Skip to content

Enable sm120 support for fused attn if cuDNN is 9.18.1+ #2693

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

Draft
KshitijLakhani wants to merge 4 commits into
base: main
Choose a base branch
from
Draft

Enable sm120 support for fused attn if cuDNN is 9.18.1+ #2693

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
fa2e2cb
Enable sm120 support for fused attn if cuDNN is 9.18.1+
KshitijLakhani Feb 20, 2026
bea8bbb
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Feb 20, 2026
b2f5864
Force intermediate tensors such as S, Sum_Exp, and Max to be BHS1 sha…
KshitijLakhani Mar 2, 2026
076420d
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Mar 2, 2026
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
12 changes: 9 additions & 3 deletions transformer_engine/common/fused_attn/fused_attn_f16_arbitrary_seqlen.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1102,6 +1102,9 @@ void fused_attn_arbitrary_seqlen_fwd(
devPtrSoftmaxOffset = input_SoftmaxOffset->data.dptr;
}

const int device_id = cuda::current_device();
const int sm_arch_ = cuda::sm_arch(device_id);

void *devPtrCuSeqlensQ = cu_seqlens_q->data.dptr;
void *devPtrCuSeqlensKV = cu_seqlens_kv->data.dptr;
void *devPtrSeqOffsetsQ = cu_seqlens_q_padded->data.dptr;
Expand All @@ -1128,15 +1131,17 @@ void fused_attn_arbitrary_seqlen_fwd(
if (return_max_logit) {
Tensor *output_Max = convertNVTETensorCheck(Aux_CTX_Tensors->tensors[i++]);
output_Max->data.dptr = nullptr;
if (q_format == NVTE_QKV_Format::NVTE_THD && cudnn_runtime_version >= 90600) {
if ((q_format == NVTE_QKV_Format::NVTE_THD && cudnn_runtime_version >= 90600) &&
!(sm_arch_ >= 120)) {
output_Max->data.shape = {num_tokens_q, num_attn_heads, 1};
} else {
output_Max->data.shape = {batch, num_attn_heads, max_seqlen_q, 1};
}
output_Max->data.dtype = DType::kFloat32;
Tensor *output_Sum_Exp = convertNVTETensorCheck(Aux_CTX_Tensors->tensors[i++]);
output_Sum_Exp->data.dptr = nullptr;
if (q_format == NVTE_QKV_Format::NVTE_THD && cudnn_runtime_version >= 90600) {
if ((q_format == NVTE_QKV_Format::NVTE_THD && cudnn_runtime_version >= 90600) &&
!(sm_arch_ >= 120)) {
output_Sum_Exp->data.shape = {num_tokens_q, num_attn_heads, 1};
} else {
output_Sum_Exp->data.shape = {batch, num_attn_heads, max_seqlen_q, 1};
Expand All @@ -1145,7 +1150,8 @@ void fused_attn_arbitrary_seqlen_fwd(
} else {
Tensor *output_S = convertNVTETensorCheck(Aux_CTX_Tensors->tensors[i++]);
output_S->data.dptr = nullptr;
if (q_format == NVTE_QKV_Format::NVTE_THD && cudnn_runtime_version >= 90600) {
if ((q_format == NVTE_QKV_Format::NVTE_THD && cudnn_runtime_version >= 90600) &&
!(sm_arch_ >= 120)) {
output_S->data.shape = {num_tokens_q, num_attn_heads, 1};
} else {
output_S->data.shape = {batch, num_attn_heads, max_seqlen_q, 1};
Expand Down
1 change: 1 addition & 0 deletions transformer_engine/common/transformer_engine.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1130,6 +1130,7 @@ int nvte_is_non_tn_fp8_gemm_supported() {
static std::vector<int> cache(num_devices, -1);
static std::vector<std::once_flag> flags(num_devices);
int device_id = transformer_engine::cuda::current_device();
// TODO: KL check if this condition is now supported or not ?
std::call_once(flags[device_id], [&]() {
int deviceComputeCapability = transformer_engine::cuda::sm_arch(device_id);
// Note: this is temporary restriction and should be lifted in the future.
Expand Down
5 changes: 3 additions & 2 deletions transformer_engine/pytorch/attention/dot_product_attention/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -610,6 +610,7 @@ def get_attention_backend(
qkv_layout,
)
use_fused_attention = False
# TODO: KL check if this condition is now supported or not ?
if (
device_compute_capability == (12, 0)
and (head_dim_qk > 128 or head_dim_qk % 8 != 0)
Expand Down Expand Up @@ -690,11 +691,11 @@ def _is_fa3_supported(num_heads, num_gqa_groups, head_dim_qk, head_dim_v, qkv_dt
"padding between sequences, i.e. [a, a, PAD, b, b, b, PAD, c, PAD]"
)
use_flash_attention = False
if device_compute_capability == (12, 0):
if device_compute_capability == (12, 0) and cudnn_version < (9, 18, 1):
if use_fused_attention:
logger.debug(
"Disabling FusedAttention as qkv_format = thd is"
" not supported for compute capability = sm120"
" not supported for compute capability = sm120 and cuDNN version < 9.18.1"
)
use_fused_attention = False

Expand Down