[upstream #3142] [upstream_ut] RuntimeError: The sycl_ext_oneapi_work_group_scratch_memory feature is not yet available for use with SYCL Graph extension.#25
Draft
Stonepia wants to merge 1 commit into
Conversation
[upstream pytorch#3142] [upstream_ut] RuntimeError: The sycl_ext_oneapi_work_group_scratch_memory feature is not yet available for use with SYCL Graph extension.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
[upstream pytorch#3142] [upstream_ut] RuntimeError: The sycl_ext_oneapi_work_group_scratch_memory feature is not yet available for use with SYCL Graph extension.
Fixes https://github.com/intel-sandbox/torch-xpu-ops-exp/issues/273
Root Cause: The oneAPI SYCL Graph extension does not yet support the sycl_ext_oneapi_work_group_scratch_memory feature, which is used by the flash attention kernel on XPU. When CUDA graph capture is attempted on XPU, the SYCL runtime raises a RuntimeError because graph recording cannot handle this memory type.
Failed Tests:
test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_256_seq_len_k_256_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_1_seq_len_q_256_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_1024_seq_len_k_256_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_True_dropout_p_0_0_float16_scale_l1_fused_kernel0_xpu_float16test_transformers_xpu.py::TestSDPACudaOnlyXPU::test_fused_attention_vs_math_ref_grads_cudagraph_batch_size_8_seq_len_q_1024_seq_len_k_1024_head_dim_64_is_causal_False_dropout_p_0_0_float16_scale0_fused_kernel0_xpu_float16Diff stat: