|
|
|
@@ -17,15 +17,15 @@
|
|
|
|
|
{"kernel":"qsa_attention_p6_s0","sha256":["068422304f4191e74610562e007ffa7ec0d4b2ff5ed619997442ca1d254073b1","6f49058a22d7c62bbf4ffba4542b7aa32a130076b07537ac5605b2c85bb97162","158b3b53061c6d3d089049f9fdb47064a8c4f3d2f3fdb0ef5089afcaeba7cc5d","d4bc3f9184f206d1c86876a7ce98f086cf94f7912b460f7dc8c809b5b78d579b","011cc9bf7a9b2f86548e46728a9013e24273ca27ee50f9ee3f5509ecbc7c851f"],"profile":6,"step":0,"rows":2048,"pos":32768,"prefix":32768,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":32768,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["prefill_flash"],"counts":{"flash_kernel":1,"compiled_selector":1},"offset":34816,"pooled_len":8704,"state_present":[true,true,false,true,true],"shapes":[[1,2048,64],[1,65536,128],[1,16384,128],[1,2,34816,256],[1,2,34816,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":0,"qk_rows_calls":1,"traces":1,"selector_dispatches":1,"capacity_transitions":0,"prefill_score_producers":{"mpp":1,"mlx":0},"modes":{"blocks":0,"prefill_blocks":1,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw65536:pool16384:out0":1},"entry_count":1,"compiled_keys":[{"source":"qk_rows","mode":"prefill_blocks","input_shape":[1,2048,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"mpp","selector_chunk_rows":2048,"label":"qk_rows:prefill_blocks:s2048:raw65536:pool16384:out0:scorempp:chunk2048:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p6_s1","sha256":["bcbfcd4ff8d59b123b6ba9e84c2d621b9982c5e7430c84b82240f2b4416e48af","d50916defa6c9151ffeb706a7e0d6b8c04b861eaa94ed2b637f82fce79c8f2c9","158b3b53061c6d3d089049f9fdb47064a8c4f3d2f3fdb0ef5089afcaeba7cc5d","7c9ca3c2fd81b2f4749ce1f575a963b7054f004b91879a64212ecda5b0c34463","773fdcd29060aa9a94684db395aeccdbe1af803e191f3d0b938abc109d715fbb"],"profile":6,"step":1,"rows":1,"pos":34816,"prefix":32768,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":32768,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["flash"],"counts":{},"offset":34817,"pooled_len":8704,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,65536,128],[1,16384,128],[1,2,35072,256],[1,2,35072,256]],"core":{"calls":2,"compiled_calls":2,"hidden_calls":0,"qk_rows_calls":2,"traces":2,"selector_dispatches":2,"capacity_transitions":0,"prefill_score_producers":{"mpp":1,"mlx":0},"modes":{"blocks":1,"prefill_blocks":1,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw65536:pool16384:out0":2},"entry_count":2,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw65536:pool16384:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"prefill_blocks","input_shape":[1,2048,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"mpp","selector_chunk_rows":2048,"label":"qk_rows:prefill_blocks:s2048:raw65536:pool16384:out0:scorempp:chunk2048:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p6_s2","sha256":["e5f8b59d40e7e1c579d22d66af242e2882d1ab57727f17c7b9eeea894a5368e7","3b29542bd05bd5525466ad6c5951dd2d59759bed880564b8e07058d8feed1d42","9bc8c466e0a3164a6f4e22c0ffe23a895126f57945d4e7f0d5a07e79c63778ad","a41386c0de5defa3cc2eb26fe6005bab14865f573f9380f998d35e324abdbb43","270285c24ae2ea32e45240ada590a6cc1c9e7b14adfc23b270c7d7440c7372d4"],"profile":6,"step":2,"rows":4,"pos":34817,"prefix":32768,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":32768,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["rows_gather"],"counts":{},"offset":34821,"pooled_len":8705,"state_present":[true,true,false,true,true],"shapes":[[1,4,64],[1,65536,128],[1,16384,128],[1,2,35072,256],[1,2,35072,256]],"core":{"calls":3,"compiled_calls":3,"hidden_calls":0,"qk_rows_calls":3,"traces":3,"selector_dispatches":3,"capacity_transitions":0,"prefill_score_producers":{"mpp":1,"mlx":0},"modes":{"blocks":1,"prefill_blocks":1,"dense_mask":0,"row_tokens":1,"update_only":0},"buckets":{"raw65536:pool16384:out0":3},"entry_count":3,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw65536:pool16384:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"prefill_blocks","input_shape":[1,2048,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"mpp","selector_chunk_rows":2048,"label":"qk_rows:prefill_blocks:s2048:raw65536:pool16384:out0:scorempp:chunk2048:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,4,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"qk_rows:row_tokens:s4:raw65536:pool16384:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s0","sha256":["f9c3c5622d5d7c346d6c2881634894e60f030654828e74d96041317d5d1b99e5","fa52f9dff396320af135c22c25b328674b3c44b6c768df9ed2d54d31bbecc71f","5385335d390b6950b06db2bd3a3e86efe8b4ca7d48d7458828f6891d0815be26","f583e84d01f4e3bfb152b8855f8520d25c5fd1639c20961343d44d3b494ee18f","ea3bda4b6f89e3552003b8afb307a5d04a67cb8d024ea18b7a7df90a172f7342"],"profile":7,"step":0,"rows":1,"pos":2051,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":2052,"pooled_len":513,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":0,"qk_rows_calls":1,"traces":1,"selector_dispatches":1,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw4096:pool1024:out0":1},"entry_count":1,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s1","sha256":["96067951d1aa0a433bcc85d4c51f0dbbf8f142f2f97b0e4aa1f0396ce6af89de","c7e437185225097e4ce2bb012fb9dea74124314e3858db5b2d3f89d175628f8e","99fb41883cf58e2cec92897b8486a4b20b6817dc32c1b98d320280e2984756b4","4af7a650ea8717de842e55d3691237c8e2af5c5c8ac78868e1323773e1a792f1","f54917b64b4e33520dd53be4677edb0d92a41e62d868d44eff6fdbc030530df8"],"profile":7,"step":1,"rows":4,"pos":2052,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,4,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":2,"compiled_calls":2,"hidden_calls":0,"qk_rows_calls":2,"traces":2,"selector_dispatches":2,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":1,"update_only":0},"buckets":{"raw4096:pool1024:out0":2},"entry_count":2,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,4,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"qk_rows:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s2","sha256":["5c48c9fcdb884b10fecce90489b58457a977528d3732d9ceb90d202f9a8bcabf","0962884629f0854f29ed23c253f65f1d44fcb8bfde142e948411c9e0ed65faf9","5b6ab0a7cc9a3a122481547531253bcba5f039fc507718fc85bef86dd8dc56ea","57eb1075e3ed3fe889a40c34bd1e59974119725a3a9ef009646a3427b3be9dd6","7fe9f3c5a939bb7dd64a147ea63354dc92c40828f2cc4d1cf8b3be331b86f4e5"],"profile":7,"step":2,"rows":2,"pos":2054,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,2,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":3,"compiled_calls":3,"hidden_calls":0,"qk_rows_calls":3,"traces":3,"selector_dispatches":3,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":2,"update_only":0},"buckets":{"raw4096:pool1024:out0":3},"entry_count":3,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,2,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":2,"label":"qk_rows:row_tokens:s2:raw4096:pool1024:out0:scorenone:chunk2:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,4,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"qk_rows:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s3","sha256":["554527077cad0087f4b548f1cf28dfbf186762ce6e3e12982922b97eec865e1d","9424991712e6561268208093dc9c9df018591b376554ac208eab7f242bdba4e4","5b6ab0a7cc9a3a122481547531253bcba5f039fc507718fc85bef86dd8dc56ea","28b0c2bfcc8486f5c6daafec65e013f6d2efbf99b24cdf75ba25372716d828d9","6d3d7b11ac7cff92650bdc798b862cbd21b383cd0154f53eee439af721d3c55f"],"profile":7,"step":3,"rows":1,"pos":2056,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":2057,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":4,"compiled_calls":4,"hidden_calls":0,"qk_rows_calls":4,"traces":3,"selector_dispatches":4,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":2,"prefill_blocks":0,"dense_mask":0,"row_tokens":2,"update_only":0},"buckets":{"raw4096:pool1024:out0":4},"entry_count":3,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,2,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":2,"label":"qk_rows:row_tokens:s2:raw4096:pool1024:out0:scorenone:chunk2:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,4,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"qk_rows:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s0","sha256":["9fc17259582838c33640717e4e21cdd556e3f5d03bd8dda03de124d43f7b55e8","48e48f96c54776ebd1a034cdfcf93ae5e027038346bff0b1cc4f6600a73bdbd4","56ff514413104108eea40eeb10e28ba89d4bcb5f273e7bfb327b650239b2608e","d1f4df60a631b8170515760665a9212d6375de4e6456f75d90eb31179c91596c","f583e84d01f4e3bfb152b8855f8520d25c5fd1639c20961343d44d3b494ee18f","ea3bda4b6f89e3552003b8afb307a5d04a67cb8d024ea18b7a7df90a172f7342"],"profile":7,"step":0,"rows":1,"pos":2051,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["flash"],"counts":{},"offset":2052,"pooled_len":513,"state_present":[true,true,true,true,true],"shapes":[[1,1,64],[1,2304,128],[1,1024,128],[1,1,128,1024],[1,2,2304,256],[1,2,2304,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s1","sha256":["24e8ced9cf6525d3fb9499e8cb45e70960458b26278552ff5e3bdb825596f33a","1ceca1a52f5057c1aaee04689fc8026d3df9e1b657d718d75b9a72aeb8a88b25","8ee668bf4a7f6eca91380e543147f88eb73d8bd3384c1a375b869d4680d95753","9221cbebf9df5094116ab9269b2c0e543d03fcd7f7ec5e9c1c745fc7b2babd0e","4af7a650ea8717de842e55d3691237c8e2af5c5c8ac78868e1323773e1a792f1","f54917b64b4e33520dd53be4677edb0d92a41e62d868d44eff6fdbc030530df8"],"profile":7,"step":1,"rows":4,"pos":2052,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["rows_gather"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,true,true,true],"shapes":[[1,4,64],[1,2304,128],[1,1024,128],[1,1,128,1024],[1,2,2304,256],[1,2,2304,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s2","sha256":["6cd50cb32aab449013ff1db077d2e4a2f480fb71f1ae11ed80499dfab78257d0","b49eaa7452b6018be70df323222b0624f230df95ea900e42a974eaba5e2ac975","a8f7c3b9add9760148121cf236dab8d6844d38d42ab5aa4a7de20bb14c342d97","d4abe49bd48fe5eeea175dd7102faf5cc4a0bcf4831d233e7521c51cf7a76932","57eb1075e3ed3fe889a40c34bd1e59974119725a3a9ef009646a3427b3be9dd6","7fe9f3c5a939bb7dd64a147ea63354dc92c40828f2cc4d1cf8b3be331b86f4e5"],"profile":7,"step":2,"rows":2,"pos":2054,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["rows_gather"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,true,true,true],"shapes":[[1,2,64],[1,2304,128],[1,1024,128],[1,1,128,1024],[1,2,2304,256],[1,2,2304,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s3","sha256":["80cb2c6d1f34cc44dfbcd30192a17c8cb4a0c2a19017924058ece49e2f5c6486","59da7bcfd15c46c7c183a322559d2f455dba99f677bbef99b3626bd32160147f","a8f7c3b9add9760148121cf236dab8d6844d38d42ab5aa4a7de20bb14c342d97","d4abe49bd48fe5eeea175dd7102faf5cc4a0bcf4831d233e7521c51cf7a76932","28b0c2bfcc8486f5c6daafec65e013f6d2efbf99b24cdf75ba25372716d828d9","6d3d7b11ac7cff92650bdc798b862cbd21b383cd0154f53eee439af721d3c55f"],"profile":7,"step":3,"rows":1,"pos":2056,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["flash"],"counts":{},"offset":2057,"pooled_len":514,"state_present":[true,true,true,true,true],"shapes":[[1,1,64],[1,2304,128],[1,1024,128],[1,1,128,1024],[1,2,2304,256],[1,2,2304,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p8_s0","sha256":["2f2d1534efe9754560b2d2289f6c9756da611928c8c9e2321786ab03dd447111","84e64cfc1690832a3a6ed516b577d1001b89b382254c08b2663f493691330f2e","7da3f843c044b333d43259efd0a68f1c0a8f5230707898663f9ce096dd1f294a"],"profile":8,"step":0,"rows":3,"pos":0,"prefix":0,"fused":0,"vision":"none","no_indexer":true,"biased":true,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":3,"pooled_len":0,"state_present":[false,false,false,true,true],"shapes":[[1,3,64],[1,2,256,256],[1,2,256,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p8_s1","sha256":["dbe815a17715596ea030903a6800bba50c5ed74b9e03a3d29826e42acd57377b","50d2e91a257e44f7554a3ba6b1e0b0c0bef2a3f9ec4f1c4b6904e150f9b7b48c","a79deff710fa918973e19986ced87ae13347fbd66972c79fd4241d299f4472dc"],"profile":8,"step":1,"rows":1,"pos":3,"prefix":0,"fused":0,"vision":"none","no_indexer":true,"biased":true,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":4,"pooled_len":0,"state_present":[false,false,false,true,true],"shapes":[[1,1,64],[1,2,256,256],[1,2,256,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p8_s2","sha256":["faeacdecf2a8829ff7c7c5751858003e0931435cae65c610e5c362a3997ba050","d4c84ff24f73effb7d50f227eb1248f997cc7ea466b6a15e6fa2876cf9481ea0","af7554d1b0575d27146ecc0b2d3755a89a645e42231c214b62cbecfaf6d81402"],"profile":8,"step":2,"rows":4,"pos":4,"prefix":0,"fused":0,"vision":"none","no_indexer":true,"biased":true,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":8,"pooled_len":0,"state_present":[false,false,false,true,true],"shapes":[[1,4,64],[1,2,256,256],[1,2,256,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p9_s0","sha256":["91fd57e3f97be79622b57e34ed7e80b9934f65ce650592a6a619f0c88b322bd9","fa52f9dff396320af135c22c25b328674b3c44b6c768df9ed2d54d31bbecc71f","5385335d390b6950b06db2bd3a3e86efe8b4ca7d48d7458828f6891d0815be26","0823d3bc139c67f69911c8970fb2c1bcf5f757971e9191f8b98d56f6689840ae","ea3bda4b6f89e3552003b8afb307a5d04a67cb8d024ea18b7a7df90a172f7342"],"profile":9,"step":0,"rows":1,"pos":2051,"prefix":2051,"fused":3,"vision":"no_axes","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":2052,"pooled_len":513,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":1,"qk_rows_calls":0,"traces":1,"selector_dispatches":1,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw4096:pool1024:out0":1},"entry_count":1,"compiled_keys":[{"source":"hidden","mode":"blocks","input_shape":[1,1,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"hidden:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p9_s1","sha256":["893d7f0e21dc6a1c5099fa172b86630a0b694dc90ea7c6db4ae928c2151e9b00","c7e437185225097e4ce2bb012fb9dea74124314e3858db5b2d3f89d175628f8e","99fb41883cf58e2cec92897b8486a4b20b6817dc32c1b98d320280e2984756b4","e38b573ffab0799d27fcadc08d1e6a6aa31e40b4e6a61c5e9c98073a7012ba24","f54917b64b4e33520dd53be4677edb0d92a41e62d868d44eff6fdbc030530df8"],"profile":9,"step":1,"rows":4,"pos":2052,"prefix":2051,"fused":3,"vision":"no_axes","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["sdpa"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,4,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":2,"compiled_calls":2,"hidden_calls":2,"qk_rows_calls":0,"traces":2,"selector_dispatches":2,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":1,"update_only":0},"buckets":{"raw4096:pool1024:out0":2},"entry_count":2,"compiled_keys":[{"source":"hidden","mode":"blocks","input_shape":[1,1,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"hidden:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"hidden","mode":"row_tokens","input_shape":[1,4,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"hidden:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p9_s0","sha256":["6fd06b68b7059ae0220c03d7afada76bdadc86001123c4180f1a16380a89bece","fa52f9dff396320af135c22c25b328674b3c44b6c768df9ed2d54d31bbecc71f","5385335d390b6950b06db2bd3a3e86efe8b4ca7d48d7458828f6891d0815be26","0823d3bc139c67f69911c8970fb2c1bcf5f757971e9191f8b98d56f6689840ae","ea3bda4b6f89e3552003b8afb307a5d04a67cb8d024ea18b7a7df90a172f7342"],"profile":9,"step":0,"rows":1,"pos":2051,"prefix":2051,"fused":3,"vision":"no_axes","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["flash"],"counts":{},"offset":2052,"pooled_len":513,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":1,"qk_rows_calls":0,"traces":1,"selector_dispatches":1,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw4096:pool1024:out0":1},"entry_count":1,"compiled_keys":[{"source":"hidden","mode":"blocks","input_shape":[1,1,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"hidden:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p9_s1","sha256":["d8762135183d3a85dfd8713c2fda95cf6397f495329ee3f626a0fc56f3cb8e2d","c7e437185225097e4ce2bb012fb9dea74124314e3858db5b2d3f89d175628f8e","99fb41883cf58e2cec92897b8486a4b20b6817dc32c1b98d320280e2984756b4","e38b573ffab0799d27fcadc08d1e6a6aa31e40b4e6a61c5e9c98073a7012ba24","f54917b64b4e33520dd53be4677edb0d92a41e62d868d44eff6fdbc030530df8"],"profile":9,"step":1,"rows":4,"pos":2052,"prefix":2051,"fused":3,"vision":"no_axes","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":true},"consumers":["rows_gather"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,4,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":2,"compiled_calls":2,"hidden_calls":2,"qk_rows_calls":0,"traces":2,"selector_dispatches":2,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":1,"update_only":0},"buckets":{"raw4096:pool1024:out0":2},"entry_count":2,"compiled_keys":[{"source":"hidden","mode":"blocks","input_shape":[1,1,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"hidden:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"hidden","mode":"row_tokens","input_shape":[1,4,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"hidden:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p0_s0","sha256":["81cabd974019bcd738e4db77d9d7b089042d747910700154be3cc3da1ce986e7","316ed43e666fc6fb252dda7a8003ecb4be8da621d37db522985096db50465844","b4f1ff4375d0b1de7607bef982261bc0c7eba872a69ddc7b7dad6fbfdd3e1859","39b0f22d255932aefce0fff5850e71c764d11fd9e48c34b05fdbe22de04e7d44","b895415c8e478aff1ca87a13b4a553231b475db7a3969586a3b3d9b0d4291cd7"],"profile":0,"step":0,"rows":1,"pos":0,"prefix":0,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":1,"pooled_len":0,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,256,128],[1,256,128],[1,2,256,256],[1,2,256,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":0,"qk_rows_calls":1,"traces":1,"selector_dispatches":0,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":0,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":1},"buckets":{"raw256:pool256:out0":1},"entry_count":1,"compiled_keys":[{"source":"qk_rows","mode":"update_only","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,256,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,256,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":0,"label":"qk_rows:update_only:s1:raw256:pool256:out0:scorenone:chunk0:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p0_s1","sha256":["305e99a624ec1ef7144f43fad916352118cde06382377139cb606a09e5079bc9","8a6f35d3a8d1a938887f03baf9718fa9af582d121d1bed815dc68cb0482db7b1","46fe8ab534bd3d133ca7aa24d80a2f771907107ecac4f4f2496bc6501da6b855","847fb2445590d5b14c84e8a3afa037193a54e60aa9ab935f11815c0c237582c3","4152bdd42eacde7f856bace140740cc1b118fc0626c091daf2bbb7abaef2bc2d","cdb4f72a8a977ed2805ee125111ac4561a05997924610cb127d456aedb868553"],"profile":0,"step":1,"rows":4,"pos":1,"prefix":0,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":5,"pooled_len":1,"state_present":[true,true,true,true,true],"shapes":[[1,4,64],[1,256,128],[1,256,128],[1,1,128,256],[1,2,256,256],[1,2,256,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":0,"qk_rows_calls":1,"traces":1,"selector_dispatches":0,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":0,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":1},"buckets":{"raw256:pool256:out0":1},"entry_count":1,"compiled_keys":[{"source":"qk_rows","mode":"update_only","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,256,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,256,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":0,"label":"qk_rows:update_only:s1:raw256:pool256:out0:scorenone:chunk0:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p0_s2","sha256":["0db2e98676e803665a722901a32da71a553b2c589a9108e3b1f79a5be93c9992","75b526cbdc562be1855bd78936eb74132ac9e8c92186a06a517106e48544f312","d9b5b65fe3b0ac1e201e710f39f49c2d7a87f97dfb34cbdd71d1598d8b7d037d","fdfd8846b635f94a215516ccbca6b106a13d69c99a088893020fe5a1f27c110e","b456ba45bb69d6aec6338b8fc22e4bdb1e463a084966b8c33ed381c43d5579a3","ebc3501a599b08fc6c3f880049f7a5b4158343195f100e0afd4b769436043880"],"profile":0,"step":2,"rows":33,"pos":5,"prefix":0,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":38,"pooled_len":9,"state_present":[true,true,true,true,true],"shapes":[[1,33,64],[1,256,128],[1,256,128],[1,1,128,256],[1,2,256,256],[1,2,256,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":0,"qk_rows_calls":1,"traces":1,"selector_dispatches":0,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":0,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":1},"buckets":{"raw256:pool256:out0":1},"entry_count":1,"compiled_keys":[{"source":"qk_rows","mode":"update_only","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,256,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,256,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":0,"label":"qk_rows:update_only:s1:raw256:pool256:out0:scorenone:chunk0:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
@@ -45,12 +45,12 @@
|
|
|
|
|
{"kernel":"qsa_attention_p6_s0","sha256":["068422304f4191e74610562e007ffa7ec0d4b2ff5ed619997442ca1d254073b1","6f49058a22d7c62bbf4ffba4542b7aa32a130076b07537ac5605b2c85bb97162","158b3b53061c6d3d089049f9fdb47064a8c4f3d2f3fdb0ef5089afcaeba7cc5d","d4bc3f9184f206d1c86876a7ce98f086cf94f7912b460f7dc8c809b5b78d579b","011cc9bf7a9b2f86548e46728a9013e24273ca27ee50f9ee3f5509ecbc7c851f"],"profile":6,"step":0,"rows":2048,"pos":32768,"prefix":32768,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":32768,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["prefill_flash"],"counts":{"flash_kernel":1,"compiled_selector":1},"offset":34816,"pooled_len":8704,"state_present":[true,true,false,true,true],"shapes":[[1,2048,64],[1,65536,128],[1,16384,128],[1,2,34816,256],[1,2,34816,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":0,"qk_rows_calls":1,"traces":1,"selector_dispatches":1,"capacity_transitions":0,"prefill_score_producers":{"mpp":1,"mlx":0},"modes":{"blocks":0,"prefill_blocks":1,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw65536:pool16384:out0":1},"entry_count":1,"compiled_keys":[{"source":"qk_rows","mode":"prefill_blocks","input_shape":[1,2048,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"mpp","selector_chunk_rows":2048,"label":"qk_rows:prefill_blocks:s2048:raw65536:pool16384:out0:scorempp:chunk2048:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p6_s1","sha256":["bcbfcd4ff8d59b123b6ba9e84c2d621b9982c5e7430c84b82240f2b4416e48af","d50916defa6c9151ffeb706a7e0d6b8c04b861eaa94ed2b637f82fce79c8f2c9","158b3b53061c6d3d089049f9fdb47064a8c4f3d2f3fdb0ef5089afcaeba7cc5d","7c9ca3c2fd81b2f4749ce1f575a963b7054f004b91879a64212ecda5b0c34463","773fdcd29060aa9a94684db395aeccdbe1af803e191f3d0b938abc109d715fbb"],"profile":6,"step":1,"rows":1,"pos":34816,"prefix":32768,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":32768,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["flash"],"counts":{},"offset":34817,"pooled_len":8704,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,65536,128],[1,16384,128],[1,2,35072,256],[1,2,35072,256]],"core":{"calls":2,"compiled_calls":2,"hidden_calls":0,"qk_rows_calls":2,"traces":2,"selector_dispatches":2,"capacity_transitions":0,"prefill_score_producers":{"mpp":1,"mlx":0},"modes":{"blocks":1,"prefill_blocks":1,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw65536:pool16384:out0":2},"entry_count":2,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw65536:pool16384:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"prefill_blocks","input_shape":[1,2048,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"mpp","selector_chunk_rows":2048,"label":"qk_rows:prefill_blocks:s2048:raw65536:pool16384:out0:scorempp:chunk2048:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p6_s2","sha256":["e5f8b59d40e7e1c579d22d66af242e2882d1ab57727f17c7b9eeea894a5368e7","3b29542bd05bd5525466ad6c5951dd2d59759bed880564b8e07058d8feed1d42","9bc8c466e0a3164a6f4e22c0ffe23a895126f57945d4e7f0d5a07e79c63778ad","a41386c0de5defa3cc2eb26fe6005bab14865f573f9380f998d35e324abdbb43","270285c24ae2ea32e45240ada590a6cc1c9e7b14adfc23b270c7d7440c7372d4"],"profile":6,"step":2,"rows":4,"pos":34817,"prefix":32768,"fused":4,"vision":"none","no_indexer":false,"biased":false,"flash_min":32768,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["rows_gather"],"counts":{},"offset":34821,"pooled_len":8705,"state_present":[true,true,false,true,true],"shapes":[[1,4,64],[1,65536,128],[1,16384,128],[1,2,35072,256],[1,2,35072,256]],"core":{"calls":3,"compiled_calls":3,"hidden_calls":0,"qk_rows_calls":3,"traces":3,"selector_dispatches":3,"capacity_transitions":0,"prefill_score_producers":{"mpp":1,"mlx":0},"modes":{"blocks":1,"prefill_blocks":1,"dense_mask":0,"row_tokens":1,"update_only":0},"buckets":{"raw65536:pool16384:out0":3},"entry_count":3,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw65536:pool16384:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"prefill_blocks","input_shape":[1,2048,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"mpp","selector_chunk_rows":2048,"label":"qk_rows:prefill_blocks:s2048:raw65536:pool16384:out0:scorempp:chunk2048:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,4,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,65536,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,16384,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"qk_rows:row_tokens:s4:raw65536:pool16384:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s0","sha256":["f9c3c5622d5d7c346d6c2881634894e60f030654828e74d96041317d5d1b99e5","fa52f9dff396320af135c22c25b328674b3c44b6c768df9ed2d54d31bbecc71f","5385335d390b6950b06db2bd3a3e86efe8b4ca7d48d7458828f6891d0815be26","f583e84d01f4e3bfb152b8855f8520d25c5fd1639c20961343d44d3b494ee18f","ea3bda4b6f89e3552003b8afb307a5d04a67cb8d024ea18b7a7df90a172f7342"],"profile":7,"step":0,"rows":1,"pos":2051,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":2052,"pooled_len":513,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":0,"qk_rows_calls":1,"traces":1,"selector_dispatches":1,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw4096:pool1024:out0":1},"entry_count":1,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s1","sha256":["96067951d1aa0a433bcc85d4c51f0dbbf8f142f2f97b0e4aa1f0396ce6af89de","c7e437185225097e4ce2bb012fb9dea74124314e3858db5b2d3f89d175628f8e","99fb41883cf58e2cec92897b8486a4b20b6817dc32c1b98d320280e2984756b4","4af7a650ea8717de842e55d3691237c8e2af5c5c8ac78868e1323773e1a792f1","f54917b64b4e33520dd53be4677edb0d92a41e62d868d44eff6fdbc030530df8"],"profile":7,"step":1,"rows":4,"pos":2052,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,4,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":2,"compiled_calls":2,"hidden_calls":0,"qk_rows_calls":2,"traces":2,"selector_dispatches":2,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":1,"update_only":0},"buckets":{"raw4096:pool1024:out0":2},"entry_count":2,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,4,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"qk_rows:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s2","sha256":["5c48c9fcdb884b10fecce90489b58457a977528d3732d9ceb90d202f9a8bcabf","0962884629f0854f29ed23c253f65f1d44fcb8bfde142e948411c9e0ed65faf9","5b6ab0a7cc9a3a122481547531253bcba5f039fc507718fc85bef86dd8dc56ea","57eb1075e3ed3fe889a40c34bd1e59974119725a3a9ef009646a3427b3be9dd6","7fe9f3c5a939bb7dd64a147ea63354dc92c40828f2cc4d1cf8b3be331b86f4e5"],"profile":7,"step":2,"rows":2,"pos":2054,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,2,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":3,"compiled_calls":3,"hidden_calls":0,"qk_rows_calls":3,"traces":3,"selector_dispatches":3,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":2,"update_only":0},"buckets":{"raw4096:pool1024:out0":3},"entry_count":3,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,2,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":2,"label":"qk_rows:row_tokens:s2:raw4096:pool1024:out0:scorenone:chunk2:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,4,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"qk_rows:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s3","sha256":["554527077cad0087f4b548f1cf28dfbf186762ce6e3e12982922b97eec865e1d","9424991712e6561268208093dc9c9df018591b376554ac208eab7f242bdba4e4","5b6ab0a7cc9a3a122481547531253bcba5f039fc507718fc85bef86dd8dc56ea","28b0c2bfcc8486f5c6daafec65e013f6d2efbf99b24cdf75ba25372716d828d9","6d3d7b11ac7cff92650bdc798b862cbd21b383cd0154f53eee439af721d3c55f"],"profile":7,"step":3,"rows":1,"pos":2056,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":2057,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":4,"compiled_calls":4,"hidden_calls":0,"qk_rows_calls":4,"traces":3,"selector_dispatches":4,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":2,"prefill_blocks":0,"dense_mask":0,"row_tokens":2,"update_only":0},"buckets":{"raw4096:pool1024:out0":4},"entry_count":3,"compiled_keys":[{"source":"qk_rows","mode":"blocks","input_shape":[1,1,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"qk_rows:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,2,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":2,"label":"qk_rows:row_tokens:s2:raw4096:pool1024:out0:scorenone:chunk2:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"qk_rows","mode":"row_tokens","input_shape":[1,4,640],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"qk_rows:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s0","sha256":["9fc17259582838c33640717e4e21cdd556e3f5d03bd8dda03de124d43f7b55e8","48e48f96c54776ebd1a034cdfcf93ae5e027038346bff0b1cc4f6600a73bdbd4","56ff514413104108eea40eeb10e28ba89d4bcb5f273e7bfb327b650239b2608e","d1f4df60a631b8170515760665a9212d6375de4e6456f75d90eb31179c91596c","f583e84d01f4e3bfb152b8855f8520d25c5fd1639c20961343d44d3b494ee18f","ea3bda4b6f89e3552003b8afb307a5d04a67cb8d024ea18b7a7df90a172f7342"],"profile":7,"step":0,"rows":1,"pos":2051,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["flash"],"counts":{},"offset":2052,"pooled_len":513,"state_present":[true,true,true,true,true],"shapes":[[1,1,64],[1,2304,128],[1,1024,128],[1,1,128,1024],[1,2,2304,256],[1,2,2304,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s1","sha256":["24e8ced9cf6525d3fb9499e8cb45e70960458b26278552ff5e3bdb825596f33a","1ceca1a52f5057c1aaee04689fc8026d3df9e1b657d718d75b9a72aeb8a88b25","8ee668bf4a7f6eca91380e543147f88eb73d8bd3384c1a375b869d4680d95753","9221cbebf9df5094116ab9269b2c0e543d03fcd7f7ec5e9c1c745fc7b2babd0e","4af7a650ea8717de842e55d3691237c8e2af5c5c8ac78868e1323773e1a792f1","f54917b64b4e33520dd53be4677edb0d92a41e62d868d44eff6fdbc030530df8"],"profile":7,"step":1,"rows":4,"pos":2052,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["rows_gather"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,true,true,true],"shapes":[[1,4,64],[1,2304,128],[1,1024,128],[1,1,128,1024],[1,2,2304,256],[1,2,2304,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s2","sha256":["6cd50cb32aab449013ff1db077d2e4a2f480fb71f1ae11ed80499dfab78257d0","b49eaa7452b6018be70df323222b0624f230df95ea900e42a974eaba5e2ac975","a8f7c3b9add9760148121cf236dab8d6844d38d42ab5aa4a7de20bb14c342d97","d4abe49bd48fe5eeea175dd7102faf5cc4a0bcf4831d233e7521c51cf7a76932","57eb1075e3ed3fe889a40c34bd1e59974119725a3a9ef009646a3427b3be9dd6","7fe9f3c5a939bb7dd64a147ea63354dc92c40828f2cc4d1cf8b3be331b86f4e5"],"profile":7,"step":2,"rows":2,"pos":2054,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["rows_gather"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,true,true,true],"shapes":[[1,2,64],[1,2304,128],[1,1024,128],[1,1,128,1024],[1,2,2304,256],[1,2,2304,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p7_s3","sha256":["80cb2c6d1f34cc44dfbcd30192a17c8cb4a0c2a19017924058ece49e2f5c6486","59da7bcfd15c46c7c183a322559d2f455dba99f677bbef99b3626bd32160147f","a8f7c3b9add9760148121cf236dab8d6844d38d42ab5aa4a7de20bb14c342d97","d4abe49bd48fe5eeea175dd7102faf5cc4a0bcf4831d233e7521c51cf7a76932","28b0c2bfcc8486f5c6daafec65e013f6d2efbf99b24cdf75ba25372716d828d9","6d3d7b11ac7cff92650bdc798b862cbd21b383cd0154f53eee439af721d3c55f"],"profile":7,"step":3,"rows":1,"pos":2056,"prefix":2051,"fused":4,"vision":"table","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["flash"],"counts":{},"offset":2057,"pooled_len":514,"state_present":[true,true,true,true,true],"shapes":[[1,1,64],[1,2304,128],[1,1024,128],[1,1,128,1024],[1,2,2304,256],[1,2,2304,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p8_s0","sha256":["2f2d1534efe9754560b2d2289f6c9756da611928c8c9e2321786ab03dd447111","84e64cfc1690832a3a6ed516b577d1001b89b382254c08b2663f493691330f2e","7da3f843c044b333d43259efd0a68f1c0a8f5230707898663f9ce096dd1f294a"],"profile":8,"step":0,"rows":3,"pos":0,"prefix":0,"fused":0,"vision":"none","no_indexer":true,"biased":true,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":3,"pooled_len":0,"state_present":[false,false,false,true,true],"shapes":[[1,3,64],[1,2,256,256],[1,2,256,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p8_s1","sha256":["dbe815a17715596ea030903a6800bba50c5ed74b9e03a3d29826e42acd57377b","50d2e91a257e44f7554a3ba6b1e0b0c0bef2a3f9ec4f1c4b6904e150f9b7b48c","a79deff710fa918973e19986ced87ae13347fbd66972c79fd4241d299f4472dc"],"profile":8,"step":1,"rows":1,"pos":3,"prefix":0,"fused":0,"vision":"none","no_indexer":true,"biased":true,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":4,"pooled_len":0,"state_present":[false,false,false,true,true],"shapes":[[1,1,64],[1,2,256,256],[1,2,256,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p8_s2","sha256":["faeacdecf2a8829ff7c7c5751858003e0931435cae65c610e5c362a3997ba050","d4c84ff24f73effb7d50f227eb1248f997cc7ea466b6a15e6fa2876cf9481ea0","af7554d1b0575d27146ecc0b2d3755a89a645e42231c214b62cbecfaf6d81402"],"profile":8,"step":2,"rows":4,"pos":4,"prefix":0,"fused":0,"vision":"none","no_indexer":true,"biased":true,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":8,"pooled_len":0,"state_present":[false,false,false,true,true],"shapes":[[1,4,64],[1,2,256,256],[1,2,256,256]],"core":null,"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p9_s0","sha256":["91fd57e3f97be79622b57e34ed7e80b9934f65ce650592a6a619f0c88b322bd9","fa52f9dff396320af135c22c25b328674b3c44b6c768df9ed2d54d31bbecc71f","5385335d390b6950b06db2bd3a3e86efe8b4ca7d48d7458828f6891d0815be26","0823d3bc139c67f69911c8970fb2c1bcf5f757971e9191f8b98d56f6689840ae","ea3bda4b6f89e3552003b8afb307a5d04a67cb8d024ea18b7a7df90a172f7342"],"profile":9,"step":0,"rows":1,"pos":2051,"prefix":2051,"fused":3,"vision":"no_axes","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":2052,"pooled_len":513,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":1,"qk_rows_calls":0,"traces":1,"selector_dispatches":1,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw4096:pool1024:out0":1},"entry_count":1,"compiled_keys":[{"source":"hidden","mode":"blocks","input_shape":[1,1,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"hidden:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p9_s1","sha256":["893d7f0e21dc6a1c5099fa172b86630a0b694dc90ea7c6db4ae928c2151e9b00","c7e437185225097e4ce2bb012fb9dea74124314e3858db5b2d3f89d175628f8e","99fb41883cf58e2cec92897b8486a4b20b6817dc32c1b98d320280e2984756b4","e38b573ffab0799d27fcadc08d1e6a6aa31e40b4e6a61c5e9c98073a7012ba24","f54917b64b4e33520dd53be4677edb0d92a41e62d868d44eff6fdbc030530df8"],"profile":9,"step":1,"rows":4,"pos":2052,"prefix":2051,"fused":3,"vision":"no_axes","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["sdpa"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,4,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":2,"compiled_calls":2,"hidden_calls":2,"qk_rows_calls":0,"traces":2,"selector_dispatches":2,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":1,"update_only":0},"buckets":{"raw4096:pool1024:out0":2},"entry_count":2,"compiled_keys":[{"source":"hidden","mode":"blocks","input_shape":[1,1,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"hidden:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"hidden","mode":"row_tokens","input_shape":[1,4,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"hidden:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p9_s0","sha256":["6fd06b68b7059ae0220c03d7afada76bdadc86001123c4180f1a16380a89bece","fa52f9dff396320af135c22c25b328674b3c44b6c768df9ed2d54d31bbecc71f","5385335d390b6950b06db2bd3a3e86efe8b4ca7d48d7458828f6891d0815be26","0823d3bc139c67f69911c8970fb2c1bcf5f757971e9191f8b98d56f6689840ae","ea3bda4b6f89e3552003b8afb307a5d04a67cb8d024ea18b7a7df90a172f7342"],"profile":9,"step":0,"rows":1,"pos":2051,"prefix":2051,"fused":3,"vision":"no_axes","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["flash"],"counts":{},"offset":2052,"pooled_len":513,"state_present":[true,true,false,true,true],"shapes":[[1,1,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":1,"compiled_calls":1,"hidden_calls":1,"qk_rows_calls":0,"traces":1,"selector_dispatches":1,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":0,"update_only":0},"buckets":{"raw4096:pool1024:out0":1},"entry_count":1,"compiled_keys":[{"source":"hidden","mode":"blocks","input_shape":[1,1,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"hidden:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|
|
|
|
|
{"kernel":"qsa_attention_p9_s1","sha256":["d8762135183d3a85dfd8713c2fda95cf6397f495329ee3f626a0fc56f3cb8e2d","c7e437185225097e4ce2bb012fb9dea74124314e3858db5b2d3f89d175628f8e","99fb41883cf58e2cec92897b8486a4b20b6817dc32c1b98d320280e2984756b4","e38b573ffab0799d27fcadc08d1e6a6aa31e40b4e6a61c5e9c98073a7012ba24","f54917b64b4e33520dd53be4677edb0d92a41e62d868d44eff6fdbc030530df8"],"profile":9,"step":1,"rows":4,"pos":2052,"prefix":2051,"fused":3,"vision":"no_axes","no_indexer":false,"biased":false,"flash_min":2049,"prefill_gather":true,"options":{"fused":true,"compiled":true,"prefill":true,"phase_prefill":true,"prefill_min_rows":32,"prefill_compile_rows":2048,"prefill_min_context":2049,"prefill_workspace":134217728,"tile_rows":0,"flash":true,"gather_decode":false,"gather":true,"gather_max_rows":8,"gather_min_context":0,"tf32":false},"consumers":["rows_gather"],"counts":{},"offset":2056,"pooled_len":514,"state_present":[true,true,false,true,true],"shapes":[[1,4,64],[1,4096,128],[1,1024,128],[1,2,2304,256],[1,2,2304,256]],"core":{"calls":2,"compiled_calls":2,"hidden_calls":2,"qk_rows_calls":0,"traces":2,"selector_dispatches":2,"capacity_transitions":0,"prefill_score_producers":{"mpp":0,"mlx":0},"modes":{"blocks":1,"prefill_blocks":0,"dense_mask":0,"row_tokens":1,"update_only":0},"buckets":{"raw4096:pool1024:out0":2},"entry_count":2,"compiled_keys":[{"source":"hidden","mode":"blocks","input_shape":[1,1,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":1,"label":"hidden:blocks:s1:raw4096:pool1024:out0:scorenone:chunk1:mlx.core.bfloat16/mlx.core.bfloat16"},{"source":"hidden","mode":"row_tokens","input_shape":[1,4,64],"input_dtype":"mlx.core.bfloat16","raw_shape":[1,4096,128],"raw_dtype":"mlx.core.bfloat16","pooled_shape":[1,1024,128],"pooled_dtype":"mlx.core.bfloat16","n_heads":4,"kv_heads":1,"head_dim":128,"block_topk":512,"compress_ratio":4,"dense_output_tokens":0,"prefill_score_producer":"none","selector_chunk_rows":4,"label":"hidden:row_tokens:s4:raw4096:pool1024:out0:scorenone:chunk4:mlx.core.bfloat16/mlx.core.bfloat16"}],"require_bucketed_backing":true,"minimum_raw_capacity":256,"minimum_pooled_capacity":256,"selector_scratch_bytes":33554432,"prefill_score_workspace_bytes":134217728,"rope_attention_scaling":1.0},"frequencies":[1.0,0.6042963862419128,0.3651741147041321,0.22067339718341827,0.1333521455526352,0.08058421313762665,0.04869674891233444,0.02942727319896221,0.017782794311642647,0.010746077634394169,0.006493816152215004,0.003924189601093531,0.0023713738191872835,0.0014330126577988267,0.0008659642189741135,0.0005232990952208638,0.0003162277862429619,0.00019109529966954142,0.00011547819303814322,6.978305464144796e-05,4.2169649532297626e-05,2.5482966520939954e-05,1.5399264157167636e-05,9.305719686381053e-06,5.62341347176698e-06,3.398208264115965e-06,2.0535248950181995e-06,1.2409376495270408e-06,7.498942409256415e-07,4.5315832153391966e-07,2.738419482284371e-07,1.6548169412544667e-07]}
|