Skip to content

GitLab

Commit 79408d6d authored by zhanghj2's avatar zhanghj2
Browse files

update error info

parent 92a05388
Hide whitespace changes
Inline Side-by-side
Showing with 4 additions and 4 deletions
csrc/extension/flash_api.h
View file @ 79408d6d
......@@ -61,9 +61,9 @@ mha_fwd_kvcache_quantization_mla(
auto q_dtype = q.dtype();
if (kv_cache_dtype == "fp8_e4m3" || kv_cache_dtype == "fp8_e5m2")
{
TORCH_CHECK(kcache.dtype() != q_dtype, "非量化情况下, query and key must have not the same dtype");
TORCH_CHECK(kcache.dtype() != q_dtype, "mha_fwd_kvcache_quantization_mla, query and key must have not the same dtype");
CHECK_DEVICE(k_scale);
TORCH_CHECK(k_scale.dtype() == torch::kFloat32, "非量化情况下, query and key must have the same dtype");
TORCH_CHECK(k_scale.dtype() == torch::kFloat32, "k_scale, dtype error");
// TORCH_CHECK(is_gfx936, "fp8_e4m3 and fp8_e5m2 Attention Forward Kernel (mha_fwd_kvcache_quantization_mla) is only supported on gfx936 architectures");
}
else
......@@ -516,9 +516,9 @@ mha_fwd_kvcache_quantization_q_nope_pe_mla(
auto q_dtype = q_nope.dtype();
if (kv_cache_dtype == "fp8_e5m2")
{
TORCH_CHECK(kcache.dtype() != q_dtype, "非量化情况下, query and key must have not the same dtype");
TORCH_CHECK(kcache.dtype() != q_dtype, "mha_fwd_kvcache_quantization_q_nope_pe_mla, query and key must have not the same dtype");
CHECK_DEVICE(k_scale);
TORCH_CHECK(k_scale.dtype() == torch::kFloat32, "非量化情况下, query and key must have the same dtype");
TORCH_CHECK(k_scale.dtype() == torch::kFloat32, "mha_fwd_kvcache_quantization_q_nope_pe_mla, k_scale dtype error");
}
else
{
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment