| /external/pytorch/benchmarks/dynamo/ |
| D | expected_ci_speedup_inductor_torchbench_cpu.csv | 2 #timm_vision_transformer,inductor,float32,static,default,1.039510755
3 phlippe_densenet,inductor,float32,static,default,1.46474287
4 basic_gnn_edgecnn,inductor,float32,dynamic,default,1.30092957
5 llama_v2_7b_16h,inductor,float32,dynamic,default,1.23234331
6 resnet50,inductor,float32,dynamic,default,1.67742767
7 #timm_efficientnet,inductor,float32,static,cpp,
8 mobilenet_v3_large,inductor,float32,static,cpp,2.63311706
9 timm_resnest,inductor,float32,dynamic,cpp,1.7321529
10 functorch_maml_omniglot,inductor,float32,dynamic,cpp,1.17617472
11 #hf_GPT2,inductor,float32,dynamic,cpp,
[all …]
|
| D | README.md | 27 The [inductor-perf-test-nightly.yml](https://github.com/pytorch/pytorch/actions/workflows/inductor-…
30 … of the [workflow](https://github.com/pytorch/pytorch/actions/workflows/inductor-perf-test-nightly…
67 - `--backend=inductor`: selects TorchInductor as the compiler backend to measure. Many more are av…
74 - `--export-aot-inductor`: benchmarks ahead-of-time compilation mode.
83 ./benchmarks/dynamo/torchbench.py --performance --training --amp --backend=inductor --output=torchb…
84 ./benchmarks/dynamo/torchbench.py --performance --inference --bfloat16 --backend=inductor --output=…
86 ./benchmarks/dynamo/huggingface.py --performance --training --amp --backend=inductor --output=huggi…
87 ./benchmarks/dynamo/huggingface.py --performance --inference --bfloat16 --backend=inductor --output…
89 ./benchmarks/dynamo/timm_models.py --performance --training --amp --backend=inductor --output=timm_…
90 ./benchmarks/dynamo/timm_models.py --performance --inference --bfloat16 --backend=inductor --output…
|
| /external/pytorch/.github/workflows/ |
| D | inductor-cu124.yml | 1 name: inductor-cu124
6 - ciflow/inductor-cu124/*
21 linux-focal-cuda12_4-py3_10-gcc9-inductor-build:
22 # Should be synced with the one in inductor.yml, but this doesn't run inductor_timm
26 sync-tag: linux-focal-cuda12_4-py3_10-gcc9-inductor-build
28 docker-image-name: pytorch-linux-focal-cuda12.4-cudnn9-py3-gcc9-inductor-benchmarks
32 { config: "inductor", shard: 1, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
33 { config: "inductor", shard: 2, num_shards: 2, runner: "linux.g5.4xlarge.nvidia.gpu" },
53 linux-focal-cuda12_4-py3_10-gcc9-inductor-test:
56 needs: linux-focal-cuda12_4-py3_10-gcc9-inductor-build
[all …]
|
| D | inductor.yml | 1 name: inductor
9 - ciflow/inductor/*
30 linux-focal-cuda12_1-py3_10-gcc9-inductor-build:
36 docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-inductor-benchmarks
41 …{ config: "inductor", shard: 1, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-typ…
42 …{ config: "inductor", shard: 2, num_shards: 2, runner: "${{ needs.get-label-type.outputs.label-typ…
64 linux-focal-cuda12_1-py3_10-gcc9-inductor-test:
67 needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-build
70 … docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.docker-image }}
71 test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.test-matrix }}
[all …]
|
| D | inductor-micro-benchmark-x86.yml | 1 name: inductor-micro-benchmark-x86
8 - ciflow/inductor-micro-benchmark-cpu-x86/*
19 linux-jammy-cpu-py3_9-gcc11-inductor-build:
20 name: linux-jammy-cpu-py3.9-gcc11-inductor
24 docker-image-name: pytorch-linux-jammy-py3.9-gcc11-inductor-benchmarks
28 …{ config: "inductor-micro-benchmark-cpu-x86", shard: 1, num_shards: 1, runner: "linux.24xl.spr-met…
31 linux-jammy-cpu-py3_9-gcc11-inductor-micro-benchmark-test:
32 name: linux-jammy-cpu-py3.9-gcc11-inductor
34 needs: linux-jammy-cpu-py3_9-gcc11-inductor-build
37 docker-image: ${{ needs.linux-jammy-cpu-py3_9-gcc11-inductor-build.outputs.docker-image }}
[all …]
|
| D | inductor-rocm.yml | 1 name: inductor-rocm
15 - ciflow/inductor-rocm/*
25 linux-focal-rocm6_2-py3_10-inductor-build:
26 name: rocm6.2-py3.10-inductor
33 { config: "inductor", shard: 1, num_shards: 2, runner: "linux.rocm.gpu.2" },
34 { config: "inductor", shard: 2, num_shards: 2, runner: "linux.rocm.gpu.2" },
37 linux-focal-rocm6_2-py3_10-inductor-test:
41 name: rocm6.2-py3.10-inductor
43 needs: linux-focal-rocm6_2-py3_10-inductor-build
46 docker-image: ${{ needs.linux-focal-rocm6_2-py3_10-inductor-build.outputs.docker-image }}
[all …]
|
| D | inductor-perf-test-nightly-x86.yml | 1 name: inductor-perf-nightly-x86
51 linux-jammy-cpu-py3_9-gcc11-inductor-build:
52 name: linux-jammy-cpu-py3.9-gcc11-inductor
56 docker-image-name: pytorch-linux-jammy-py3.9-gcc11-inductor-benchmarks
77 linux-jammy-cpu-py3_9-gcc11-inductor-test-nightly:
78 name: linux-jammy-cpu-py3.9-gcc11-inductor
80 needs: linux-jammy-cpu-py3_9-gcc11-inductor-build
85 docker-image: ${{ needs.linux-jammy-cpu-py3_9-gcc11-inductor-build.outputs.docker-image }}
86 test-matrix: ${{ needs.linux-jammy-cpu-py3_9-gcc11-inductor-build.outputs.test-matrix }}
93 linux-jammy-cpu-py3_9-gcc11-inductor-test:
[all …]
|
| D | inductor-micro-benchmark.yml | 1 name: inductor-micro-benchmark
8 - ciflow/inductor-micro-benchmark/*
19 linux-focal-cuda12_1-py3_10-gcc9-inductor-micro-benchmark-build:
24 docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-inductor-benchmarks
28 { config: "inductor-micro-benchmark", shard: 1, num_shards: 1, runner: "linux.gcp.a100" },
31 linux-focal-cuda12_1-py3_10-gcc9-inductor-micro-benchmark-test:
34 needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-micro-benchmark-build
37 …docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-micro-benchmark-build.outputs.do…
38 …test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-micro-benchmark-build.outputs.tes…
|
| D | inductor-perf-test-nightly.yml | 1 name: inductor-A100-perf-nightly
69 linux-focal-cuda12_1-py3_10-gcc9-inductor-build:
74 docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-inductor-benchmarks
95 linux-focal-cuda12_1-py3_10-gcc9-inductor-test-nightly:
98 needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-build
103 … docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.docker-image }}
104 test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.test-matrix }}
110 linux-focal-cuda12_1-py3_10-gcc9-inductor-test-weekly:
113 needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-build
118 … docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.docker-image }}
[all …]
|
| D | inductor-perf-compare.yml | 1 name: inductor-A100-perf-compare
6 - ciflow/inductor-perf-compare/*
16 linux-focal-cuda12_1-py3_10-gcc9-inductor-build:
21 docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-inductor-benchmarks
33 linux-focal-cuda12_1-py3_10-gcc9-inductor-test:
36 needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-build
39 … docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.docker-image }}
40 test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.test-matrix }}
|
| D | inductor-perf-test-nightly-aarch64.yml | 1 name: inductor-perf-nightly-aarch64
53 linux-jammy-aarch64-py3_10-inductor-build:
54 name: linux-jammy-aarch64-py3.10-inductor
59 docker-image-name: pytorch-linux-jammy-aarch64-py3.10-gcc11-inductor-benchmarks
104 linux-jammy-aarch64-py3_10-inductor-test-nightly:
105 name: linux-jammy-aarch64-py3.10-inductor
107 needs: linux-jammy-aarch64-py3_10-inductor-build
114 docker-image: ${{ needs.linux-jammy-aarch64-py3_10-inductor-build.outputs.docker-image }}
115 test-matrix: ${{ needs.linux-jammy-aarch64-py3_10-inductor-build.outputs.test-matrix }}
122 linux-jammy-aarch64-py3_10-inductor-test:
[all …]
|
| D | inductor-perf-test-nightly-a10g.yml | 1 name: inductor-perf-nightly-A10g
71 linux-focal-cuda12_1-py3_10-gcc9-inductor-build:
76 docker-image-name: pytorch-linux-focal-cuda12.1-cudnn9-py3-gcc9-inductor-benchmarks
97 linux-focal-cuda12_1-py3_10-gcc9-inductor-test-nightly:
100 needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-build
105 … docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.docker-image }}
106 test-matrix: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.test-matrix }}
112 linux-focal-cuda12_1-py3_10-gcc9-inductor-test:
115 needs: linux-focal-cuda12_1-py3_10-gcc9-inductor-build
120 … docker-image: ${{ needs.linux-focal-cuda12_1-py3_10-gcc9-inductor-build.outputs.docker-image }}
[all …]
|
| /external/pytorch/test/inductor/ |
| D | test_decompose_mem_bound_mm.py | 1 # Owner(s): ["module: inductor"]
88 torch._logging.set_logs(inductor=logging.DEBUG)
104 counters["inductor"]["decompose_bmm"],
115 counters["inductor"]["decompose_bmm"],
126 torch._logging.set_logs(inductor=logging.DEBUG)
142 counters["inductor"]["decompose_addmm"],
147 counters["inductor"]["decompose_mm"],
150 decompose_mm_fwd = counters["inductor"]["decompose_mm"]
159 counters["inductor"]["decompose_mm"] - decompose_mm_fwd,
173 torch._logging.set_logs(inductor=logging.DEBUG)
[all …]
|
| D | test_codecache.py | 1 # Owner(s): ["module: inductor"]
148 self.assertEqual(counters["inductor"]["fxgraph_cache_miss"], 1)
149 self.assertEqual(counters["inductor"]["fxgraph_cache_hit"], 0)
150 self.assertEqual(counters["inductor"]["fxgraph_lookup_write_file"], 0)
158 self.assertEqual(counters["inductor"]["fxgraph_cache_miss"], 1)
159 self.assertEqual(counters["inductor"]["fxgraph_cache_hit"], 1)
160 self.assertEqual(counters["inductor"]["fxgraph_lookup_write_file"], 1)
224 self.assertGreater(counters["inductor"]["fxgraph_cache_miss"], 0)
225 self.assertEqual(counters["inductor"]["fxgraph_cache_hit"], 0)
232 self.assertEqual(counters["inductor"]["fxgraph_cache_miss"], 0)
[all …]
|
| D | test_select_algorithm.py | 1 # Owner(s): ["module: inductor"]
67 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 1)
84 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 1)
101 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 1)
113 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 1)
127 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 1)
140 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 0)
153 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 1)
165 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 1)
179 self.assertEqual(counters["inductor"]["select_algorithm_autotune"], 1)
[all …]
|
| D | test_standalone_compile.py | 1 # Owner(s): ["module: inductor"]
3 from torch import _dynamo as dynamo, _inductor as inductor unknown
45 mod_opt = inductor.compile(symbolic_trace(mod), [inp])
53 mod_opt = inductor.compile(symbolic_trace(mod), [inp])
61 mod_opt = inductor.compile(symbolic_trace(mod), [inp])
69 mod_opt = inductor.compile(make_fx(mod)(inp), [inp])
78 mod_opt = inductor.compile(mod, [inp])
87 mod_opt = inductor.compile(gm, [inp])
96 mod_opt = inductor.compile(gm, [inp])
109 mod_opt = inductor.compile(mod, inp)
|
| /external/pytorch/benchmarks/dynamo/microbenchmarks/ |
| D | matmul_relu.py | 11 @torch._dynamo.optimize("inductor", nopython=True)
58 time_with_torch_timer(inductor_mm, (a, b), string_id="inductor mm")
68 inductor mm mean: 0.0653 ms
72 inductor mm mean: 0.0252 ms
76 inductor mm mean: 0.0274 ms
80 inductor mm mean: 0.0244 ms
84 inductor mm mean: 0.0290 ms
88 inductor mm mean: 0.0319 ms
92 inductor mm mean: 0.0255 ms
96 inductor mm mean: 0.5090 ms
[all …]
|
| D | inductor_mm.py | 17 @torch._dynamo.optimize("inductor", nopython=True)
22 @torch._dynamo.optimize("inductor", nopython=True)
36 print("shape; torch mm; triton mm; inductor aten mm; inductor triton mm")
65 print("shape; torch mm; triton mm; inductor aten mm; inductor triton mm")
115 shape; torch mm; triton mm; inductor aten mm; inductor triton mm
127 shape; torch mm; triton mm; inductor aten mm; inductor triton mm
|
| /external/pytorch/torch/csrc/inductor/aoti_eager/ |
| D | kernel_holder.h | 9 #include <torch/csrc/inductor/aoti_eager/kernel_meta_info.h>
10 #include <torch/csrc/inductor/aoti_runner/model_container_runner.h>
15 namespace torch::inductor {
46 // The AOTIPythonKernelHolder class uses the AOT Inductor to generate a kernel
52 // Inductor is called again to generate the kernel library.
94 // Invoke python utility function on the Inductor side to produce AOTI kernel
96 // Inductor utility function -
102 // Invoke python utility function on the Inductor side to load AOTI kernel for
104 // Inductor utility function - torch._inductor.utils.load_aoti_eager_cache
111 } // namespace torch::inductor
|
| /external/pytorch/torch/_logging/ |
| D | _registrations.py | 20 register_log("inductor", ["torch._inductor", "torch._inductor.cudagraph_trees"])
24 "Logs information from wrapping inductor generated code with cudagraphs.",
85 …generated by AOTDispatch, after partitioning. Useful to understand what's being given to Inductor",
99 …nerated by post grad passes. Useful to understand what's being given to Inductor after post grad p…
139 "Prints the code that Inductor generates (either Triton or C++)",
145 "Prints the code that Inductor generates (on a per-kernel basis)",
151 "Inductor scheduler information. Useful if working on Inductor fusion algo",
158 "Detailed Inductor fusion decisions. More detailed than 'schedule'",
168 "Detailed Inductor compute/comm overlap decisions",
188 "Detailed Inductor benchmarking information.",
|
| /external/pytorch/.github/ |
| D | pytorch-probot.yml | 8 - ciflow/inductor
9 - ciflow/inductor-rocm
10 - ciflow/inductor-perf-compare
11 - ciflow/inductor-micro-benchmark
12 - ciflow/inductor-micro-benchmark-cpu-x86
13 - ciflow/inductor-cu124
|
| /external/pytorch/test/distributed/_composable/ |
| D | test_replicate_with_compiler.py | 12 from torch import _inductor as inductor, nn unknown
66 return inductor.compile(gm_, example_inputs_)
76 A version of MultiProcessTestCase that derives from the Inductor TestCase
77 to handle isolation of the inductor cache dir.
219 @unittest.skipIf(not has_triton(), "Inductor+gpu needs triton and recent GPU arch")
226 @unittest.skipIf(not has_triton(), "Inductor+gpu needs triton and recent GPU arch")
233 @unittest.skipIf(not has_triton(), "Inductor+gpu needs triton and recent GPU arch")
247 @unittest.skipIf(not has_triton(), "Inductor+gpu needs triton and recent GPU arch")
259 # TODO: figure out why we need to disable Inductor to avoid test errors.
264 @unittest.skipIf(not has_triton(), "Inductor+gpu needs triton and recent GPU arch")
[all …]
|
| /external/pytorch/.github/scripts/ |
| D | drci_mocks.json.gz | |
| /external/pytorch/torch/utils/_sympy/ |
| D | symbol.py | 26 # Inductor: The intermediates in inner_fn tmp0, one generated per ops call.
30 # Inductor: Placeholder variable that is later replaced with TMP
32 # Inductor: Some size expressions are replaced with a precomputed size ps0
36 # Inductor: An indexing variable i0 in loops IR which ranges over non-reduced
39 # Inductor: A reduction indexing r0 variable in loops IR which ranges over
42 # Inductor: In templated kernels torch._inductor.kernel, we have a hook to
47 # Inductor: iteration domain for blockIdx.x/blockIdx.y
50 # Inductor: this is used solely for dynamic_reshape_indexer
|
| /external/pytorch/.ci/pytorch/ |
| D | test.sh | 305 --exclude-inductor-tests \
317 python test/run_test.py -i inductor/test_torchinductor.py -k test_multi_gpu --verbose
318 python test/run_test.py -i inductor/test_aot_inductor.py -k test_non_default_cuda_device --verbose
319 python test/run_test.py -i inductor/test_aot_inductor.py -k test_replicate_on_devices --verbose
349 python test/run_test.py --inductor \
354 …# Do not add --inductor for the following inductor unit tests, otherwise we will fail because of n…
356 …--include inductor/test_torchinductor inductor/test_torchinductor_opinfo inductor/test_aot_inducto…
377 echo "Testing Inductor cpp wrapper mode with TORCHINDUCTOR_ABI_COMPATIBLE=1"
379 PYTORCH_TESTING_DEVICE_ONLY_FOR="" python test/run_test.py --include inductor/test_cpu_cpp_wrapper
380 python test/run_test.py --include inductor/test_cuda_cpp_wrapper
[all …]
|