[Bugfix][DiffusionGemma] Read the causal mask as int32 and compile the sample step (#59992)

Signed-off-by: ojeda-e <eojeda@league.com>
This commit is contained in:
Mia Ojeda
2026-10-06 19:25:51 +08:00
committed by GitHub
parent 09e0ce3d82
commit 3313825983
5 changed files with 24 additions and 16 deletions
+6 -4
View File
@@ -78,8 +78,9 @@ torch::Tensor get_scheduler_metadata(
input.causal = causal;
input.isa = isa;
input.enable_kv_split = enable_kv_split;
input.dynamic_causal =
dynamic_causal.has_value() ? dynamic_causal->data_ptr<bool>() : nullptr;
input.dynamic_causal = dynamic_causal.has_value()
? dynamic_causal->data_ptr<int32_t>()
: nullptr;
const int64_t kv_cache_idx =
static_cast<int64_t>(parse_fp8_kv_dtype(kv_cache_dtype));
@@ -256,8 +257,9 @@ void cpu_attention_with_kv_cache(
input.s_aux = nullptr;
input.s_aux_is_bf16 = false;
}
input.dynamic_causal =
dynamic_causal.has_value() ? dynamic_causal->data_ptr<bool>() : nullptr;
input.dynamic_causal = dynamic_causal.has_value()
? dynamic_causal->data_ptr<int32_t>()
: nullptr;
input.scale = scale;
input.causal = causal;
input.sliding_window_size = sliding_window;