{ "metadata": { "width": 1344, "height": 768, "frames": 124, "steps": 12, "sampler_step": 1, "seed": 440420, "text_tokens": 100, "block_index": 24, "attention": "sage2", "hidden_shape": [ 37810, 5376 ], "h_msa_shape": [ 37810, 5376 ], "segments": [ [ 0, 100, 1 ], [ 100, 514, 2 ], [ 514, 37810, 0 ] ] }, "segments": [ [ 0, 100, 1 ], [ 100, 514, 2 ], [ 514, 37810, 0 ] ], "warmup": 3, "iterations": 5, "results": [ { "segment": "all", "segment_tuple": null, "backend": "sage2", "layout_mode": "hnd", "q_shape": [ 1, 37810, 56, 128 ], "output_shape": [ 37810, 7168 ], "timings": { "q_transpose_contiguous": { "count": 5, "mean_s": 0.005225043802056462, "p50_s": 0.005116002983413637, "p90_s": 0.005561528832186013, "p95_s": 0.0055678074364550415, "p99_s": 0.005572830319870263, "min_s": 0.0048384820111095905, "max_s": 0.005574086040724069 }, "k_transpose_contiguous": { "count": 5, "mean_s": 0.005020070588216186, "p50_s": 0.004905874957330525, "p90_s": 0.005268368811812252, "p95_s": 0.005289322405587882, "p99_s": 0.005306085280608386, "min_s": 0.004802018986083567, "max_s": 0.005310275999363512 }, "v_transpose_contiguous": { "count": 5, "mean_s": 0.004939833201933652, "p50_s": 0.004868466989137232, "p90_s": 0.005137888411991298, "p95_s": 0.0051993222092278305, "p99_s": 0.005248469247017056, "min_s": 0.00475086597725749, "max_s": 0.005260756006464362 }, "attention_kernel": { "count": 5, "mean_s": 0.2837455251952633, "p50_s": 0.2643922639545053, "p90_s": 0.32986050880281254, "p95_s": 0.3455152763985097, "p99_s": 0.3580390904750675, "min_s": 0.2538328950176947, "max_s": 0.36117004399420694 }, "output_reshape": { "count": 5, "mean_s": 0.0048204828053712845, "p50_s": 0.004661265993490815, "p90_s": 0.005151034612208605, "p95_s": 0.005268343805801124, "p99_s": 0.005362191160675139, "min_s": 0.004617601982317865, "max_s": 0.005385652999393642 } }, "layout_attention_total_mean_s": 0.3037509555928409, "reference_backend": "sage2:hnd", "reference_diff": { "max": 0.0, "mean": 0.0 }, "status": "ok" }, { "segment": "all", "segment_tuple": null, "backend": "sage2", "layout_mode": "sage_nhd", "q_shape": [ 1, 37810, 56, 128 ], "output_shape": [ 37810, 7168 ], "timings": { "q_nhd_contiguous": { "count": 5, "mean_s": 0.005327214219141752, "p50_s": 0.005574870039708912, "p90_s": 0.005795779020991177, "p95_s": 0.005863789014983922, "p99_s": 0.005918197010178119, "min_s": 0.004763827018905431, "max_s": 0.005931799008976668 }, "k_nhd_contiguous": { "count": 5, "mean_s": 0.004842757794540376, "p50_s": 0.004699617973528802, "p90_s": 0.005195734999142587, "p95_s": 0.005358382000122219, "p99_s": 0.005488499600905925, "min_s": 0.004638209997210652, "max_s": 0.0055210290011018515 }, "v_nhd_contiguous": { "count": 5, "mean_s": 0.004678740992676467, "p50_s": 0.004741937969811261, "p90_s": 0.004805538605432957, "p95_s": 0.004823794809635728, "p99_s": 0.004838399772997945, "min_s": 0.004513168998528272, "max_s": 0.0048420510138385 }, "attention_kernel": { "count": 5, "mean_s": 0.3031308094039559, "p50_s": 0.2796200959710404, "p90_s": 0.3695915882126428, "p95_s": 0.3924628216191195, "p99_s": 0.4107598083443008, "min_s": 0.25944606703706086, "max_s": 0.41533405502559617 }, "output_reshape": { "count": 5, "mean_s": 5.465606227517128e-06, "p50_s": 4.223955329507589e-06, "p90_s": 8.115207310765982e-06, "p95_s": 9.385612793266772e-06, "p99_s": 1.0401937179267406e-05, "min_s": 4.00003045797348e-06, "max_s": 1.0656018275767565e-05 } }, "layout_attention_total_mean_s": 0.317984988016542, "reference_backend": "sage2:hnd", "reference_diff": { "max": 0.0, "mean": 0.0 }, "status": "ok" }, { "segment": "all", "segment_tuple": null, "backend": "sage2", "layout_mode": "sage_strided_nhd", "q_shape": [ 1, 37810, 56, 128 ], "output_shape": [ 37810, 7168 ], "timings": { "attention_kernel": { "count": 5, "mean_s": 0.260253090003971, "p50_s": 0.2601712809992023, "p90_s": 0.2612823422066867, "p95_s": 0.26159851960837843, "p99_s": 0.26185146152973177, "min_s": 0.25904494104906917, "max_s": 0.2619146970100701 }, "output_reshape": { "count": 5, "mean_s": 4.3424079194664955e-06, "p50_s": 4.175992216914892e-06, "p90_s": 5.068804603070021e-06, "p95_s": 5.350401625037192e-06, "p99_s": 5.5756792426109315e-06, "min_s": 3.8080033846199512e-06, "max_s": 5.631998647004366e-06 } }, "layout_attention_total_mean_s": 0.2602574324118905, "reference_backend": "sage2:hnd", "reference_diff": { "max": 0.0, "mean": 0.0 }, "status": "ok" } ] }