ttnn.permute

This table is a trace for ttnn.permute op. Traces are generated from nightly tt-torch runs. To see nightly runs: Nightly Runs

NameInput ShapesInput LayoutsAttributesOutput ShapesOutput LayoutsPCCATOL
ttnn.permutetensor<[1,128,512,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 512 + d2, d3), memory_config: (2048, 16, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,512,512,128,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 262144 + d1 * 512 + d2, d3), memory_config: (8192, 4, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,512,512,128,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 262144 + d1 * 512 + d2, d3), memory_config: (8192, 4, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,128,512,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 512 + d2, d3), memory_config: (2048, 16, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,128,512,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 512 + d2, d3), memory_config: (2048, 16, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,512,512,128,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 262144 + d1 * 512 + d2, d3), memory_config: (8192, 4, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,512,512,3,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 262144 + d1 * 512 + d2, d3), memory_config: (8192, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,3,512,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 1536 + d1 * 512 + d2, d3), memory_config: (48, 16, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,16,14,14,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 512 + d1 * 32 + d2, d3), memory_config: (16, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,14,14,16,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 448 + d1 * 32 + d2, d3), memory_config: (14, 1, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,14,14,4,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 448 + d1 * 32 + d2, d3), memory_config: (14, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,4,14,14,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 128 + d1 * 32 + d2, d3), memory_config: (4, 1, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,256,512,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 131072 + d1 * 512 + d2, d3), memory_config: (4096, 16, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,512,512,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 262144 + d1 * 512 + d2, d3), memory_config: (8192, 8, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,512,512,128,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 262144 + d1 * 512 + d2, d3), memory_config: (8192, 4, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,128,512,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 512 + d2, d3), memory_config: (2048, 16, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,256,512,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 131072 + d1 * 512 + d2, d3), memory_config: (4096, 16, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,512,512,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 262144 + d1 * 512 + d2, d3), memory_config: (8192, 8, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,512,512,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 262144 + d1 * 512 + d2, d3), memory_config: (8192, 8, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,256,512,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 131072 + d1 * 512 + d2, d3), memory_config: (4096, 16, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,3,1024,1024,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 3072 + d1 * 1024 + d2, d3), memory_config: (96, 32, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,1024,1024,3,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 1048576 + d1 * 1024 + d2, d3), memory_config: (32768, 1, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,1024,1024,64,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 1048576 + d1 * 1024 + d2, d3), memory_config: (32768, 2, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,64,1024,1024,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 1024 + d2, d3), memory_config: (2048, 32, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,480,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 122880 + d1 * 256 + d2, d3), memory_config: (3840, 8, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,256,256,480,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 15, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,256,256,64,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 2, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,64,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 16384 + d1 * 256 + d2, d3), memory_config: (512, 8, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,512,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 131072 + d1 * 256 + d2, d3), memory_config: (4096, 8, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,256,256,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 16, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,256,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 8, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,256,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 8, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,512,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 131072 + d1 * 256 + d2, d3), memory_config: (4096, 8, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,256,256,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 16, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,256,256,512,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 16, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,512,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 131072 + d1 * 256 + d2, d3), memory_config: (4096, 8, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,64,1024,1024,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 1024 + d2, d3), memory_config: (2048, 32, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,1024,1024,64,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 1048576 + d1 * 1024 + d2, d3), memory_config: (32768, 2, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,1024,1024,48,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 1048576 + d1 * 1024 + d2, d3), memory_config: (32768, 2, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,48,1024,1024,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 49152 + d1 * 1024 + d2, d3), memory_config: (1536, 32, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,64,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 16384 + d1 * 256 + d2, d3), memory_config: (512, 8, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,256,256,64,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 2, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,256,256,98,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 65536 + d1 * 256 + d2, d3), memory_config: (2048, 4, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,98,256,256,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 25088 + d1 * 256 + d2, d3), memory_config: (784, 8, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,64,480,640,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 30720 + d1 * 480 + d2, d3), memory_config: (960, 20, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,480,640,64,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 307200 + d1 * 640 + d2, d3), memory_config: (9600, 2, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,480,640,1,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 307200 + d1 * 640 + d2, d3), memory_config: (9600, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,1,480,640,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 480 + d1 * 480 + d2, d3), memory_config: (15, 20, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,64,480,640,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 30720 + d1 * 480 + d2, d3), memory_config: (960, 20, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,480,640,64,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 307200 + d1 * 640 + d2, d3), memory_config: (9600, 2, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,480,640,64,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 307200 + d1 * 640 + d2, d3), memory_config: (9600, 2, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,64,480,640,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 30720 + d1 * 480 + d2, d3), memory_config: (960, 20, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,960,128,128,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 122880 + d1 * 128 + d2, d3), memory_config: (3840, 4, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,128,128,960,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 16384 + d1 * 128 + d2, d3), memory_config: (512, 30, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,128,128,64,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 16384 + d1 * 128 + d2, d3), memory_config: (512, 2, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,64,128,128,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 8192 + d1 * 128 + d2, d3), memory_config: (256, 4, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,16,28,28,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 512 + d1 * 32 + d2, d3), memory_config: (16, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,28,28,16,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 896 + d1 * 32 + d2, d3), memory_config: (28, 1, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,14,14,16,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 448 + d1 * 32 + d2, d3), memory_config: (14, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,16,14,14,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 512 + d1 * 32 + d2, d3), memory_config: (16, 1, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,4,14,14,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 128 + d1 * 32 + d2, d3), memory_config: (4, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,14,14,4,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 448 + d1 * 32 + d2, d3), memory_config: (14, 1, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,7,7,4,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 224 + d1 * 32 + d2, d3), memory_config: (7, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,4,7,7,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 128 + d1 * 32 + d2, d3), memory_config: (4, 1, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,528,14,14,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 16896 + d1 * 32 + d2, d3), memory_config: (528, 1, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 2, 3, 1>tensor<[1,14,14,528,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 448 + d1 * 32 + d2, d3), memory_config: (14, 17, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,14,14,528,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 448 + d1 * 32 + d2, d3), memory_config: (14, 17, 'tile<32x32, bf16>', 'dram')permutation: array<i64: 0, 3, 1, 2>tensor<[1,528,14,14,bf16]>mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 16896 + d1 * 32 + d2, d3), memory_config: (528, 1, 'tile<32x32, bf16>', 'dram')nannan
ttnn.permutetensor<[1,220,12,1,768,f32]>mapping_from: (d0, d1, d2, d3, d4), mapping_to: (d0 * 84480 + d1 * 384 + d2 * 32 + d3, d4), memory_config: (2640, 24, 'tile<32x32, f32>', 'dram')permutation: array<i64: 1, 2, 4, 0, 3>tensor<[220,12,768,1,1,f32]>mapping_from: (d0, d1, d2, d3, d4), mapping_to: (d0 * 294912 + d1 * 24576 + d2 * 32 + d3, d4), memory_config: (2027520, 1, 'tile<32x32, f32>', 'dram')nannan