Microbenchmarking — pytorch Architecture
Granular timing for individual operator or layer tests
Entity Profile
Relationship Graph
Domain
Functions
- AssertionError()
- AssertionError()
- __init__()
- __init__()
- __init__()
- __init__()
- __init__()
- __init__()
- __init__()
- __init__()
- __init__()
- __init__()
- __init__()
- __post_init__()
- _accuracy()
- _download_model()
- _normalize_bench_inputs()
- _produce_dynamic_shapes_for_export()
- _register_dataclass_output_as_pytree()
- _run_torchbench_from_args()
- _run_torchbench_model()
- _write_results_to_json()
- a_shapes()
- abs_norm()
- add()
- amp_dtype_bfloat16()
- args()
- args()
- batch_size_finder()
- benchmark_single_shape_for_backend()
- cast_based_on_args()
- cast_to_bf16()
- cast_to_fp64()
- check_accuracy()
- check_accuracy()
- check_csv()
- check_graph_breaks()
- check_tolerance()
- common()
- compiled()
- compiled()
- compiled()
- compiled()
- compiled()
- compiled()
- compute_loss()
- compute_mean_rstd()
- coverage_experiment()
- disable_cudagraph_models()
- eager()
- eager()
- eager()
- empty_gpu_cache()
- exit_after()
- export()
- export_nativert()
- failing_fx2trt_models()
- fb()
- fn()
- fns()
- force_amp_for_fp16_bf16_models()
- force_fp16_for_bf16_models()
- forward()
- fp32_only_models()
- fsdp_checkpointing_base()
- fusion_type()
- generate_inputs_for_model()
- get_compile_time()
- get_field()
- get_field()
- get_fsdp_auto_wrap_policy()
- get_iou_threshold()
- get_memory_bytes()
- get_memory_bytes()
- get_memory_bytes()
- get_memory_bytes()
- get_memory_bytes()
- get_output_amp_train_process_func()
- get_sequence_length()
- get_shapes()
- get_shapes()
- get_shapes()
- get_shapes()
- get_suite_from_model_iter_fn()
- gn()
- help()
- huge_graph()
- huge_graph()
- huggingface_llm_models()
- inductor_aten_mm()
- inductor_scatter_add()
- inductor_triton_bmm()
- init_optimizer()
- inline_inbuilt_nn_modules_models()
- iter_model_names()
- iter_models()
- latency_experiment_summary()
- liger()
- liger()
- liger()
- liger()
- load()
- load()
- load_model_from_path()
- load_yaml_file()
- log_operator_inputs()
- longest_common_prefix()
- main()
- main()
- main()
- main()
- main()
- main()
- main()
- main()
- main()
- main()
- main()
- maybe_cast()
- maybe_init_distributed()
- maybe_mark_step()
- maybe_preserve_compile_debug()
- microbenchmark()
- mm_add()
- mm_add_relu()
- model_iter_fn()
- model_names()
- non_deterministic_models()
- nothing()
- null_experiment()
- open()
- optimize_templates()
- optimizer_zero_grad()
- output_csv()
- overhead_experiment()
- parse_args()
- parse_args()
- parse_cmd_args()
- parser()
- parser()
- pip_install()
- print()
- print()
- print_aten_ops()
- print_summary()
- print_summary_table()
- process_entry()
- process_hf_reformer_output()
- profile()
- pytorch()
- quack()
- quack()
- quack()
- quack()
- read_batch_size_from_file()
- recompile_profiler_experiment()
- report_geomean_speedup()
- rms_norm_ref()
- rms_norm_ref()
- run()
- run_model()
- run_performance_test()
- scale()
- setattr()
- setup()
- setup_determinism_for_accuracy_test()
- shapes()
- skip_accuracy_check_as_eager_non_deterministic()
- skip_models_due_to_control_flow()
- skip_models_for_cpu()
- skip_models_for_cpu()
- skip_models_for_cpu_aarch64()
- skip_models_for_cuda()
- skip_models_for_freezing_cpu()
- skip_models_for_freezing_cuda()
- skip_models_for_xpu()
- skip_not_suitable_for_training_models()
- slow_models()
- sum()
- summarize_graph_break()
- symbolic_convert_overhead_stress_test()
- test_total_time()
- test_total_time()
- test_total_time()
- timed()
- torch()
- torch()
- torch_mm()
- torch_mm_relu()
- torch_xla()
- torchscript_jit_trace()
- torchviz_model()
- triton()
- tti()
- use_iou_for_bool_accuracy()
- use_larger_multiplier_for_smaller_tensor()
- use_larger_multiplier_for_smaller_tensor()
- visualize_comparison()
- xla()
Frequently Asked Questions
What is the Microbenchmarking subdomain?
Microbenchmarking is a subdomain in the pytorch codebase, part of the DynamoBenchmarks domain. Granular timing for individual operator or layer tests It contains 0 source files.
Which domain does Microbenchmarking belong to?
Microbenchmarking belongs to the DynamoBenchmarks domain.
What functions are in Microbenchmarking?
The Microbenchmarking subdomain contains 199 function(s): AssertionError, AssertionError, __init__, __init__, __init__, __init__, __init__, __init__, and 191 more.
Analyze Your Own Codebase
Get architecture documentation, dependency graphs, and domain analysis for your codebase in minutes.
Try Supermodel Free