _custom_ops as ops
 
 from sglang.srt.utils import is_hip
 
@@ -12,13 +11,6 @@ is_hip_ = is_hip()
 fp8_type_ = torch.float8_e4m3fnuz if is_hip_ else torch.float8_e4m3fn
 
 
-def vllm_scaled_fp8_quant(
-    input: torch.Tensor,
-    scale: Optional[torch.Tensor] = None,
-) -> Tuple[torch.Tensor, torch.Tensor]:
-    return ops.scaled_fp8_quant(input, scale)
-
-
 def sglang_scaled_fp8_quant(
     input: torch.Tensor,
     scale: Optional[torch.Tensor] = None,
@@ -34,6 +26,16 @@ def sglang_scaled_fp8_quant(
     return output, scale
 
 
+def torch_scaled_fp8_quant(tensor, inv_scale):
+    # The reference implementation that fully aligns to
+    # the kernel being tested.
+    finfo = torch.finfo(torch.float8_e4m3fn)
+    scale = inv_scale.reciprocal()
+    qweight = (tensor.to(torch.float32) * scale).clamp(min=finfo.min, max=finfo.max)
+    qweight = qweight.to(torch.float8_e4m3fn)
+    return qweight
+
+
 @pytest.mark.parametrize(
     "num_tokens,hidden_dim",
     list(itertools.product([128, 256, 512], [512, 2048, 4096])),
@@ -45,21 +47,19 @@ def test_per_tensor_quant_compare_implementations(
     device = torch.device("cuda")
     x = torch.rand((num_tokens, hidden_dim), dtype=torch.float16, device=device)
 
-    vllm_out, vllm_scale = vllm_scaled_fp8_quant(x)
     sglang_out, sglang_scale = sglang_scaled_fp8_quant(x)
+    torch_out = torch_scaled_fp8_quant(x, sglang_scale)
 
-    torch.testing.assert_close(vllm_scale, sglang_scale, rtol=1e-3, atol=1e-3)
     torch.testing.assert_close(
-        vllm_out.float(), sglang_out.float(), rtol=1e-3, atol=1e-3
+        sglang_out.float(), torch_out.float(), rtol=1e-3, atol=1e-3
     )
 
     scale = torch.rand(1, dtype=torch.float32, device=device)
-    vllm_out, vllm_scale = vllm_scaled_fp8_quant(x, scale)
     sglang_out, sglang_scale = sglang_scaled_fp8_quant(x, scale)
+    torch_out = torch_scaled_fp8_quant(x, scale)
 
-    torch.testing.assert_close(vllm_scale, sglang_scale, rtol=1e-3, atol=1e-3)
     torch.testing.assert_close(
-        vllm_out.float(), sglang_out.float(), rtol=1e-3, atol=1e-3
+        sglang_out.float(), torch_out.float(), rtol=1e-3, atol=1e-3
     )
 
 

sgl-kernel/tests/test_per_token_quant_fp8.py

@@ -4,7 +4,6 @@ from typing import Optional, Tuple
 import pytest
 import torch
 from sgl_kernel import sgl_per_token_quant_fp8
-from vllm import _custom_ops as ops
 
 from sglang.srt.utils import is_hip
 
@@ -12,10 +11,15 @@ is_hip_ = is_hip()
 fp8_type_ = torch.float8_e4m3fnuz if is_hip_ else torch.float8_e4m3fn
 
 
-def vllm_per_token_quant_fp8(
-    input: torch.Tensor,
-) -> Tuple[torch.Tensor, torch.Tensor]:
-    return ops.scaled_fp8_quant(input, use_per_token_if_dynamic=True)
+def torch_per_token_quant_fp8(tensor, inv_scale):
+    # The reference implementation that fully aligns to
+    # the kernel being tested.
+    finfo = torch.finfo(torch.float8_e4m3fn)
+    inv_scale = inv_scale.view(-1, 1)
+    scale = inv_scale.reciprocal()
+    qweight = (tensor.to(torch.float32) * scale).clamp(min=finfo.min, max=finfo.max)
+    qweight = qweight.to(torch.float8_e4m3fn)
+    return qweight
 
 
 def sglang_per_token_quant_fp8(
@@ -41,12 +45,11 @@ def test_per_token_quant_compare_implementations(
     device = torch.device("cuda")
     x = torch.rand((num_tokens, hidden_dim), dtype=torch.float16, device=device)
 
-    vllm_out, vllm_scale = vllm_per_token_quant_fp8(x)
     sglang_out, sglang_scale = sglang_per_token_quant_fp8(x)
+    torch_out = torch_per_token_quant_fp8(x, sglang_scale)
 
-    torch.testing.assert_close(vllm_scale, sglang_scale, rtol=1e-3, atol=1e-3)
     torch.testing.assert_close(
-        vllm_out.float(), sglang_out.float(), rtol=1e-3, atol=1e-3
+        sglang_out.float(), torch_out.float(), rtol=1e-3, atol=1e-3
     )