DeepSeek/DeepGEMM
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepGEMM synced 2025-05-09 12:39:21 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

pytest Integration

Browse Source
This commit is contained in:
A-transformer 2025-02-27 11:43:23 +04:00 committed by GitHub
parent 58046b4e01
commit f9a6da9ac2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 90 additions and 93 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

143
tests/test_core.py
View File

@ -1,10 +1,16 @@
import pytest
import random import random
import torch import torch
from typing import Tuple from typing import Tuple
import deep_gemm import deep_gemm
from deep_gemm import bench_kineto, calc_diff, cell_div, get_col_major_tma_aligned_tensor from deep_gemm import bench_kineto, calc_diff, ceil_div, get_col_major_tma_aligned_tensor
# Set seeds and TF32 flags up front
torch.backends.cuda.matmul.allow_tf32 = True
torch.backends.cudnn.allow_tf32 = True
torch.manual_seed(0)
random.seed(0)
def per_token_cast_to_fp8(x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]: def per_token_cast_to_fp8(x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
assert x.dim() == 2 and x.size(1) % 128 == 0 assert x.dim() == 2 and x.size(1) % 128 == 0
@ -13,18 +19,16 @@ def per_token_cast_to_fp8(x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
x_amax = x_view.abs().float().amax(dim=2).view(m, -1).clamp(1e-4) x_amax = x_view.abs().float().amax(dim=2).view(m, -1).clamp(1e-4)
return (x_view * (448.0 / x_amax.unsqueeze(2))).to(torch.float8_e4m3fn).view(m, n), (x_amax / 448.0).view(m, -1) return (x_view * (448.0 / x_amax.unsqueeze(2))).to(torch.float8_e4m3fn).view(m, n), (x_amax / 448.0).view(m, -1)
def per_block_cast_to_fp8(x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]: def per_block_cast_to_fp8(x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
assert x.dim() == 2 assert x.dim() == 2
m, n = x.shape m, n = x.shape
x_padded = torch.zeros((cell_div(m, 128) * 128, cell_div(n, 128) * 128), dtype=x.dtype, device=x.device) x_padded = torch.zeros((ceil_div(m, 128) * 128, ceil_div(n, 128) * 128), dtype=x.dtype, device=x.device)
x_padded[:m, :n] = x x_padded[:m, :n] = x
x_view = x_padded.view(-1, 128, x_padded.size(1) // 128, 128) x_view = x_padded.view(-1, 128, x_padded.size(1) // 128, 128)
x_amax = x_view.abs().float().amax(dim=(1, 3), keepdim=True).clamp(1e-4) x_amax = x_view.abs().float().amax(dim=(1, 3), keepdim=True).clamp(1e-4)
x_scaled = (x_view * (448.0 / x_amax)).to(torch.float8_e4m3fn) x_scaled = (x_view * (448.0 / x_amax)).to(torch.float8_e4m3fn)
return x_scaled.view_as(x_padded)[:m, :n].contiguous(), (x_amax / 448.0).view(x_view.size(0), x_view.size(2)) return x_scaled.view_as(x_padded)[:m, :n].contiguous(), (x_amax / 448.0).view(x_view.size(0), x_view.size(2))
def construct(m: int, k: int, n: int) -> \ def construct(m: int, k: int, n: int) -> \
Tuple[Tuple[torch.Tensor, torch.Tensor], Tuple[torch.Tensor, torch.Tensor], torch.Tensor, torch.Tensor]: Tuple[Tuple[torch.Tensor, torch.Tensor], Tuple[torch.Tensor, torch.Tensor], torch.Tensor, torch.Tensor]:
x = torch.randn((m, k), device='cuda', dtype=torch.bfloat16) x = torch.randn((m, k), device='cuda', dtype=torch.bfloat16)
@ -37,7 +41,6 @@ def construct(m: int, k: int, n: int) -> \
x_fp8 = (x_fp8[0], get_col_major_tma_aligned_tensor(x_fp8[1])) x_fp8 = (x_fp8[0], get_col_major_tma_aligned_tensor(x_fp8[1]))
return x_fp8, y_fp8, out, ref_out return x_fp8, y_fp8, out, ref_out
def construct_grouped(num_groups: int, m: int, k: int, n: int, is_masked: bool) -> \ def construct_grouped(num_groups: int, m: int, k: int, n: int, is_masked: bool) -> \
Tuple[Tuple[torch.Tensor, torch.Tensor], Tuple[torch.Tensor, torch.Tensor], torch.Tensor, torch.Tensor]: Tuple[Tuple[torch.Tensor, torch.Tensor], Tuple[torch.Tensor, torch.Tensor], torch.Tensor, torch.Tensor]:
x = torch.randn((num_groups, m, k), device='cuda', dtype=torch.bfloat16) x = torch.randn((num_groups, m, k), device='cuda', dtype=torch.bfloat16)
@ -46,8 +49,14 @@ def construct_grouped(num_groups: int, m: int, k: int, n: int, is_masked: bool)
ref_out = torch.einsum('gmk,gnk->gmn', x, y) ref_out = torch.einsum('gmk,gnk->gmn', x, y)
assert m % 4 == 0, f'TMA alignment error: {m}' assert m % 4 == 0, f'TMA alignment error: {m}'
x_fp8 = (torch.empty_like(x, dtype=torch.float8_e4m3fn), torch.empty((num_groups, m, k // 128), device='cuda', dtype=torch.float)) x_fp8 = (
y_fp8 = (torch.empty_like(y, dtype=torch.float8_e4m3fn), torch.empty((num_groups, (n + 127) // 128, k // 128), device='cuda', dtype=torch.float)) torch.empty_like(x, dtype=torch.float8_e4m3fn),
torch.empty((num_groups, m, k // 128), device='cuda', dtype=torch.float)
)
y_fp8 = (
torch.empty_like(y, dtype=torch.float8_e4m3fn),
torch.empty((num_groups, (n + 127) // 128, k // 128), device='cuda', dtype=torch.float)
)
for i in range(num_groups): for i in range(num_groups):
x_fp8[0][i], x_fp8[1][i] = per_token_cast_to_fp8(x[i]) x_fp8[0][i], x_fp8[1][i] = per_token_cast_to_fp8(x[i])
y_fp8[0][i], y_fp8[1][i] = per_block_cast_to_fp8(y[i]) y_fp8[0][i], y_fp8[1][i] = per_block_cast_to_fp8(y[i])
@ -61,98 +70,86 @@ def construct_grouped(num_groups: int, m: int, k: int, n: int, is_masked: bool)
x_fp8 = (x_fp8[0], get_col_major_tma_aligned_tensor(x_fp8[1])) x_fp8 = (x_fp8[0], get_col_major_tma_aligned_tensor(x_fp8[1]))
return x_fp8, y_fp8, out, ref_out return x_fp8, y_fp8, out, ref_out
@pytest.mark.parametrize("m", [64, 128, 4096])
def test_gemm() -> None: @pytest.mark.parametrize("k,n", [(7168, 2112), (1536, 24576), (512, 32768), (16384, 7168), (7168, 4096), (2048, 7168)])
print('Testing GEMM:') def test_gemm(m, k, n):
for m in (64, 128, 4096): """Test single GEMM with various dimensions."""
for k, n in [(7168, 2112), (1536, 24576), (512, 32768), (16384, 7168), (7168, 4096), (2048, 7168)]:
x_fp8, y_fp8, out, ref_out = construct(m, k, n) x_fp8, y_fp8, out, ref_out = construct(m, k, n)
deep_gemm.gemm_fp8_fp8_bf16_nt(x_fp8, y_fp8, out) deep_gemm.gemm_fp8_fp8_bf16_nt(x_fp8, y_fp8, out)
diff = calc_diff(out, ref_out) diff = calc_diff(out, ref_out)
assert diff < 0.001, f'{m=}, {k=}, {n=}, {diff:.5f}' assert diff < 0.001, f'GEMM mismatch: m={m}, k={k}, n={n}, diff={diff:.5f}'
# noinspection PyShadowingNames # (Optional) performance timing
def test_func(): def test_func():
# Construct new tensors every time to avoid L2 cache acceleration # Construct new tensors every time to avoid L2 cache effects
x_fp8, y_fp8, out, ref_out = construct(m, k, n) x_fp8_local, y_fp8_local, out_local, _ = construct(m, k, n)
deep_gemm.gemm_fp8_fp8_bf16_nt(x_fp8, y_fp8, out) deep_gemm.gemm_fp8_fp8_bf16_nt(x_fp8_local, y_fp8_local, out_local)
t = bench_kineto(test_func, 'fp8_gemm', suppress_kineto_output=True) t = bench_kineto(test_func, 'fp8_gemm', suppress_kineto_output=True)
print(f' > Performance (m={m:5}, n={n:5}, k={k:5}): {t * 1e6:4.0f} us | ' print(f'[test_gemm] (m={m:5}, n={n:5}, k={k:5}): {t * 1e6:6.0f} us | '
f'throughput: {2 * m * n * k / t / 1e12:4.0f} TFLOPS, ' f'{2 * m * n * k / t / 1e12:5.2f} TFLOPS')
f'{(m * k + k * n + m * n * 2) / 1e9 / t:4.0f} GB/s')
print()
@pytest.mark.parametrize("num_groups,m,k,n", [
def test_m_grouped_gemm_contiguous() -> None: (4, 8192, 7168, 4096),
print('Testing grouped contiguous GEMM:') (4, 8192, 2048, 7168),
(8, 4096, 7168, 4096),
for num_groups, m, k, n in ((4, 8192, 7168, 4096), (4, 8192, 2048, 7168), (8, 4096, 7168, 4096), (8, 4096, 2048, 7168)): (8, 4096, 2048, 7168)
# TODO: make a stronger test ])
def test_m_grouped_gemm_contiguous(num_groups, m, k, n):
"""Test grouped GEMM with merged group+M dimension."""
x_fp8, y_fp8, out, ref_out = construct_grouped(num_groups, m, k, n, is_masked=False) x_fp8, y_fp8, out, ref_out = construct_grouped(num_groups, m, k, n, is_masked=False)
m_indices = torch.arange(0, num_groups, device='cuda', dtype=torch.int) m_indices = torch.arange(0, num_groups, device='cuda', dtype=torch.int)
m_indices = m_indices.unsqueeze(-1).expand(num_groups, m).contiguous().view(-1) m_indices = m_indices.unsqueeze(-1).expand(num_groups, m).contiguous().view(-1)
deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(x_fp8, y_fp8, out, m_indices) deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(x_fp8, y_fp8, out, m_indices)
diff = calc_diff(out, ref_out) diff = calc_diff(out, ref_out)
assert diff < 0.001, f'm={m * num_groups}, {k=}, {n=}, {diff:.5f}' assert diff < 0.001, f'Grouped GEMM mismatch: m={m*num_groups}, k={k}, n={n}, diff={diff:.5f}'
# noinspection PyShadowingNames # (Optional) performance timing
def test_func(): def test_func():
# Construct new tensors every time to avoid L2 cache acceleration x_fp8_local, y_fp8_local, out_local, _ = construct_grouped(num_groups, m, k, n, is_masked=False)
x_fp8, y_fp8, out, ref_out = construct_grouped(num_groups, m, k, n, is_masked=False) m_indices_local = torch.arange(0, num_groups, device='cuda', dtype=torch.int)
m_indices = torch.arange(0, num_groups, device='cuda', dtype=torch.int) m_indices_local = m_indices_local.unsqueeze(-1).expand(num_groups, m).contiguous().view(-1)
m_indices = m_indices.unsqueeze(-1).expand(num_groups, m).contiguous().view(-1) deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(x_fp8_local, y_fp8_local, out_local, m_indices_local)
deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_contiguous(x_fp8, y_fp8, out, m_indices)
t = bench_kineto(test_func, 'fp8_gemm', suppress_kineto_output=True) t = bench_kineto(test_func, 'fp8_gemm', suppress_kineto_output=True)
print(f' > Performance ({num_groups=}, m_per_group={m:4}, n={n:4}, k={k:4}): {t * 1e6:4.0f} us | ' print(f'[test_m_grouped_gemm_contiguous] (groups={num_groups}, m={m}, n={n}, k={k}): '
f'throughput: {2 * num_groups * m * n * k / t / 1e12:4.0f} TFLOPS, ' f'{t * 1e6:6.0f} us | {2 * num_groups * m * n * k / t / 1e12:5.2f} TFLOPS')
f'{(num_groups * (m * k + k * n + m * n * 2)) / 1e9 / t:4.0f} GB/s')
print()
@pytest.mark.parametrize("num_groups,m", [
def test_m_grouped_gemm_masked() -> None: (1, 1024),
print('Testing grouped masked GEMM:') (2, 512),
(4, 256),
for num_groups, m in ((1, 1024), (2, 512), (4, 256)): ])
for k, n in ((7168, 4096), (2048, 7168), ): @pytest.mark.parametrize("k,n", [
# Test correctness (7168, 4096),
(2048, 7168),
])
def test_m_grouped_gemm_masked(num_groups, m, k, n):
"""Test grouped GEMM where each group can have a different 'masked' M dimension."""
# Test correctness with random partial masks
masked_m_candidates = list(filter(lambda candidate: candidate <= m, (64, 128, 192, 256, 320, 384))) masked_m_candidates = list(filter(lambda candidate: candidate <= m, (64, 128, 192, 256, 320, 384)))
for i in range(10): for _ in range(10):
x_fp8, y_fp8, out, ref_out = construct_grouped(num_groups, m, k, n, is_masked=True) x_fp8, y_fp8, out, ref_out = construct_grouped(num_groups, m, k, n, is_masked=True)
masked_m = torch.empty((num_groups, ), device='cuda', dtype=torch.int) masked_m = torch.empty((num_groups,), device='cuda', dtype=torch.int)
for j in range(num_groups): for j in range(num_groups):
masked_m[j] = random.choice(masked_m_candidates) masked_m[j] = random.choice(masked_m_candidates)
expected_m = int(masked_m.float().mean()) + 1 # We pick an expected M that covers the average usage
expected_m = min(int(masked_m.float().mean()) + 1, m)
deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_masked(x_fp8, y_fp8, out, masked_m, expected_m) deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_masked(x_fp8, y_fp8, out, masked_m, expected_m)
for j in range(num_groups): for j in range(num_groups):
diff = calc_diff(out[j, :masked_m[j].item()], ref_out[j, :masked_m[j].item()]) diff = calc_diff(out[j, :masked_m[j].item()], ref_out[j, :masked_m[j].item()])
assert diff < 0.001, f'{m=}, {k=}, {n=}, {j=}, masked_m={masked_m[j]}, {num_groups=}, {diff:.5f}' assert diff < 0.001, (
f'Grouped Masked GEMM mismatch: m={m}, k={k}, n={n}, group={j}, '
f'masked_m={masked_m[j]}, groups={num_groups}, diff={diff:.5f}'
)
# noinspection PyShadowingNames # Test performance with full dimension (no actual mask)
def test_func(): def test_func():
# Construct new tensors every time to avoid L2 cache acceleration x_fp8_local, y_fp8_local, out_local, _ = construct_grouped(num_groups, m, k, n, is_masked=True)
x_fp8, y_fp8, out, ref_out = construct_grouped(num_groups, m, k, n, is_masked=True) masked_m = torch.ones((num_groups,), device='cuda', dtype=torch.int) * m
masked_m = torch.ones((num_groups, ), device='cuda', dtype=torch.int) * m deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_masked(x_fp8_local, y_fp8_local, out_local, masked_m, m)
deep_gemm.m_grouped_gemm_fp8_fp8_bf16_nt_masked(x_fp8, y_fp8, out, masked_m, m)
# Test performance with fixed shapes
t = bench_kineto(test_func, 'fp8_gemm', suppress_kineto_output=True) t = bench_kineto(test_func, 'fp8_gemm', suppress_kineto_output=True)
print(f' > Performance ({num_groups=}, m_per_group={m:4}, n={n:4}, k={k:4}): {t * 1e6:4.0f} us | ' print(f'[test_m_grouped_gemm_masked] (groups={num_groups}, m={m}, n={n}, k={k}): '
f'throughput: {2 * num_groups * m * n * k / t / 1e12:4.0f} TFLOPS, ' f'{t * 1e6:6.0f} us | {2 * num_groups * m * n * k / t / 1e12:5.2f} TFLOPS')
f'{(num_groups * (m * k + k * n + m * n * 2)) / 1e9 / t:4.0f} GB/s')
print()
if __name__ == '__main__':
torch.backends.cuda.matmul.allow_tf32 = True
torch.backends.cudnn.allow_tf32 = True
torch.manual_seed(0)
random.seed(0)
print('Library path:')
print(f' > {deep_gemm.__path__}\n')
test_gemm()
test_m_grouped_gemm_contiguous()
test_m_grouped_gemm_masked()