[Kernel][Triton] Add Triton implementation for scaled_mm_triton to support fp8 and int8 SmoothQuant, symmetric case

tests/kernels/test_scaled_mm_triton.py

@@ -0,0 +1,96 @@
+"""Tests for the scaled_mm_triton kernel
+
+Run `pytest tests/kernels/test_scaled_mm_triton.py`.
+"""
+import importlib
+from typing import Optional, Type
+
+import pytest
+import torch
+
+from vllm.utils import seed_everything
+
+device = "cuda"
+
+
+def scaled_mm_torch(a: torch.Tensor,
+                    b: torch.Tensor,
+                    scale_a: torch.Tensor,
+                    scale_b: torch.Tensor,
+                    out_dtype: Type[torch.dtype],
+                    bias: Optional[torch.Tensor] = None) -> torch.Tensor:
+    out = torch.mm(a.to(torch.float32), b.to(torch.float32))
+    out = scale_a * out
+    out = scale_b.T * out
+    out = out.to(out_dtype)
+    if bias is not None:
+        out = out + bias
+
+    return out
+
+
+@pytest.mark.parametrize("M", [1, 16, 32, 64, 128, 256, 512, 222, 33, 1])
+@pytest.mark.parametrize("N", [2048, 8192, 16384, 256, 1024])
+@pytest.mark.parametrize("K", [128, 496, 1024])
+@pytest.mark.parametrize("out_dtype", [torch.float16, torch.bfloat16])
+@pytest.mark.parametrize("in_dtype", [torch.int8])
+@pytest.mark.parametrize("use_scalar_scale_a", [True, False])
+@pytest.mark.parametrize("use_scalar_scale_b", [True, False])
+@pytest.mark.parametrize("use_bias", [True, False])
+def test_scaled_mm(M, N, K, in_dtype, out_dtype, use_scalar_scale_a,
+                   use_scalar_scale_b, use_bias):
+    is_floating_point_type = lambda t: torch.tensor([1, 1], dtype=t
+                                                    ).is_floating_point()
+
+    seed_everything(0)
+
+    # NOTE: There are cases, where if the matrix is large enough, an output
+    # like 65504.4 can be produced, and can easily turn into inf when
+    # multiplied when using float16/bfloat16.  This means one function, e.g.,
+    # testing function, and another function, e.g. golden function, can
+    # produce a non-inf value while the other produces an inf value, and
+    # will cause assert_close/allclose to fail, even though if overflow
+    # wouldn't have occurred, the values would have been "close."
+    #
+    # So, the values here are kept small enough to avoid this situation.
+    if is_floating_point_type(in_dtype):
+        a = (0.25 * torch.rand(
+            (M, K), dtype=torch.float32, device=device)).to(in_dtype)
+        b = (0.25 * torch.rand(
+            (K, N), dtype=torch.float32, device=device)).to(in_dtype)
+    else:
+        a = torch.randint(-32, 32, (M, K), dtype=in_dtype, device=device)
+        b = torch.randint(-32, 32, (K, N), dtype=in_dtype, device=device)
+
+    if use_scalar_scale_a:
+        scale_a = torch.rand((1, 1), device=device)
+    else:
+        scale_a = 0.25 * torch.rand((M, 1), device=device)
+
+    if use_scalar_scale_b:
+        scale_b = torch.rand((1, 1), device=device)
+    else:
+        scale_b = 0.25 * torch.rand((1, 1), device=device)
+
+    bias = None
+    if use_bias:
+        bias = torch.rand((N, ), device=device, dtype=out_dtype)
+
+    scaled_mm_triton_module = importlib.import_module(
+        "vllm.model_executor.layers.quantization.compressed_tensors."
+        "scaled_mm_triton")
+    scaled_mm_triton = scaled_mm_triton_module.scaled_mm_triton
+
+    c_check = scaled_mm_triton(a, b, scale_a, scale_b, out_dtype, bias)
+
+    a_cpu = a.cpu()
+    b_cpu = b.cpu()
+    scale_a_cpu = scale_a.cpu()
+    scale_b_cpu = scale_b.cpu()
+    bias_cpu = None if bias is None else bias.cpu()
+
+    c_actual = scaled_mm_torch(a_cpu, b_cpu, scale_a_cpu, scale_b_cpu,
+                               out_dtype, bias_cpu)
+
+    c_check_cpu = c_check.cpu()
+    torch.testing.assert_close(c_check_cpu, c_actual, rtol=1e-1, atol=1e-1)

vllm/_custom_ops.py

@@ -1,5 +1,6 @@
 import contextlib
 import functools
+import importlib
 from typing import TYPE_CHECKING, List, Optional, Tuple, Union
 
 import torch
@@ -486,6 +487,14 @@ def cutlass_scaled_mm(a: torch.Tensor,
 
     m = a.shape[0]
     n = b.shape[1]
+
+    if current_platform.is_rocm():
+        scaled_mm_triton_module = importlib.import_module(
+            "vllm.model_executor.layers.quantization.compressed_tensors."
+            "scaled_mm_triton")
+        scaled_mm_triton = scaled_mm_triton_module.scaled_mm_triton
+        return scaled_mm_triton(a, b, scale_a, scale_b, out_dtype, bias)
+
     out = torch.empty((m, n), dtype=out_dtype, device=a.device)
 
     torch.ops._C.cutlass_scaled_mm(out, a, b, scale_a, scale_b, bias)

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

@@ -0,0 +1,207 @@
+from typing import Optional, Type
+
+import torch
+import triton
+import triton.language as tl
+
+
+# This function handles some cases that can cause certain failure, e.g.
+# a tensor that has shape = (72, 48) but stride = (5120, 1).  It can happen,
+# for example by saving a tensor using torch.save() and then adjusting its
+# size afterwards and then trying to use it.  Unfortunately,
+# torch.is_contiguous() doesn't help since a transposed tensor doesn't return
+# True, even though it can be stored contiguously in memory.
+#
+# There is a way to handle this case, which I learned about from here:
+#
+# https://github.com/pytorch/pytorch/blob/
+# a874ec85e83cfe75e7238296022d53d7e20860df/aten/src/ATen/native/
+# cuda/Blas.cpp#L58
+#
+# This doesn't happen very often fortunately, because the only solution is
+# inefficient.
+def prepare_matrix_for_triton(x: torch.Tensor):
+    strides = x.stride()
+    sizes = x.shape
+    is_not_transpose = strides[0] == 1 and (strides[1] >= max(1, sizes[0]))
+    is_transpose = strides[1] == 1 and (strides[0] >= max(1, sizes[1]))
+    if not is_not_transpose and not is_transpose:
+        return torch.clone(x, memory_format=torch.contiguous_format)
+    return x
+
+
+@triton.jit
+def scaled_mm_kernel(a_ptr, b_ptr, scale_a_ptr, scale_b_ptr, c_ptr, bias_ptr,
+                     M, N, K, stride_am, stride_ak, stride_bk, stride_bn,
+                     stride_cm, stride_cn, ACCUMULATOR_DTYPE: tl.constexpr,
+                     BLOCK_SIZE_M: tl.constexpr, BLOCK_SIZE_N: tl.constexpr,
+                     BLOCK_SIZE_K: tl.constexpr,
+                     BLOCK_SIZE_SCALE_A: tl.constexpr,
+                     BLOCK_SIZE_SCALE_B: tl.constexpr):
+    pid = tl.program_id(axis=0)
+
+    num_pid_n = tl.cdiv(N, BLOCK_SIZE_N)
+
+    pid_m = pid // num_pid_n
+    pid_n = pid % num_pid_n
+
+    accumulator_dtype = ACCUMULATOR_DTYPE
+    accumulator = tl.zeros((BLOCK_SIZE_M, BLOCK_SIZE_N),
+                           dtype=accumulator_dtype)
+
+    # NOTE: Some tensor inputs are so large, they will cause int32 overflow
+    # so it is necessary to use tl.int64 for all the offsets, else SEGV will
+    # eventually occur.
+
+    # Offsets and masks.
+    offsets_am = pid_m * BLOCK_SIZE_M + tl.arange(0, BLOCK_SIZE_M).to(tl.int64)
+    masks_am = offsets_am < M
+
+    offsets_bn = pid_n * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N).to(tl.int64)
+    masks_bn = offsets_bn < N
+
+    offsets_k = tl.arange(0, BLOCK_SIZE_K).to(tl.int64)
+    offsets_a = (stride_am * offsets_am[:, None] +
+                 stride_ak * offsets_k[None, :])
+    offsets_b = (stride_bk * offsets_k[:, None] +
+                 stride_bn * offsets_bn[None, :])
+
+    # NOTE: BLOCK_SIZE_SCALE_A could be 1 or BLOCK_SIZE_M, so need to create
+    # appropriate offsets and masks for each case. Same goes for
+    # BLOCK_SIZE_SCALE_B.
+    offsets_scale_am = (tl.arange(0, BLOCK_SIZE_SCALE_A) +
+                        (BLOCK_SIZE_SCALE_A > 1) * pid_m * BLOCK_SIZE_M)
+    masks_scale_am = offsets_scale_am < M
+
+    offsets_scale_bn = (tl.arange(0, BLOCK_SIZE_SCALE_B) +
+                        (BLOCK_SIZE_SCALE_B > 1) * pid_n * BLOCK_SIZE_N)
+    masks_scale_bn = offsets_scale_bn < N
+
+    offsets_scale_a = (offsets_scale_am[:, None].to(tl.int64) +
+                       tl.arange(0, 1)[None, :].to(tl.int64))
+    offsets_scale_b = (offsets_scale_bn[:, None].to(tl.int64) +
+                       tl.arange(0, 1)[None, :].to(tl.int64))
+
+    a_ptrs = a_ptr + offsets_a
+    b_ptrs = b_ptr + offsets_b
+
+    scale_a_ptrs = scale_a_ptr + offsets_scale_a
+    scale_b_ptrs = scale_b_ptr + offsets_scale_b
+
+    for k in range(0, tl.cdiv(K, BLOCK_SIZE_K)):
+        masks_k = offsets_k < K
+        masks_a = masks_am[:, None] & masks_k[None, :]
+        a = tl.load(a_ptrs, mask=masks_a)
+
+        masks_b = masks_k[:, None] & masks_bn[None, :]
+        b = tl.load(b_ptrs, mask=masks_b)
+
+        # Accumulate results.
+        accumulator = tl.dot(a, b, accumulator, out_dtype=accumulator_dtype)
+
+        offsets_k += BLOCK_SIZE_K
+        a_ptrs += BLOCK_SIZE_K * stride_ak
+        b_ptrs += BLOCK_SIZE_K * stride_bk
+
+    # Apply scale at end.
+    masks_scale_a = masks_scale_am[:, None] & (tl.arange(0, 1) < 1)[:, None]
+    scale_a = tl.load(scale_a_ptrs, masks_scale_a)
+    # Need to broadcast to the appropriate size, if scale_a is already
+    # (BLOCK_SIZE_M, 1) then it will broadcast to its own shape. Same goes
+    # for scale_b below.
+    scale_a = scale_a.broadcast_to((BLOCK_SIZE_M, 1))
+    accumulator = scale_a * accumulator.to(tl.float32)
+
+    masks_scale_b = masks_scale_bn[:, None] & (tl.arange(0, 1) < 1)[None, :]
+    scale_b = tl.load(scale_b_ptrs, masks_scale_b)
+    scale_b = scale_b.broadcast_to((BLOCK_SIZE_N, 1))
+    accumulator = scale_b.T * accumulator.to(tl.float32)
+
+    # Convert to output format.
+    c = accumulator.to(c_ptr.type.element_ty)
+
+    # Add bias, it's already in output format, so add it after conversion.
+    if bias_ptr:
+        offsets_bias = offsets_bn
+        bias_ptrs = bias_ptr + offsets_bias
+        bias_mask = offsets_bias < N
+        bias = tl.load(bias_ptrs, bias_mask)
+        c += bias
+
+    # Save output
+    offs_cm = pid_m * BLOCK_SIZE_M + tl.arange(0, BLOCK_SIZE_M).to(tl.int64)
+    offs_cn = pid_n * BLOCK_SIZE_N + tl.arange(0, BLOCK_SIZE_N).to(tl.int64)
+    offs_cm = offs_cm.to(tl.int64)
+    offs_cn = offs_cn.to(tl.int64)
+    c_ptrs = (c_ptr + stride_cm * offs_cm[:, None] +
+              stride_cn * offs_cn[None, :])
+    c_mask = (offs_cm[:, None] < M) & (offs_cn[None, :] < N)
+
+    tl.store(c_ptrs, c, mask=c_mask)
+
+
+# input   - [M, K]
+# weight - [K, N]
+def scaled_mm_triton(input: torch.Tensor,
+                     weight: torch.Tensor,
+                     scale_a: torch.Tensor,
+                     scale_b: torch.Tensor,
+                     out_dtype: Type[torch.dtype],
+                     bias: Optional[torch.Tensor] = None,
+                     block_size_m: int = 32,
+                     block_size_n: int = 32,
+                     block_size_k: int = 32) -> torch.Tensor:
+    M, K = input.shape
+    N = weight.shape[1]
+
+    assert N > 0 and K > 0 and M > 0
+    assert weight.shape[0] == K
+    assert input.dtype == weight.dtype
+    assert scale_a.dtype == scale_b.dtype and scale_a.is_floating_point()
+    assert scale_a.shape == torch.Size([1, 1]) or scale_a.shape == torch.Size(
+        [M, 1])
+    assert scale_b.shape == torch.Size([1, 1]) or scale_b.shape == torch.Size(
+        [N, 1])
+    assert torch.empty((1, 1), dtype=out_dtype).is_floating_point()
+    assert bias is None or bias.is_floating_point()
+
+    grid = lambda META: (triton.cdiv(M, META['BLOCK_SIZE_M']) * triton.cdiv(
+        N, META['BLOCK_SIZE_N']), )
+
+    result = torch.empty((M, N), dtype=out_dtype, device=input.device)
+
+    has_scalar = lambda x: x.shape[0] == 1 and x.shape[1] == 1
+
+    block_size_sa = 1 if has_scalar(scale_a) else block_size_m
+    block_size_sb = 1 if has_scalar(scale_b) else block_size_n
+
+    input = prepare_matrix_for_triton(input)
+    weight = prepare_matrix_for_triton(weight)
+
+    accumulator_dtype = tl.float32 if input.is_floating_point() else tl.int32
+
+    # A = input, B = weight, C = result
+    # A = M x K, B = K x N, C = M x N
+    scaled_mm_kernel[grid](input,
+                           weight,
+                           scale_a,
+                           scale_b,
+                           result,
+                           bias,
+                           M,
+                           N,
+                           K,
+                           input.stride(0),
+                           input.stride(1),
+                           weight.stride(0),
+                           weight.stride(1),
+                           result.stride(0),
+                           result.stride(1),
+                           accumulator_dtype,
+                           BLOCK_SIZE_M=block_size_m,
+                           BLOCK_SIZE_N=block_size_n,
+                           BLOCK_SIZE_K=block_size_k,
+                           BLOCK_SIZE_SCALE_A=block_size_sa,
+                           BLOCK_SIZE_SCALE_B=block_size_sb)
+
+    return result.to(out_dtype)