Reduce formatting diffs #20348
reviewdog [misspell] report
reported by reviewdog 🐶
Findings (0)
Filtered Findings (432)
tools/ci_build/amd_hipify.py|100 col 38| "marshalling" is a misspelling of "marshaling"
tools/ci_build/github/azure-pipelines/web-ci-pipeline.yml|38 col 41| "varaibles" is a misspelling of "variables"
tools/ci_build/github/azure-pipelines/npm-packaging-pipeline.yml|18 col 41| "varaibles" is a misspelling of "variables"
tools/ci_build/github/windows/install_third_party_deps.ps1|4 col 62| "spectre" is a misspelling of "specter"
tools/ci_build/github/linux/copy_strip_binary.sh|59 col 19| "licence" is a misspelling of "license"
tools/ci_build/build.py|657 col 82| "targetting" is a misspelling of "targeting"
tools/ci_build/build.py|2485 col 52| "behaviour" is a misspelling of "behavior"
csharp/tools/Microsoft.ML.OnnxRuntime.PerfTool/Program.cs|35 col 69| "exection" is a misspelling of "execution"
csharp/ApiDocs/_exported_templates/default/styles/docfx.vendor.js|1 col 205613| "EXPORTAS" is a misspelling of "EXPORTS"
csharp/ApiDocs/_exported_templates/default/styles/docfx.vendor.js|1 col 235884| "strat" is a misspelling of "start"
csharp/ApiDocs/_exported_templates/default/styles/docfx.vendor.js|1 col 262510| "AVERAGEIFS" is a misspelling of "AVERAGES"
csharp/ApiDocs/_exported_templates/default/partials/title.tmpl.partial|42| "Explict" is a misspelling of "Explicit"
csharp/sample/Microsoft.ML.OnnxRuntime.ResNet50v2Sample/LabelMap.cs|94 col 57| "sulphur" is a misspelling of "sulfur"
csharp/sample/Microsoft.ML.OnnxRuntime.ResNet50v2Sample/LabelMap.cs|330 col 57| "sulphur" is a misspelling of "sulfur"
csharp/test/Microsoft.ML.OnnxRuntime.Tests.Common/InferenceTest.cs|1895 col 32| "thrid" is a misspelling of "third"
csharp/src/Microsoft.ML.OnnxRuntime/OrtValue.shared.cs|25 col 96| "UNKOWN" is a misspelling of "UNKNOWN"
csharp/src/Microsoft.ML.OnnxRuntime/NativeMethods.shared.cs|1329 col 129| "transfered" is a misspelling of "transferred"
include/onnxruntime/core/session/onnxruntime_c_api.h|465 col 26| "overriden" is a misspelling of "overridden"
include/onnxruntime/core/session/onnxruntime_c_api.h|471 col 26| "overriden" is a misspelling of "overridden"
include/onnxruntime/core/session/onnxruntime_c_api.h|551 col 26| "overriden" is a misspelling of "overridden"
include/onnxruntime/core/session/onnxruntime_c_api.h|557 col 26| "overriden" is a misspelling of "overridden"
include/onnxruntime/core/session/onnxruntime_c_api.h|2765 col 60| "extention" is a misspelling of "extension"
include/onnxruntime/core/optimizer/graph_transformer_config.h|16 col 31| "redundent" is a misspelling of "redundant"
include/onnxruntime/core/platform/EigenNonBlockingThreadPool.h|1132 col 24| "avaiable" is a misspelling of "available"
winml/api/Microsoft.AI.MachineLearning.Experimental.idl|131 col 79| "specifiy" is a misspelling of "specify"
winml/api/Windows.AI.MachineLearning.idl|12 col 18| "attirbute" is a misspelling of "attribute"
winml/lib/Api/impl/NumericData.h|18 col 5| "Privte" is a misspelling of "Private"
winml/lib/Api/LearningModelBinding.cpp|33 col 12| "insensetive" is a misspelling of "insensitive"
winml/test/api/LearningModelSessionAPITest.cpp|318 col 50| "overriden" is a misspelling of "overridden"
winml/test/model/model_tests.cpp|173 col 77| "accomodate" is a misspelling of "accommodate"
winml/test/common/test.h|21 col 14| "SUPRESS" is a misspelling of "SUPPRESS"
winml/test/common/googleTestMacros.h|67 col 39| "SUPRESS" is a misspelling of "SUPPRESS"
winml/test/common/taefTestMacros.h|51 col 8| "SUPRESS" is a misspelling of "SUPPRESS"
java/src/test/java/sample/ScoreMNIST.java|244 col 30| "probabilites" is a misspelling of "probabilities"
java/src/main/java/ai/onnxruntime/providers/package-info.java|6 col 32| "behaviour" is a misspelling of "behavior"
java/src/main/java/ai/onnxruntime/OnnxRuntime.java|435 col 60| "recognised" is a misspelling of "recognized"
java/build.gradle|45 col 36| "signalling" is a misspelling of "signaling"
orttraining/tools/scripts/layer_norm_transform.py|167 col 29| "shoudl" is a misspelling of "should"
orttraining/tools/scripts/opset12_model_transform.py|5 col 51| "orginal" is a misspelling of "original"
orttraining/tools/scripts/model_transform.py|276 col 50| "differnt" is a misspelling of "different"
orttraining/tools/scripts/model_transform.py|276 col 59| "verison" is a misspelling of "version"
orttraining/orttraining/python/training/model_desc_validation.py|20 col 73| "accross" is a misspelling of "across"
orttraining/orttraining/python/training/model_desc_validation.py|21 col 59| "withing" is a misspelling of "within"
orttraining/orttraining/python/training/orttrainer_options.py|410 col 47| "accross" is a misspelling of "across"
orttraining/orttraining/python/training/torchdynamo/ort_backend.py|136 col 23| "placehoder" is a misspelling of "placeholder"
orttraining/orttraining/python/training/postprocess.py|292 col 25| "ouput" is a misspelling of "output"
orttraining/orttraining/python/training/ortmodule/_custom_gradient_registry.py|241 col 59| "convienent" is a misspelling of "convenient"
orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/cuda/fused_ops/fused_ops_frontend.cpp|13 col 41| "implict" is a misspelling of "implicit"
orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/cuda/fused_ops/fused_ops_frontend.cpp|153 col 22| "threshhold" is a misspelling of "threshold"
orttraining/orttraining/python/training/ortmodule/torch_cpp_extensions/cuda/fused_ops/fused_ops_frontend.cpp|170 col 26| "threshhold" is a misspelling of "threshold"
orttraining/orttraining/python/training/ortmodule/_execution_agent.py|103 col 45| "foward" is a misspelling of "forward"
orttraining/orttraining/python/training/ortmodule/init.py|58 col 53| "proceding" is a misspelling of "proceeding"
orttraining/orttraining/python/training/ortmodule/init.py|93 col 2| "Initalized" is a misspelling of "Initialized"
orttraining/orttraining/python/training/onnxblock/blocks.py|350 col 14| "Supress" is a misspelling of "Suppress"
orttraining/orttraining/python/training/onnxblock/blocks.py|355 col 18| "Supress" is a misspelling of "Suppress"
orttraining/orttraining/python/ort_trainer.py|706 col 58| "overlaping" is a misspelling of "overlapping"
orttraining/orttraining/test/distributed/partition_utils.h|341 col 93| "origional" is a misspelling of "original"
orttraining/orttraining/test/distributed/partition_utils.h|417 col 48| "infomation" is a misspelling of "information"
orttraining/orttraining/test/python/qat_poc_example/qat.py|27 col 9| "Begining" is a misspelling of "Beginning"
orttraining/orttraining/test/python/orttraining_test_ortmodule_api.py|266 col 56| "begining" is a misspelling of "beginning"
orttraining/orttraining/test/graph/bert_toy_fetches.h|11 col 14| "arrary" is a misspelling of "array"
orttraining/orttraining/test/training_ops/cuda/cross_entropy_test.cc|1041 col 26| "seperately" is a misspelling of "separately"
orttraining/orttraining/models/mnist/main.cc|72 col 7| "seperated" is a misspelling of "separated"
orttraining/orttraining/models/mnist/main.cc|304 col 42| "cancelled" is a misspelling of "canceled"
orttraining/orttraining/models/bert/main.cc|210 col 7| "seperated" is a misspelling of "separated"
orttraining/orttraining/models/bert/main.cc|574 col 50| "substract" is a misspelling of "subtract"
orttraining/orttraining/models/runner/training_runner.cc|1191 col 46| "avaliable" is a misspelling of "available"
orttraining/orttraining/core/graph/optimizer_graph_builder.h|128 col 26| "overriden" is a misspelling of "overridden"
orttraining/orttraining/core/graph/training_op_defs.cc|1740 col 59| "udpated" is a misspelling of "updated"
orttraining/orttraining/core/graph/training_op_defs.cc|1823 col 35| "graident" is a misspelling of "gradient"
orttraining/orttraining/core/graph/training_op_defs.cc|2496 col 15| "ouput" is a misspelling of "output"
orttraining/orttraining/core/graph/training_op_defs.cc|2533 col 22| "compresssed" is a misspelling of "compressed"
orttraining/orttraining/core/graph/training_op_defs.cc|2571 col 28| "compresssed" is a misspelling of "compressed"
orttraining/orttraining/core/graph/training_op_defs.cc|2609 col 22| "compresssed" is a misspelling of "compressed"
orttraining/orttraining/core/graph/training_op_defs.cc|2685 col 22| "compresssed" is a misspelling of "compressed"
orttraining/orttraining/core/graph/training_op_defs.cc|3194 col 48| "continous" is a misspelling of "continuous"
orttraining/orttraining/core/graph/training_op_defs.cc|3243 col 56| "devided" is a misspelling of "divided"
orttraining/orttraining/core/graph/training_op_defs.cc|3639 col 45| "ouput" is a misspelling of "output"
orttraining/orttraining/core/graph/training_op_defs.cc|3692 col 45| "ouput" is a misspelling of "output"
orttraining/orttraining/core/graph/training_op_defs.cc|4743 col 49| "behaviour" is a misspelling of "behavior"
orttraining/orttraining/core/graph/mixed_precision_transformer.cc|49 col 8| "possibile" is a misspelling of "possible"
orttraining/orttraining/core/framework/pipeline.cc|196 col 11| "Foward" is a misspelling of "Forward"
orttraining/orttraining/core/graph/pipeline_transformer.cc|449 col 7| "diferent" is a misspelling of "different"
orttraining/orttraining/core/graph/pipeline_transformer.cc|1618 col 42| "accross" is a misspelling of "across"
orttraining/orttraining/core/framework/ortmodule_graph_builder.cc|226 col 22| "accomodate" is a misspelling of "accommodate"
orttraining/orttraining/core/framework/adasum/adasum_interface.h|141 col 18| "desireable" is a misspelling of "desirable"
orttraining/orttraining/core/framework/torch/torch_proxy.h|25 col 3| "Explaination" is a misspelling of "Explanation"
orttraining/orttraining/core/session/training_session.cc|1428 col 22| "recieved" is a misspelling of "received"
orttraining/orttraining/core/session/training_session.cc|1447 col 56| "recieves" is a misspelling of "receives"
orttraining/orttraining/core/optimizer/graph_transformer_config.h|20 col 43| "ouput" is a misspelling of "output"
orttraining/orttraining/core/optimizer/insert_output_rewriter.h|10 col 31| "addtional" is a misspelling of "additional"
orttraining/orttraining/core/optimizer/insert_output_rewriter.h|27 col 31| "addtional" is a misspelling of "additional"
orttraining/orttraining/training_ops/cuda/math/div_grad_impl.cu|13 col 57| "funciton" is a misspelling of "function"
orttraining/orttraining/training_ops/cuda/optimizer/lamb.cc|585 col 35| "accomodate" is a misspelling of "accommodate"
orttraining/orttraining/training_ops/cuda/optimizer/lamb_impl.cu|195 col 16| "coefficeint" is a misspelling of "coefficient"
orttraining/orttraining/training_ops/cpu/activation/activations_grad.cc|85 col 25| "implentation" is a misspelling of "implementation"
.gitattributes|1 col 24| "behaviour" is a misspelling of "behavior"
onnxruntime/wasm/api.h|6 col 87| "funtion" is a misspelling of "function"
onnxruntime/python/onnxruntime_pybind_schema.cc|18 col 39| "registed" is a misspelling of "registered"
onnxruntime/python/tools/quantization/quant_utils.py|270 col 44| "trasnform" is a misspelling of "transform"
onnxruntime/python/tools/quantization/operators/direct_q8.py|16 col 28| "compatiblity" is a misspelling of "compatibility"
onnxruntime/python/tools/transformers/fusion_embedlayer.py|51 col 11| "distil" is a misspelling of "distill"
onnxruntime/python/tools/transformers/fusion_embedlayer.py|58 col 15| "distil" is a misspelling of "distill"
onnxruntime/python/tools/transformers/fusion_embedlayer.py|105 col 14| "distil" is a misspelling of "distill"
onnxruntime/python/tools/transformers/fusion_embedlayer.py|558 col 80| "distil" is a misspelling of "distill"
onnxruntime/python/tools/transformers/fusion_embedlayer.py|637 col 80| "distil" is a misspelling of "distill"
onnxruntime/python/tools/transformers/fusion_embedlayer.py|673 col 80| "distil" is a misspelling of "distill"
onnxruntime/python/tools/transformers/notebooks/PyTorch_Bert-Squad_OnnxRuntime_GPU.ipynb|1668 col 43| "verison" is a misspelling of "version"
onnxruntime/python/tools/transformers/onnx_model_tnlr.py|20 col 29| "addtion" is a misspelling of "addition"
onnxruntime/python/tools/transformers/shape_optimizer.py|6 col 123| "senarios" is a misspelling of "scenarios"
onnxruntime/python/tools/transformers/benchmark.py|924 col 42| "avaiable" is a misspelling of "available"
onnxruntime/python/tools/transformers/README.md|32 col 19| "ususally" is a misspelling of "usually"
onnxruntime/python/tools/transformers/dev_benchmark.cmd|44 col 16| "mutliple" is a misspelling of "multiple"
onnxruntime/python/tools/transformers/onnx_model_t5.py|725 col 50| "expeced" is a misspelling of "expected"
onnxruntime/python/tools/transformers/models/gpt2/convert_to_onnx.py|108 col 18| "aboslute" is a misspelling of "absolute"
onnxruntime/python/tools/transformers/models/gpt2/gpt2_tester.py|140 col 10| "Emtpy" is a misspelling of "Empty"
onnxruntime/python/tools/transformers/models/gpt2/benchmark_gpt2.py|196 col 11| "scirpt" is a misspelling of "script"
onnxruntime/python/tools/transformers/optimizer.py|492 col 21| "symoblic" is a misspelling of "symbolic"
onnxruntime/python/tools/transformers/run_benchmark.sh|62 col 14| "mutliple" is a misspelling of "multiple"
onnxruntime/python/onnxruntime_pybind_state.cc|160 col 34| "destory" is a misspelling of "destroy"
onnxruntime/python/onnxruntime_pybind_state.cc|729 col 33| "Neverthless" is a misspelling of "Nevertheless"
onnxruntime/python/onnxruntime_pybind_state.cc|748 col 31| "Neverthless" is a misspelling of "Nevertheless"
onnxruntime/python/onnxruntime_pybind_state.cc|1146 col 15| "prefered" is a misspelling of "preferred"
onnxruntime/python/onnxruntime_pybind_sparse_tensor.cc|45 col 27| "corresponing" is a misspelling of "corresponding"
onnxruntime/python/onnxruntime_pybind_sparse_tensor.cc|411 col 63| "modificaiton" is a misspelling of "modification"
onnxruntime/test/testdata/transform/model_creation_for_testing.ipynb|312 col 86| "overriden" is a misspelling of "overridden"
onnxruntime/test/testdata/transform/model_parallel/self_attention_megatron_basic_test.py|10 col 14| "attension" is a misspelling of "attention"
onnxruntime/test/python/quantization/test_quantize_static_resnet.py|90 col 36| "guaranted" is a misspelling of "guaranteed"
onnxruntime/test/python/transformers/test_generation.py|50 col 36| "centre" is a misspelling of "center"
onnxruntime/test/python/onnxruntime_test_python_sparse_matmul.py|57 col 8| "Mutliple" is a misspelling of "Multiple"
onnxruntime/test/python/onnxruntime_test_python_mlops.py|176 col 49| "behaviour" is a misspelling of "behavior"
onnxruntime/test/mlas/unittest/test_pool2d_fixture.h|10 col 53| "seperately" is a misspelling of "separately"
onnxruntime/test/mlas/unittest/test_fgemm_fixture.h|12 col 53| "seperately" is a misspelling of "separately"
onnxruntime/test/mlas/unittest/test_symm_qgemm_fixture.h|10 col 53| "seperately" is a misspelling of "separately"
onnxruntime/test/mlas/unittest/test_qgemm_fixture.h|10 col 53| "seperately" is a misspelling of "separately"
onnxruntime/test/mlas/unittest/test_halfgemm.cpp|20 col 53| "seperately" is a misspelling of "separately"
onnxruntime/test/mlas/unittest/test_pool3d_fixture.h|10 col 53| "seperately" is a misspelling of "separately"
onnxruntime/test/providers/internal_testing/internal_testing_partitioning_tests.cc|86 col 13| "funtion" is a misspelling of "function"
onnxruntime/test/providers/cuda/test_cases/allocator_cuda_test.cc|17 col 27| "avaliable" is a misspelling of "available"
onnxruntime/test/providers/dnnl/transformer/matmul_post_op_transform_test.cc|17 col 67| "occured" is a misspelling of "occurred"
onnxruntime/test/providers/dnnl/transformer/matmul_post_op_transform_test.cc|20 col 63| "durring" is a misspelling of "during"
onnxruntime/test/providers/dnnl/transformer/matmul_post_op_transform_test.cc|676 col 27| "accross" is a misspelling of "across"
onnxruntime/test/providers/cpu/rnn/deep_cpu_lstm_op_test.cc|285 col 31| "behaviour" is a misspelling of "behavior"
onnxruntime/test/providers/cpu/rnn/deep_cpu_lstm_op_test.cc|334 col 31| "behaviour" is a misspelling of "behavior"
onnxruntime/test/providers/cpu/controlflow/scan_test.cc|158 col 64| "behaviour" is a misspelling of "behavior"
onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc|424 col 47| "accuarcy" is a misspelling of "accuracy"
onnxruntime/test/providers/cpu/nn/string_normalizer_test.cc|137 col 12| "INSENSETIVE" is a misspelling of "INSENSITIVE"
onnxruntime/test/fuzzing/include/BetaDistribution.h|86 col 43| "probabilty" is a misspelling of "probability"
onnxruntime/test/fuzzing/include/BetaDistribution.h|140 col 18| "probabilty" is a misspelling of "probability"
onnxruntime/test/fuzzing/src/test.cpp|368 col 75| "preceeding" is a misspelling of "preceding"
onnxruntime/test/framework/tunable_op_test.cc|668 col 49| "corretly" is a misspelling of "correctly"
onnxruntime/test/framework/allocation_planner_test.cc|1228 col 106| "seperately" is a misspelling of "separately"
onnxruntime/test/shared_lib/test_inference.cc|1363 col 6| "accomodate" is a misspelling of "accommodate"
onnxruntime/test/shared_lib/test_inference.cc|1414 col 6| "accomodate" is a misspelling of "accommodate"
onnxruntime/test/platform/android/cxa_demangle_test.cc|30 col 26| "omited" is a misspelling of "omitted"
onnxruntime/test/contrib_ops/attention_lstm_op_test.cc|269 col 34| "lenghts" is a misspelling of "lengths"
onnxruntime/test/contrib_ops/attention_lstm_op_test.cc|270 col 40| "lenghts" is a misspelling of "lengths"
onnxruntime/test/contrib_ops/attention_lstm_op_test.cc|355 col 82| "lenghts" is a misspelling of "lengths"
onnxruntime/test/contrib_ops/attention_lstm_op_test.cc|392 col 82| "lenghts" is a misspelling of "lengths"
onnxruntime/test/contrib_ops/attention_lstm_op_test.cc|431 col 82| "lenghts" is a misspelling of "lengths"
onnxruntime/test/contrib_ops/attention_lstm_op_test.cc|470 col 82| "lenghts" is a misspelling of "lengths"
onnxruntime/test/contrib_ops/attention_lstm_op_test.cc|524 col 82| "lenghts" is a misspelling of "lengths"
onnxruntime/test/contrib_ops/attention_lstm_op_test.cc|581 col 84| "lenghts" is a misspelling of "lengths"
onnxruntime/test/ir/schema_registry_manager_test.cc|92 col 23| "registery" is a misspelling of "registry"
onnxruntime/test/ir/schema_registry_manager_test.cc|129 col 73| "targetted" is a misspelling of "targeted"
onnxruntime/test/ir/graph_test.cc|467 col 29| "verison" is a misspelling of "version"
onnxruntime/test/ir/graph_test.cc|484 col 22| "funciton" is a misspelling of "function"
onnxruntime/test/ir/graph_test.cc|484 col 43| "untill" is a misspelling of "until"
onnxruntime/test/optimizer/runtime_optimization/graph_runtime_optimization_test.cc|71 col 10| "NTO" is a misspelling of "NOT"
onnxruntime/test/optimizer/runtime_optimization/graph_runtime_optimization_test.cc|72 col 4| "NTO" is a misspelling of "NOT"
onnxruntime/test/optimizer/runtime_optimization/graph_runtime_optimization_test.cc|72 col 25| "NTO" is a misspelling of "NOT"
onnxruntime/test/optimizer/runtime_optimization/graph_runtime_optimization_test.cc|73 col 4| "NTO" is a misspelling of "NOT"
onnxruntime/test/optimizer/runtime_optimization/graph_runtime_optimization_test.cc|73 col 25| "NTO" is a misspelling of "NOT"
onnxruntime/test/optimizer/runtime_optimization/graph_runtime_optimization_test.cc|74 col 4| "NTO" is a misspelling of "NOT"
onnxruntime/test/optimizer/runtime_optimization/graph_runtime_optimization_test.cc|74 col 25| "NTO" is a misspelling of "NOT"
onnxruntime/test/optimizer/transpose_optimizer_test.cc|4660 col 92| "cancelled" is a misspelling of "canceled"
onnxruntime/test/perftest/ReadMe.txt|13 col 89| "direcotry" is a misspelling of "directory"
onnxruntime/test/perftest/ort_test_session.cc|918 col 49| "overriden" is a misspelling of "overridden"
onnxruntime/tool/etw/main.cc|22 col 49| "CANCELLED" is a misspelling of "CANCELED"
onnxruntime/tool/etw/compare_two_sessions.cc|29 col 49| "CANCELLED" is a misspelling of "CANCELED"
onnxruntime/tool/etw/eparser.cc|10 col 42| "attribue" is a misspelling of "attribute"
onnxruntime/tool/etw/eparser.cc|19 col 40| "attribue" is a misspelling of "attribute"
onnxruntime/contrib_ops/cuda/bert/longformer_attention_softmax.cu|151 col 27| "agressive" is a misspelling of "aggressive"
onnxruntime/contrib_ops/cuda/bert/decoder_attention.cc|285 col 5| "calcualte" is a misspelling of "calculate"
onnxruntime/contrib_ops/cuda/bert/embed_layer_norm_impl.cu|144 col 54| "toghether" is a misspelling of "together"
onnxruntime/contrib_ops/cuda/bert/tensorrt_fused_multihead_attention/cross_attention/fmha_cross_attention.h|44 col 63| "lenghts" is a misspelling of "lengths"
onnxruntime/contrib_ops/cuda/bert/tensorrt_fused_multihead_attention/cross_attention/fmha_cross_attention.h|72 col 63| "lenghts" is a misspelling of "lengths"
onnxruntime/contrib_ops/cuda/collective/sharding_spec.h|41 col 9| "proccess" is a misspelling of "process"
onnxruntime/contrib_ops/cuda/math/cufft_plan_cache.h|37 col 5| "contenst" is a misspelling of "contents"
onnxruntime/contrib_ops/cuda/math/cufft_plan_cache.h|54 col 5| "contenst" is a misspelling of "contents"
onnxruntime/contrib_ops/cuda/math/fft_ops.cc|90 col 48| "dimention" is a misspelling of "dimension"
onnxruntime/contrib_ops/cpu/transformers/sampling_cpu_helper.h|155 col 18| "presense" is a misspelling of "presence"
onnxruntime/contrib_ops/cpu/inverse.cc|56 col 51| "containes" is a misspelling of "contains"
onnxruntime/contrib_ops/cpu/cdist.cc|70 col 102| "optimising" is a misspelling of "optimizing"
onnxruntime/contrib_ops/cpu/attnlstm/deep_cpu_attn_lstm.h|22 col 7| "equivilent" is a misspelling of "equivalent"
onnxruntime/contrib_ops/cpu/attnlstm/attention_wrapper.cc|60 col 26| "calulated" is a misspelling of "calculated"
onnxruntime/core/mlas/lib/convolve.cpp|64 col 53| "underyling" is a misspelling of "underlying"
onnxruntime/core/mlas/lib/convolve.cpp|270 col 53| "underyling" is a misspelling of "underlying"
onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorEinSum.cpp|171 col 22| "accomodate" is a misspelling of "accommodate"
onnxruntime/core/providers/dml/DmlExecutionProvider/src/Operators/DmlOperatorFusedMatMul.cpp|47 col 67| "becuase" is a misspelling of "because"
onnxruntime/core/providers/migraphx/hip_allocator.cc|45 col 50| "appropiate" is a misspelling of "appropriate"
onnxruntime/core/providers/cuda/reduction/reduction_ops.cc|229 col 58| "comparsion" is a misspelling of "comparison"
onnxruntime/core/providers/cuda/cuda_allocator.cc|63 col 50| "appropiate" is a misspelling of "appropriate"
onnxruntime/core/providers/cuda/math/softmax_blockwise_impl.cuh|33 col 75| "utilisation" is a misspelling of "utilization"
onnxruntime/core/providers/cuda/tensor/resize_impl.cu|267 col 20| "occured" is a misspelling of "occurred"
onnxruntime/core/providers/cuda/tensor/resize_impl.cu|272 col 18| "occured" is a misspelling of "occurred"
onnxruntime/core/providers/cuda/tensor/resize_impl.cu|275 col 34| "occured" is a misspelling of "occurred"
onnxruntime/core/providers/cuda/tensor/transpose_impl.cu|148 col 37| "luanching" is a misspelling of "launching"
onnxruntime/core/providers/cuda/tensor/transpose_impl.cu|264 col 37| "luanching" is a misspelling of "launching"
onnxruntime/core/providers/cuda/nn/conv.cc|14 col 47| "versoin" is a misspelling of "version"
onnxruntime/core/providers/cuda/nn/conv.cc|218 col 31| "expecially" is a misspelling of "especially"
onnxruntime/core/providers/cuda/object_detection/roialign_impl.cu|119 col 35| "compatiblity" is a misspelling of "compatibility"
onnxruntime/core/providers/dnnl/dnnl_node_capability.cc|434 col 82| "gaurantee" is a misspelling of "guarantee"
onnxruntime/core/providers/dnnl/dnnl_node_capability.cc|468 col 10| "dimentions" is a misspelling of "dimensions"
onnxruntime/core/providers/dnnl/dnnl_node_capability.cc|618 col 30| "dimentional" is a misspelling of "dimensional"
onnxruntime/core/providers/dnnl/subgraph/dnnl_transpose.cc|59 col 89| "dimentions" is a misspelling of "dimensions"
onnxruntime/core/providers/dnnl/subgraph/dnnl_transpose.cc|68 col 32| "dimentions" is a misspelling of "dimensions"
onnxruntime/core/providers/dnnl/subgraph/dnnl_matmul.cc|110 col 36| "dimentions" is a misspelling of "dimensions"
onnxruntime/core/providers/dnnl/subgraph/dnnl_matmul.cc|130 col 36| "dimentions" is a misspelling of "dimensions"
onnxruntime/core/providers/dnnl/subgraph/dnnl_matmul.cc|177 col 79| "propogation" is a misspelling of "propagation"
onnxruntime/core/providers/dnnl/subgraph/dnnl_matmul.cc|178 col 82| "physcial" is a misspelling of "physical"
onnxruntime/core/providers/dnnl/subgraph/dnnl_dequantizelinear.cc|71 col 66| "substract" is a misspelling of "subtract"
onnxruntime/core/providers/dnnl/subgraph/dnnl_conv.h|35 col 16| "infered" is a misspelling of "inferred"
onnxruntime/core/providers/dnnl/subgraph/dnnl_conv.h|37 col 64| "infered" is a misspelling of "inferred"
onnxruntime/core/providers/dnnl/subgraph/dnnl_conv.h|50 col 42| "infered" is a misspelling of "inferred"
onnxruntime/core/providers/dnnl/subgraph/dnnl_conv.h|52 col 43| "infered" is a misspelling of "inferred"
onnxruntime/core/providers/dnnl/subgraph/dnnl_conv.h|69 col 16| "dimentions" is a misspelling of "dimensions"
onnxruntime/core/providers/dnnl/subgraph/dnnl_conv.h|69 col 85| "dimention" is a misspelling of "dimension"
onnxruntime/core/providers/dnnl/subgraph/dnnl_conv.h|70 col 37| "dimention" is a misspelling of "dimension"
onnxruntime/core/providers/dnnl/subgraph/dnnl_convgrad.h|42 col 42| "infered" is a misspelling of "inferred"
onnxruntime/core/providers/dnnl/subgraph/dnnl_convgrad.h|44 col 43| "infered" is a misspelling of "inferred"
onnxruntime/core/providers/dnnl/subgraph/dnnl_convgrad.h|52 col 16| "dimentions" is a misspelling of "dimensions"
onnxruntime/core/providers/dnnl/subgraph/dnnl_convgrad.h|52 col 85| "dimention" is a misspelling of "dimension"
onnxruntime/core/providers/dnnl/subgraph/dnnl_convgrad.h|53 col 37| "dimention" is a misspelling of "dimension"
onnxruntime/core/providers/dnnl/subgraph/dnnl_reduce.cc|138 col 23| "equivelent" is a misspelling of "equivalent"
onnxruntime/core/providers/dnnl/subgraph/dnnl_reduce.cc|143 col 26| "equivelent" is a misspelling of "equivalent"
onnxruntime/core/providers/dnnl/subgraph/dnnl_reduce.cc|147 col 26| "equivelent" is a misspelling of "equivalent"
onnxruntime/core/providers/dnnl/subgraph/dnnl_reduce.cc|301 col 31| "directy" is a misspelling of "directly"
onnxruntime/core/providers/dnnl/subgraph/dnnl_reduce.cc|301 col 47| "efficent" is a misspelling of "efficient"
onnxruntime/core/providers/dnnl/subgraph/dnnl_subgraph_primitive.h|68 col 63| "dimentional" is a misspelling of "dimensional"
onnxruntime/core/providers/dnnl/subgraph/dnnl_convgrad.cc|43 col 3| "acheive" is a misspelling of "achieve"
onnxruntime/core/providers/dnnl/dnnl_execution_provider.cc|348 col 16| "expectes" is a misspelling of "expects"
onnxruntime/core/providers/rknpu/rknpu_execution_provider.cc|31 col 26| "Exection" is a misspelling of "Execution"
onnxruntime/core/providers/rknpu/rknpu_execution_provider.cc|285 col 40| "Exection" is a misspelling of "Execution"
onnxruntime/core/providers/rknpu/rknpu_execution_provider.cc|285 col 62| "Exection" is a misspelling of "Execution"
onnxruntime/core/providers/nnapi/nnapi_builtin/nnapi_lib/nnapi_implementation.cc|231 col 36| "behaviour" is a misspelling of "behavior"
onnxruntime/core/providers/nnapi/nnapi_builtin/nnapi_lib/nnapi_implementation.h|2101 col 5| "simulatenously" is a misspelling of "simultaneously"
... (Too many findings. Dropped some findings)