pyc-kernel-lab / inference /artifacts /data /03_ground_truth.json
Allanatrix's picture
Publish inference portal artifacts and sanitized plots
91279d3 verified
Raw History Blame Contribute Delete
5.1 kB
{
"status": "ok",
"meta": {
"device": "cuda",
"dtype": "torch.float16",
"hidden": 2048,
"hidden_effective": 2048,
"warmup": 20,
"iters": 120,
"bucket_lengths": [
256
],
"batch_sizes": [
64
],
"torch_compile": false,
"compile_mode": "reduce-overhead",
"compile_backend": "inductor",
"inductor_gemm_backends": "",
"inductor_gemm_backends_applied": false,
"allow_tf32": true,
"inference_mode": true,
"use_inplace_residual": true,
"pad_hidden_to": 16,
"use_cuda_graph_arena": false,
"single_shape_seq_len": 0,
"skip_batch_scaling": false,
"single_pass_after_warmup": false
},
"global_counters": {
"compile_calls": 0,
"graph_build_calls": 1
},
"bucket_results": [
{
"batch": 64,
"seq_len": 256,
"samples_ms": [
15.9429,
15.7665,
15.7304,
15.7467,
15.7367,
15.738,
15.7319,
15.737,
15.7349,
15.7461,
15.7352,
15.7354,
15.7378,
15.7434,
15.7315,
15.7399,
15.7312,
15.7378,
15.7373,
15.6223,
15.625,
15.6175,
15.6176,
15.6112,
15.6271,
15.615,
15.6278,
15.6073,
15.7262,
15.7417,
15.7207,
15.7334,
15.7325,
15.7482,
15.7397,
15.7416,
15.7395,
15.7391,
15.7395,
15.7571,
15.7672,
15.7491,
15.7439,
15.7374,
15.7409,
15.7355,
15.7334,
15.7374,
15.8271,
15.9085,
15.9924,
16.0947,
16.2181,
16.2091,
16.2112,
16.2214,
16.2135,
16.215,
16.1081,
16.0299,
15.9637,
15.9626,
15.9622,
15.9693,
15.9665,
15.9676,
15.9612,
15.8494,
15.8491,
15.8454,
15.8467,
15.8452,
15.8658,
15.9136,
15.7971,
15.7391,
15.7324,
15.7331,
15.7476,
15.7337,
15.7441,
15.7335,
15.7376,
15.7428,
15.744,
15.7347,
15.7351,
15.7337,
15.7332,
15.7373,
15.736,
15.7307,
15.7341,
15.6804,
15.8221,
15.7297,
15.7291,
15.7368,
15.7362,
15.7448,
15.7317,
15.7328,
15.7313,
15.7262,
15.7311,
15.7286,
15.725,
15.7398,
15.7301,
15.7367,
15.8377,
15.8488,
15.8526,
15.8613,
15.846,
15.8507,
15.8445,
15.8495,
15.8635,
15.8477
],
"latency_ms": {
"mean": 15.7974,
"p50": 15.7395,
"p95": 16.1081,
"min": 15.6073,
"max": 16.2214
},
"compile_calls_during_steady_state": 0,
"graph_build_calls_during_steady_state": 0,
"memory_pre": {
"allocated_bytes": 142774272,
"reserved_bytes": 694157312,
"peak_allocated_bytes": 679636992,
"allocation_events": 208,
"segment_alloc_events": 7
},
"memory_post": {
"allocated_bytes": 142774272,
"reserved_bytes": 694157312,
"peak_allocated_bytes": 679636992,
"allocation_events": 1408,
"segment_alloc_events": 7
},
"allocation_event_delta": 1200,
"segment_alloc_event_delta": 0,
"memory_stable": false
}
],
"batch_scaling": [
{
"batch": 64,
"seq_len": 256,
"samples_ms": [
15.853,
15.8565,
15.8498,
15.8425,
15.8513,
15.8473,
15.8429,
15.8548,
15.8463,
15.8502,
15.857,
15.8419,
15.8482,
15.8903,
15.9548,
15.7347,
15.7385,
15.7272,
15.7425,
15.7383,
15.7319,
15.7265,
15.7277,
15.7271,
15.7264,
15.7286,
15.7269,
15.7407,
15.7404,
15.8234
],
"latency_ms": {
"mean": 15.7989,
"p50": 15.8234,
"p95": 15.8903,
"min": 15.7264,
"max": 15.9548
},
"compile_calls_during_steady_state": 0,
"graph_build_calls_during_steady_state": 0,
"memory_pre": {
"allocated_bytes": 142774272,
"reserved_bytes": 694157312,
"peak_allocated_bytes": 679636992,
"allocation_events": 1508,
"segment_alloc_events": 9
},
"memory_post": {
"allocated_bytes": 142774272,
"reserved_bytes": 694157312,
"peak_allocated_bytes": 679636992,
"allocation_events": 1808,
"segment_alloc_events": 9
},
"allocation_event_delta": 300,
"segment_alloc_event_delta": 0,
"memory_stable": false
}
]
}