DeepSeek/DeepGEMM
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepGEMM synced 2025-06-26 23:15:49 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Rename variables

Browse Source
This commit is contained in:
yukuai 2025-05-20 16:40:56 +08:00
parent dc89674f47
commit e7e38ed222
1 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
tests/test_core.py
View File

@ -207,7 +207,7 @@ def test_m_grouped_gemm_contiguous() -> None:
t = bench_kineto(test_func, 'fp8_gemm', suppress_kineto_output=True) t = bench_kineto(test_func, 'fp8_gemm', suppress_kineto_output=True)
sum_m = (m_indices != -1).sum().item() sum_m = (m_indices != -1).sum().item()
print(f' > Performance ({num_groups=}, m={m:4}, n={n:4}, k={k:4}): {t * 1e6:4.0f} us | ' print(f' > Performance ({num_groups=}, expected_m_per_group={expected_m_per_group:4}, n={n:4}, k={k:4}): {t * 1e6:4.0f} us | '
f'throughput: {2 * sum_m * n * k / t / 1e12:4.0f} TFLOPS, ' f'throughput: {2 * sum_m * n * k / t / 1e12:4.0f} TFLOPS, '
f'{(sum_m * k + num_groups * k * n + sum_m * n * 2) / 1e9 / t:4.0f} GB/s') f'{(sum_m * k + num_groups * k * n + sum_m * n * 2) / 1e9 / t:4.0f} GB/s')
print() print()
@ -217,14 +217,14 @@ def test_m_grouped_gemm_masked() -> None:
print('Testing grouped masked GEMM:') print('Testing grouped masked GEMM:')
m = 4096 m = 4096
for num_groups, excepted_m in ((1, 1024), (2, 512), (4, 256)): for num_groups, expected_m_per_group in ((1, 1024), (2, 512), (4, 256)):
for k, n in ((7168, 4096), (2048, 7168), ): for k, n in ((7168, 4096), (2048, 7168), ):
# Test correctness # Test correctness
for i in range(10): for i in range(10):
x_fp8, y_fp8, out, ref_out = construct_masked_grouped(num_groups, m, k, n) x_fp8, y_fp8, out, ref_out = construct_masked_grouped(num_groups, m, k, n)
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.randint(int(excepted_m * 0.7), int(excepted_m * 1.3)) masked_m[j] = random.randint(int(expected_m_per_group * 0.7), int(expected_m_per_group * 1.3))
expected_m = min(int(masked_m.float().mean()) + 1, m) 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):
@ -234,7 +234,7 @@ def test_m_grouped_gemm_masked() -> None:
# Construct new tensors only once to avoid L2 cache acceleration (creating them puts them in L2) # Construct new tensors only once to avoid L2 cache acceleration (creating them puts them in L2)
x_fp8, y_fp8, out, ref_out = construct_masked_grouped(num_groups, m, k, n) x_fp8, y_fp8, out, ref_out = construct_masked_grouped(num_groups, m, k, n)
for j in range(num_groups): for j in range(num_groups):
masked_m[j] = random.randint(int(excepted_m * 0.7), int(excepted_m * 1.3)) masked_m[j] = random.randint(int(expected_m_per_group * 0.7), int(expected_m_per_group * 1.3))
expected_m = min(int(masked_m.float().mean()) + 1, m) expected_m = min(int(masked_m.float().mean()) + 1, m)
sum_m = masked_m.sum().item() sum_m = masked_m.sum().item()
@ -244,7 +244,7 @@ def test_m_grouped_gemm_masked() -> None:
# Test performance with fixed shapes # 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' > Performance ({num_groups=}, expected_m_per_group={expected_m_per_group:4}, n={n:4}, k={k:4}): {t * 1e6:4.0f} us | '
f'throughput: {2 * sum_m * n * k / t / 1e12:4.0f} TFLOPS, ' f'throughput: {2 * sum_m * n * k / t / 1e12:4.0f} TFLOPS, '
f'{(sum_m * k + num_groups * k * n + sum_m * n * 2) / 1e9 / t:4.0f} GB/s') f'{(sum_m * k + num_groups * k * n + sum_m * n * 2) / 1e9 / t:4.0f} GB/s')
print() print()