From adc752b7ca202c75d168b965c6fd3813b98f3479 Mon Sep 17 00:00:00 2001 From: mengfeil Date: Wed, 12 Feb 2025 10:47:39 +0800 Subject: [PATCH] Update --- .../inductor_huggingface_inference.csv | 138 ------- .../inductor_huggingface_training.csv | 141 ------- .../inductor_timm_models_inference.csv | 186 ---------- .../inductor_timm_models_training.csv | 195 ---------- .../inductor_torchbench_inference.csv | 342 ----------------- .../inductor_torchbench_training.csv | 345 +----------------- .github/workflows/pull.yml | 2 +- src/comm/DeviceProperties.h | 14 +- 8 files changed, 10 insertions(+), 1353 deletions(-) diff --git a/.github/ci_expected_accuracy/inductor_huggingface_inference.csv b/.github/ci_expected_accuracy/inductor_huggingface_inference.csv index c3ab0f32a..a75d3d225 100644 --- a/.github/ci_expected_accuracy/inductor_huggingface_inference.csv +++ b/.github/ci_expected_accuracy/inductor_huggingface_inference.csv @@ -1,185 +1,47 @@ name,float32,bfloat16,float16,amp_bf16,amp_fp16 - - - AlbertForMaskedLM,pass,pass,pass,pass,pass - - - AlbertForQuestionAnswering,pass,pass,pass,pass,pass - - - AllenaiLongformerBase,pass,pass,pass,pass,pass - - - BartForCausalLM,pass,pass,pass,pass,pass - - - BartForConditionalGeneration,pass,pass,pass,pass,pass - - - BertForMaskedLM,pass,pass,pass,pass,pass - - - BertForQuestionAnswering,pass,pass,pass,pass,pass - - - BlenderbotForCausalLM,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - BlenderbotSmallForCausalLM,pass,pass,pass,pass,pass - - - BlenderbotSmallForConditionalGeneration,pass,pass,pass,pass,pass - - - CamemBert,pass,pass,pass,pass,pass - - - DebertaForMaskedLM,pass,pass,pass,pass,pass - - - DebertaForQuestionAnswering,pass,pass,pass,pass,pass - - - DebertaV2ForMaskedLM,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - DebertaV2ForQuestionAnswering,pass,pass,pass,pass,pass - - - DistilBertForMaskedLM,pass,pass,pass,pass,pass - - - DistilBertForQuestionAnswering,pass,pass,pass,pass,pass - - - DistillGPT2,pass,pass,pass,pass,pass - - - ElectraForCausalLM,pass,pass,pass,pass,pass - - - ElectraForQuestionAnswering,pass,pass,pass,pass,pass - - - GPT2ForSequenceClassification,pass,pass,pass,pass,pass - - - GoogleFnet,pass,pass,pass,pass,pass - - - LayoutLMForMaskedLM,pass,pass,pass,pass,pass - - - LayoutLMForSequenceClassification,pass,pass,pass,pass,pass - - - M2M100ForConditionalGeneration,pass,pass,pass,pass,pass - - - MBartForCausalLM,pass,pass,pass,pass,pass - - - MBartForConditionalGeneration,pass,pass,pass,pass,pass - - - MT5ForConditionalGeneration,pass,pass,pass,pass,pass - - - MegatronBertForCausalLM,pass,pass,pass,pass,pass - - - MegatronBertForQuestionAnswering,pass,pass,pass,pass,pass - - - MobileBertForMaskedLM,pass,pass,pass,pass,pass - - - MobileBertForQuestionAnswering,pass,pass,pass,pass,pass - - - OPTForCausalLM,pass,pass,pass,pass,pass - - - PLBartForCausalLM,pass,pass,pass,pass,pass - - - PLBartForConditionalGeneration,pass,pass,pass,pass,pass - - - PegasusForCausalLM,pass,pass,pass,pass,pass - - - PegasusForConditionalGeneration,pass,pass,pass,pass,pass - - - RobertaForCausalLM,pass,pass,pass,pass,pass - - - RobertaForQuestionAnswering,pass,pass,pass,pass,pass - - - Speech2Text2ForCausalLM,pass,pass,pass,pass,pass - - - T5ForConditionalGeneration,pass,pass,pass,pass,pass - - - T5Small,pass,pass,pass,pass,pass - - - TrOCRForCausalLM,pass,pass,pass,pass,pass - - - XGLMForCausalLM,pass,pass,pass,pass,pass - - - XLNetLMHeadModel,pass,pass,pass,pass,pass - - - YituTechConvBert,pass,pass,pass,pass,pass diff --git a/.github/ci_expected_accuracy/inductor_huggingface_training.csv b/.github/ci_expected_accuracy/inductor_huggingface_training.csv index 43f23f826..e2d5645e2 100644 --- a/.github/ci_expected_accuracy/inductor_huggingface_training.csv +++ b/.github/ci_expected_accuracy/inductor_huggingface_training.csv @@ -1,189 +1,48 @@ name,float32,bfloat16,float16,amp_bf16,amp_fp16 - - - AlbertForMaskedLM,pass,pass,pass,pass,pass - - - AlbertForQuestionAnswering,pass,pass,pass,pass,pass - - - AllenaiLongformerBase,pass,pass,pass,pass,pass - - - BartForCausalLM,pass,pass,pass,pass,pass - - - BartForConditionalGeneration,pass,pass,pass,pass,pass - - - BertForMaskedLM,pass,pass,pass,pass,pass - - - BertForQuestionAnswering,pass,pass,pass,pass,pass - - - BlenderbotForCausalLM,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - BlenderbotSmallForCausalLM,pass,pass,pass,pass,pass - - - BlenderbotSmallForConditionalGeneration,pass,pass,pass,pass,pass - - - CamemBert,pass,pass,pass,pass,pass - - - DebertaForMaskedLM,pass,pass,pass,pass,pass - - - DebertaForQuestionAnswering,pass,pass,pass,pass,pass - - - DebertaV2ForMaskedLM,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - # Skip DebertaV2ForQuestionAnswering issue: https://github.com/intel/torch-xpu-ops/issues/1216 - - - DebertaV2ForQuestionAnswering,fail_accuracy,fail_accuracy,fail_accuracy,pass,pass - - - DistilBertForMaskedLM,pass,pass,pass,pass,pass - - - DistilBertForQuestionAnswering,pass,pass,pass,pass,pass - - - DistillGPT2,pass,pass,pass,pass,pass - - - ElectraForCausalLM,pass,pass,pass,pass,pass - - - ElectraForQuestionAnswering,pass,pass,pass,pass,pass - - - GPT2ForSequenceClassification,pass,pass,pass,pass,pass - - - GoogleFnet,pass,pass,pass,pass,pass - - - LayoutLMForMaskedLM,pass,pass,pass,pass,pass - - - LayoutLMForSequenceClassification,pass,pass,pass,pass,pass - - - M2M100ForConditionalGeneration,pass,pass,pass,pass,pass - - - MBartForCausalLM,pass,pass,pass,pass,pass - - - MBartForConditionalGeneration,pass,pass,pass,pass,pass - - - MT5ForConditionalGeneration,pass,pass,pass,pass,pass - - - MegatronBertForCausalLM,pass,pass,pass,pass,pass - - - MegatronBertForQuestionAnswering,pass,pass,pass,pass,pass - - - MobileBertForMaskedLM,pass,pass,pass,pass,pass - - - MobileBertForQuestionAnswering,pass,pass,pass,pass,pass - - - OPTForCausalLM,pass,pass,pass,pass,pass - - - PLBartForCausalLM,pass,pass,pass,pass,pass - - - PLBartForConditionalGeneration,pass,pass,pass,pass,pass - - - PegasusForCausalLM,pass,pass,pass,pass,pass - - - PegasusForConditionalGeneration,pass,pass,pass,pass,pass - - - RobertaForCausalLM,pass,pass,pass,pass,pass - - - RobertaForQuestionAnswering,pass,pass,pass,pass,pass - - - Speech2Text2ForCausalLM,pass,pass,pass,pass,pass - - - T5ForConditionalGeneration,pass,pass,pass,pass,pass - - - T5Small,pass,pass,pass,pass,pass - - - TrOCRForCausalLM,pass,pass,pass,pass,pass - - - XGLMForCausalLM,pass,pass,pass,pass,pass - - - XLNetLMHeadModel,pass,pass,pass,pass,pass - - - YituTechConvBert,pass,pass,pass,pass,pass diff --git a/.github/ci_expected_accuracy/inductor_timm_models_inference.csv b/.github/ci_expected_accuracy/inductor_timm_models_inference.csv index 2ade4d4ab..6cbd2e854 100644 --- a/.github/ci_expected_accuracy/inductor_timm_models_inference.csv +++ b/.github/ci_expected_accuracy/inductor_timm_models_inference.csv @@ -1,249 +1,63 @@ name,float32,bfloat16,float16,amp_bf16,amp_fp16 - - - adv_inception_v3,pass,pass,pass,pass,pass - - - beit_base_patch16_224,pass,pass,pass,pass,pass - - - botnet26t_256,pass,pass,pass,pass,pass - - - cait_m36_384,pass,pass,pass,pass,pass - - - coat_lite_mini,pass,pass,pass,pass,pass - - - convit_base,pass,pass,pass,pass,pass - - - convmixer_768_32,pass,pass,pass,pass,pass - - - convnext_base,pass,pass,pass,pass,pass - - - crossvit_9_240,pass,pass,pass,pass,pass - - - cspdarknet53,pass,pass,pass,pass,pass - - - deit_base_distilled_patch16_224,pass,pass,pass,pass,pass - - - dla102,pass,pass,pass,pass,pass - - - dm_nfnet_f0,pass,pass,pass,pass,pass - - - dpn107,pass,pass,pass,pass,pass - - - eca_botnext26ts_256,pass,pass,pass,pass,pass - - - eca_halonext26ts,pass,pass,pass,pass,pass - - - ese_vovnet19b_dw,pass,pass,pass,pass,pass - - - fbnetc_100,pass,pass,pass,pass,pass - - - fbnetv3_b,pass,pass,pass,pass,pass - - - gernet_l,pass,pass,pass,pass,pass - - - ghostnet_100,pass,pass,pass,pass,pass - - - gluon_inception_v3,pass,pass,pass,pass,pass - - - gmixer_24_224,pass,pass,pass,pass,pass - - - gmlp_s16_224,pass,pass,pass,pass,pass - - - hrnet_w18,pass,pass,pass,pass,pass - - - inception_v3,pass,pass,pass,pass,pass - - - jx_nest_base,pass,pass,pass,pass,pass - - - lcnet_050,pass,pass,pass,pass,pass - - - # https://github.com/pytorch/pytorch/pull/145112 - - - levit_128,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - mixer_b16_224,pass,pass,pass,pass,pass - - - mixnet_l,pass,pass,pass,pass,pass - - - mnasnet_100,pass,pass,pass,pass,pass - - - mobilenetv2_100,pass,pass,pass,pass,pass - - - mobilenetv3_large_100,pass,pass,pass,pass,pass - - - mobilevit_s,pass,pass,pass,pass,pass - - - nfnet_l0,pass,pass,pass,pass,pass - - - pit_b_224,pass,pass,pass,pass,pass - - - pnasnet5large,pass,pass,pass,pass,pass - - - poolformer_m36,pass,pass,pass,pass,pass - - - regnety_002,pass,pass,pass,pass,pass - - - repvgg_a2,pass,pass,pass,pass,pass - - - res2net101_26w_4s,pass,pass,pass,pass,pass - - - res2net50_14w_8s,pass,pass,pass,pass,pass - - - res2next50,pass,pass,pass,pass,pass - - - resmlp_12_224,pass,pass,pass,pass,pass - - - resnest101e,pass,pass,pass,pass,pass - - - rexnet_100,pass,pass,pass,pass,pass - - - sebotnet33ts_256,pass,pass,pass,pass,pass - - - selecsls42b,pass,pass,pass,pass,pass - - - spnasnet_100,pass,pass,pass,pass,pass - - - swin_base_patch4_window7_224,pass,pass,pass,pass,pass - - - swsl_resnext101_32x16d,pass,pass,pass,pass,pass - - - tf_efficientnet_b0,pass,pass,pass,pass,pass - - - tf_mixnet_l,pass,pass,pass,pass,pass - - - tinynet_a,pass,pass,pass,pass,pass - - - tnt_s_patch16_224,pass,pass,pass,pass,pass - - - twins_pcpvt_base,pass,pass,pass,pass,pass - - - visformer_small,pass,pass,pass,pass,pass - - - vit_base_patch16_224,pass,pass,pass,pass,pass - - - volo_d1_224,pass,pass,pass,pass,pass - - - xcit_large_24_p8_224,pass,pass,pass,pass,pass diff --git a/.github/ci_expected_accuracy/inductor_timm_models_training.csv b/.github/ci_expected_accuracy/inductor_timm_models_training.csv index 0990fca43..710ec855c 100644 --- a/.github/ci_expected_accuracy/inductor_timm_models_training.csv +++ b/.github/ci_expected_accuracy/inductor_timm_models_training.csv @@ -1,261 +1,66 @@ name,float32,bfloat16,float16,amp_bf16,amp_fp16 - - - adv_inception_v3,pass,pass,pass,pass,pass - - - beit_base_patch16_224,pass,pass,pass,pass,pass - - - botnet26t_256,pass,pass,pass,pass,pass - - - cait_m36_384,pass,pass,pass,pass,pass - - - coat_lite_mini,pass,pass,pass,pass,pass - - - convit_base,pass,pass,pass,pass,pass - - - convmixer_768_32,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1274 - - - convnext_base,pass,fail_accuracy,fail_accuracy,pass,pass - - - crossvit_9_240,pass,pass,pass,pass,pass - - - cspdarknet53,pass,pass,pass,pass,pass - - - deit_base_distilled_patch16_224,pass,pass,pass,pass,pass - - - dla102,pass,pass,pass,pass,pass - - - dm_nfnet_f0,pass,pass,pass,pass,pass - - - dpn107,pass,pass,pass,pass,pass - - - eca_botnext26ts_256,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1275 - - - eca_halonext26ts,pass,pass,pass,fail_accuracy,pass - - - ese_vovnet19b_dw,pass,pass,pass,pass,pass - - - fbnetc_100,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1275 - - - fbnetv3_b,pass,pass,pass,fail_accuracy,pass - - - gernet_l,pass,pass,pass,pass,pass - - - ghostnet_100,pass,pass,pass,pass,pass - - - gluon_inception_v3,pass,pass,pass,pass,pass - - - gmixer_24_224,pass,pass,pass,pass,pass - - - gmlp_s16_224,pass,pass,pass,pass,pass - - - hrnet_w18,pass,pass,pass,pass,pass - - - inception_v3,pass,pass,pass,pass,pass - - - jx_nest_base,pass,pass,pass,pass,pass - - - lcnet_050,pass,pass,pass,pass,pass - - - # https://github.com/pytorch/pytorch/pull/145112 - - - levit_128,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - mixer_b16_224,pass,pass,pass,pass,pass - - - mixnet_l,pass,pass,pass,pass,pass - - - mnasnet_100,pass,pass,pass,pass,pass - - - mobilenetv2_100,pass,pass,pass,pass,pass - - - mobilenetv3_large_100,pass,pass,pass,pass,pass - - - mobilevit_s,pass,pass,pass,pass,pass - - - nfnet_l0,pass,pass,pass,pass,pass - - - pit_b_224,pass,pass,pass,pass,pass - - - pnasnet5large,pass,pass,pass,pass,pass - - - poolformer_m36,pass,pass,pass,pass,pass - - - regnety_002,pass,pass,pass,pass,pass - - - repvgg_a2,pass,pass,pass,pass,pass - - - res2net101_26w_4s,pass,pass,pass,pass,pass - - - res2net50_14w_8s,pass,pass,pass,pass,pass - - - res2next50,pass,pass,pass,pass,pass - - - resmlp_12_224,pass,pass,pass,pass,pass - - - resnest101e,pass,pass,pass,pass,pass - - - rexnet_100,pass,pass,pass,pass,pass - - - sebotnet33ts_256,pass,pass,pass,pass,pass - - - selecsls42b,pass,pass,pass,pass,pass - - - spnasnet_100,pass,pass,pass,pass,pass - - - swin_base_patch4_window7_224,pass,pass,pass,pass,pass - - - swsl_resnext101_32x16d,pass,pass,pass,pass,pass - - - tf_efficientnet_b0,pass,pass,pass,pass,pass - - - tf_mixnet_l,pass,pass,pass,pass,pass - - - tinynet_a,pass,pass,pass,pass,pass - - - tnt_s_patch16_224,pass,pass,pass,pass,pass - - - twins_pcpvt_base,pass,pass,pass,pass,pass - - - visformer_small,pass,pass,pass,pass,pass - - - vit_base_patch16_224,pass,pass,pass,pass,pass - - - volo_d1_224,pass,pass,pass,pass,pass - - - xcit_large_24_p8_224,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip diff --git a/.github/ci_expected_accuracy/inductor_torchbench_inference.csv b/.github/ci_expected_accuracy/inductor_torchbench_inference.csv index 801bdd1db..5988681b5 100644 --- a/.github/ci_expected_accuracy/inductor_torchbench_inference.csv +++ b/.github/ci_expected_accuracy/inductor_torchbench_inference.csv @@ -1,457 +1,115 @@ name,float32,bfloat16,float16,amp_bf16,amp_fp16 - - - # https://github.com/intel/torch-xpu-ops/issues/1221 - - - torchrec_dlrm,pass,eager_fail_to_run,eager_fail_to_run,fail_to_run,fail_to_run - - - BERT_pytorch,pass,pass,pass,pass,pass - - - Background_Matting,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - DALLE2_pytorch,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - # https://github.com/intel/torch-xpu-ops/issues/1263 - - - LearningToPaint,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - Super_SloMo,pass,pass,pass,pass,pass - - - alexnet,pass,pass,pass,pass,pass - - - basic_gnn_edgecnn,pass,pass,pass,pass,pass - - - basic_gnn_gcn,pass,pass,pass,pass,pass - - - basic_gnn_gin,pass,pass,pass,pass,pass - - - basic_gnn_sage,pass,pass,pass,pass,pass - - - cm3leon_generate,pass,pass,pass,pass,pass - - - dcgan,pass,pass,pass,pass,pass - - - demucs,pass,pass,pass,pass,pass - - - densenet121,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1278 - - - detectron2_fasterrcnn_r_101_c4,pass,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_fasterrcnn_r_101_dc5,pass,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_fasterrcnn_r_101_fpn,pass,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_fasterrcnn_r_50_c4,pass,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_fasterrcnn_r_50_dc5,pass,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_fasterrcnn_r_50_fpn,pass,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_fcos_r_50_fpn,pass,pass,pass,pass,pass - - - detectron2_maskrcnn,fail_to_run,eager_fail_to_run,fail_to_run,eager_fail_to_run,fail_to_run - - - detectron2_maskrcnn_r_101_c4,fail_accuracy,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_maskrcnn_r_101_fpn,fail_accuracy,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_maskrcnn_r_50_c4,pass,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - detectron2_maskrcnn_r_50_fpn,pass,eager_fail_to_run,fail_accuracy,eager_fail_to_run,fail_accuracy - - - dlrm,pass,pass,pass,pass,pass - - - doctr_det_predictor,pass,pass,pass,pass,pass - - - doctr_reco_predictor,pass,pass,pass,pass,pass - - - drq,pass,pass,pass,pass,pass - - - fastNLP_Bert,pass,pass,pass,pass,pass - - - functorch_dp_cifar10,pass,pass,pass,pass,pass - - - functorch_maml_omniglot,pass,pass,pass,pass,pass - - - hf_Albert,pass,pass,pass,pass,pass - - - hf_Bart,pass,pass,pass,pass,pass - - - hf_Bert,pass,pass,pass,pass,pass - - - hf_Bert_large,pass,pass,pass,pass,pass - - - hf_BigBird,pass,pass,pass,pass,pass - - - hf_DistilBert,pass,pass,pass,pass,pass - - - hf_GPT2,pass,pass,pass,pass,pass - - - hf_GPT2_large,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - hf_Longformer,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1262 - - - hf_Reformer,eager_two_runs_differ,eager_two_runs_differ,eager_two_runs_differ,eager_two_runs_differ,eager_two_runs_differ - - - hf_T5,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1276 - - - hf_T5_base,pass,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - hf_T5_generate,pass,pass,pass,pass,pass - - - hf_T5_large,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - hf_Whisper,pass,pass,pass,pass,pass - - - hf_clip,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - hf_distil_whisper,pass,pass,pass,pass,pass - - - lennard_jones,pass,pass,pass,pass,pass - - - llama,pass,pass,pass,pass,pass - - - llama_v2_7b_16h,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - # https://github.com/intel/torch-xpu-ops/issues/1277 - - - llava,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - maml,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - maml_omniglot,pass,pass,pass,pass,pass - - - microbench_unbacked_tolist_sum,pass,pass,pass,pass,pass - - - mnasnet1_0,pass,pass,pass,pass,pass - - - mobilenet_v2,pass,pass,pass,pass,pass - - - mobilenet_v2_quantized_qat,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - mobilenet_v3_large,pass,pass,pass,pass,pass - - - moco,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - moondream,pass,pass,pass,pass,pass - - - nanogpt,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1260 - - - nvidia_deeprecommender,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - opacus_cifar10,pass,pass,pass,pass,pass - - - phlippe_densenet,pass,pass,pass,pass,pass - - - phlippe_resnet,pass,pass,pass,pass,pass - - - pyhpc_equation_of_state,pass,pass,pass,pass,pass - - - pyhpc_isoneutral_mixing,pass,pass,pass,pass,pass - - - pyhpc_turbulent_kinetic_energy,pass,pass,pass,pass,pass - - - pytorch_CycleGAN_and_pix2pix,pass,pass,pass,pass,pass - - - pytorch_stargan,pass,pass,pass,pass,pass - - - pytorch_unet,pass,pass,pass,pass,pass - - - resnet152,pass,pass,pass,pass,pass - - - resnet18,pass,pass,pass,pass,pass - - - resnet50,pass,pass,pass,pass,pass - - - resnet50_quantized_qat,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - resnext50_32x4d,pass,pass,pass,pass,pass - - - sam,pass,pass,pass,pass,pass - - - sam_fast,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - shufflenet_v2_x1_0,pass,pass,pass,pass,pass - - - simple_gpt,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - simple_gpt_tp_manual,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - # https://github.com/intel/torch-xpu-ops/issues/1273 - - - soft_actor_critic,pass,fail_accuracy,pass,pass,pass - - - speech_transformer,pass,pass,pass,pass,pass - - - squeezenet1_1,pass,fail_accuracy,pass,pass,pass - - - stable_diffusion_text_encoder,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1261 - - - stable_diffusion_unet,eager_fail_to_run,pass_due_to_skip,pass_due_to_skip,eager_fail_to_run,eager_fail_to_run - - - tacotron2,pass,pass,pass,fail_to_run,fail_to_run - - - timm_efficientdet,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - timm_efficientnet,pass,pass,pass,pass,pass - - - timm_nfnet,pass,pass,pass,pass,pass - - - timm_regnet,pass,pass,pass,pass,pass - - - timm_resnest,pass,pass,pass,pass,pass - - - timm_vision_transformer,pass,pass,pass,pass,pass - - - timm_vision_transformer_large,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - timm_vovnet,pass,pass,pass,pass,pass - - - torch_multimodal_clip,pass,pass,pass,pass,pass - - - tts_angular,pass,pass,pass,pass,pass - - - vgg16,pass,pass,pass,pass,pass - - - vision_maskrcnn,pass,pass,pass,eager_fail_to_run,eager_fail_to_run - - - yolov3,pass,pass,pass,pass,pass - - - hf_Roberta_base,pass,pass,pass,pass,pass diff --git a/.github/ci_expected_accuracy/inductor_torchbench_training.csv b/.github/ci_expected_accuracy/inductor_torchbench_training.csv index 90f4f2cbf..2a007a86f 100644 --- a/.github/ci_expected_accuracy/inductor_torchbench_training.csv +++ b/.github/ci_expected_accuracy/inductor_torchbench_training.csv @@ -1,457 +1,116 @@ name,float32,bfloat16,float16,amp_bf16,amp_fp16 - - - torchrec_dlrm,pass,eager_fail_to_run,eager_fail_to_run,pass,pass - - - BERT_pytorch,pass,pass,pass,pass,pass - - - Background_Matting,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - DALLE2_pytorch,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - # https://github.com/intel/torch-xpu-ops/issues/1263 - - - LearningToPaint,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - # https://github.com/intel/torch-xpu-ops/issues/1256 - - - Super_SloMo,eager_two_runs_differ,pass,pass,eager_two_runs_differ,pass - - - alexnet,pass,pass,pass,pass,pass - - - basic_gnn_edgecnn,pass,pass,pass,pass,pass - - - basic_gnn_gcn,pass,pass,pass,pass,pass - - - basic_gnn_gin,pass,pass,pass,pass,pass - - - basic_gnn_sage,pass,pass,pass,pass,pass - - - cm3leon_generate,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - dcgan,pass,pass,pass,pass,pass - - - demucs,eager_two_runs_differ,eager_fail_to_run,eager_fail_to_run,eager_two_runs_differ,eager_two_runs_differ - - - densenet121,pass,pass,pass,pass,pass - - - detectron2_fasterrcnn_r_101_c4,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_fasterrcnn_r_101_dc5,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_fasterrcnn_r_101_fpn,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_fasterrcnn_r_50_c4,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_fasterrcnn_r_50_dc5,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_fasterrcnn_r_50_fpn,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_fcos_r_50_fpn,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - detectron2_maskrcnn,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_maskrcnn_r_101_c4,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_maskrcnn_r_101_fpn,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_maskrcnn_r_50_c4,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - detectron2_maskrcnn_r_50_fpn,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - dlrm,pass,pass,pass,pass,pass - - - doctr_det_predictor,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - doctr_reco_predictor,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - drq,pass,pass,pass,pass,pass - - - fastNLP_Bert,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/508 - - - functorch_dp_cifar10,fail_accuracy,fail_accuracy,fail_accuracy,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/510 - - - functorch_maml_omniglot,pass,pass,pass,fail_accuracy,pass - - - hf_Albert,pass,pass,pass,pass,pass - - - hf_Bart,pass,pass,pass,pass,pass - - - hf_Bert,pass,pass,pass,pass,pass - - - hf_Bert_large,pass,pass,pass,pass,pass - - - hf_BigBird,pass,pass,pass,pass,pass - - - hf_DistilBert,pass,pass,pass,pass,pass - - - hf_GPT2,pass,pass,pass,pass,pass - - - hf_GPT2_large,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - hf_Longformer,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1262 - - - hf_Reformer,eager_two_runs_differ,eager_two_runs_differ,eager_two_runs_differ,eager_two_runs_differ,eager_two_runs_differ - - - hf_T5,pass,pass,pass,pass,pass - - - hf_T5_base,pass,pass,pass,pass,pass - - - hf_T5_generate,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - hf_T5_large,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - hf_Whisper,pass,pass,pass,pass,pass - - - hf_clip,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - hf_distil_whisper,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - lennard_jones,pass,pass,pass,pass,pass - - - llama,pass,pass,pass,pass,pass - - - llama_v2_7b_16h,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - llava,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - maml,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - maml_omniglot,pass,pass,pass,pass,pass - - - microbench_unbacked_tolist_sum,pass,pass,pass,pass,pass - - - mnasnet1_0,pass,pass,pass,pass,pass - - - mobilenet_v2,pass,pass,pass,pass,pass - - - mobilenet_v2_quantized_qat,pass,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - mobilenet_v3_large,pass,pass,pass,pass,pass - - - moco,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - moondream,pass,pass,pass,pass,pass - - - nanogpt,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1260 - - - nvidia_deeprecommender,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - opacus_cifar10,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - phlippe_densenet,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/509 - - - phlippe_resnet,pass,fail_accuracy,pass,fail_accuracy,pass - - - pyhpc_equation_of_state,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - pyhpc_isoneutral_mixing,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - pyhpc_turbulent_kinetic_energy,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - pytorch_CycleGAN_and_pix2pix,eager_two_runs_differ,eager_two_runs_differ,pass,eager_two_runs_differ,pass - - - pytorch_stargan,pass,pass,pass,pass,pass - - - pytorch_unet,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - resnet152,pass,pass,pass,pass,pass - - - resnet18,pass,pass,pass,pass,pass - - - resnet50,pass,pass,pass,pass,pass - - - resnet50_quantized_qat,pass,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - resnext50_32x4d,pass,pass,pass,pass,pass - - - sam,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - sam_fast,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - shufflenet_v2_x1_0,pass,pass,pass,pass,pass - - - simple_gpt,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - simple_gpt_tp_manual,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - soft_actor_critic,pass,pass,pass,pass,pass - - - speech_transformer,pass,pass,pass,pass,pass - - - squeezenet1_1,pass,pass,pass,pass,pass - - - stable_diffusion_text_encoder,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1261 - - - stable_diffusion_unet,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - tacotron2,fail_to_run,fail_to_run,fail_to_run,fail_to_run,fail_to_run - - - timm_efficientdet,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load,model_fail_to_load - - - timm_efficientnet,pass,pass,pass,pass,pass - - - timm_nfnet,pass,pass,pass,pass,pass - - - -timm_regnet,pass,pass,pass,pass,pass - - - +# https://github.com/intel/torch-xpu-ops/issues/1334 +timm_regnet,pass,fail_accuracy,pass,pass,pass timm_resnest,pass,pass,pass,pass,pass - - - timm_vision_transformer,pass,pass,pass,pass,pass - - - timm_vision_transformer_large,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip,pass_due_to_skip - - - timm_vovnet,pass,pass,pass,pass,pass - - - torch_multimodal_clip,pass,pass,pass,pass,pass - - - tts_angular,pass,pass,pass,pass,pass - - - vgg16,pass,pass,pass,pass,pass - - - # https://github.com/intel/torch-xpu-ops/issues/1264 - - - vision_maskrcnn,eager_fail_to_run,pass,pass,eager_fail_to_run,eager_fail_to_run - - - yolov3,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run,eager_fail_to_run - - - hf_Roberta_base,pass,pass,pass,pass,pass diff --git a/.github/workflows/pull.yml b/.github/workflows/pull.yml index a25df15ec..46aa2428c 100644 --- a/.github/workflows/pull.yml +++ b/.github/workflows/pull.yml @@ -30,7 +30,7 @@ jobs: uses: actions/checkout@v4 - name: Run lint check run: | - export ADDITIONAL_LINTRUNNER_ARGS="--skip CLANGTIDY,CLANGFORMAT --all-files" + export ADDITIONAL_LINTRUNNER_ARGS="--skip CLANGTIDY,CLANGFORMAT,MERGE_CONFLICTLESS_CSV --all-files" bash .github/scripts/lintrunner.sh preci-linux-build: diff --git a/src/comm/DeviceProperties.h b/src/comm/DeviceProperties.h index 4638bb90f..a45d7b08e 100644 --- a/src/comm/DeviceProperties.h +++ b/src/comm/DeviceProperties.h @@ -147,25 +147,25 @@ uint32_t syclPrefVectorWidth( uint32_t vec_width = 16; if (std::is_same::value) { - return vec_width/sizeof(char); + return vec_width / sizeof(char); } if (std::is_same::value) { - return vec_width/sizeof(short); + return vec_width / sizeof(short); } if (std::is_same::value) { - return vec_width/sizeof(int); + return vec_width / sizeof(int); } if (std::is_same::value) { - return vec_width/sizeof(int64_t); + return vec_width / sizeof(int64_t); } if (std::is_same::value) { - return vec_width/sizeof(float); + return vec_width / sizeof(float); } if (std::is_same::value) { - return vec_width/sizeof(double); + return vec_width / sizeof(double); } if (std::is_same::value) { - return vec_width/sizeof(::sycl::half); + return vec_width / sizeof(::sycl::half); } throw std::invalid_argument( "Invalid data type to fetch preferred vector width!");