Skip to content

Fix gluon attention example for fp8 #10399

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
masahi merged 1 commit into from
May 28, 2026
+13 −8
Merged

Fix gluon attention example for fp8 #10399

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
21 changes: 13 additions & 8 deletions python/examples/gluon/01-attention-forward.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -383,7 +383,8 @@ def get_loop_bounds(self, STAGE: gl.constexpr):

@gluon.jit
def _borrow_s_as_p(config, s_tmem):
p_tmem = s_tmem._reinterpret(config.dtype, [config.SPLIT_M, 2 * config.BLOCK_N], config.p_tmem_layout)
cols: gl.constexpr = s_tmem.dtype.primitive_bitwidth // config.dtype.primitive_bitwidth
p_tmem = s_tmem._reinterpret(config.dtype, [config.SPLIT_M, cols * config.BLOCK_N], config.p_tmem_layout)
return p_tmem.slice(0, config.BLOCK_N)


Expand Down Expand Up @@ -1162,7 +1163,7 @@ def attention_forward(q, k, v, causal, sm_scale, o=None, M=None, *, use_tmem_red
@pytest.mark.parametrize("N_CTX", [1024, 2048, 4096, 8192])
@pytest.mark.parametrize("HEAD_DIM", [64, 128])
@pytest.mark.parametrize("causal", [False, True])
@pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16])
@pytest.mark.parametrize("dtype", [torch.float16, torch.bfloat16, torch.float8_e5m2])
@pytest.mark.parametrize("use_tmem_red", [False, True] if is_blackwell_ultra() else [False])
@pytest.mark.parametrize("cta_layout", [(), ((1, 0), ), ((1, 0), (2, 0))], ids=["1cta", "2ctas", "4ctas"])
@pytest.mark.skipif(not is_blackwell(), reason="Gluon attention is only supported on Blackwell GPUs")
Expand All @@ -1178,15 +1179,19 @@ def alloc_fn(size: int, alignment: int, stream):
pytest.skip("TMEM reduction is only supported on Blackwell Ultra GPUs")

torch.manual_seed(42)
q = (torch.empty((Z, H, N_CTX, HEAD_DIM), dtype=dtype, device=device).normal_(mean=0.0, std=0.5).requires_grad_())
k = (torch.empty((Z, H, N_CTX, HEAD_DIM), dtype=dtype, device=device).normal_(mean=0.0, std=0.5).requires_grad_())
v = (torch.empty((Z, H, N_CTX, HEAD_DIM), dtype=dtype, device=device).normal_(mean=0.0, std=0.5).requires_grad_())
q = (torch.empty((Z, H, N_CTX, HEAD_DIM), device=device).normal_(mean=0.0, std=0.5).to(dtype).requires_grad_())
k = (torch.empty((Z, H, N_CTX, HEAD_DIM), device=device).normal_(mean=0.0, std=0.5).to(dtype).requires_grad_())
v = (torch.empty((Z, H, N_CTX, HEAD_DIM), device=device).normal_(mean=0.0, std=0.5).to(dtype).requires_grad_())
sm_scale = 0.5

ref_out = torch.nn.functional.scaled_dot_product_attention(q, k, v, scale=sm_scale, is_causal=causal)

tri_out, _ = attention_forward(q, k, v, causal, sm_scale, use_tmem_red=use_tmem_red, cta_layout=cta_layout)
torch.testing.assert_close(ref_out, tri_out, atol=1e-2, rtol=0)
if dtype == torch.float8_e5m2:
ref_out = torch.nn.functional.scaled_dot_product_attention(q.float(), k.float(), v.float(), scale=sm_scale,
is_causal=causal)
torch.testing.assert_close(ref_out.to(dtype).float(), tri_out.float(), atol=0.25, rtol=0.25)
else:
ref_out = torch.nn.functional.scaled_dot_product_attention(q, k, v, scale=sm_scale, is_causal=causal)
torch.testing.assert_close(ref_out, tri_out, atol=1e-2, rtol=0)


# ===-----------------------------------------------------------------------===#
Expand Down
Loading