ttnn.matmul
This table is a trace for ttnn.matmul op. Traces are generated from nightly tt-torch runs. To see nightly runs: Nightly Runs
| Name | Input Shapes | Input Layouts | Attributes | Output Shapes | Output Layouts | PCC | ATOL |
|---|---|---|---|---|---|---|---|
| ttnn.matmul | tensor<[16,250,250,3,f32]> tensor<[3,1,f32]> | mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 64000 + d1 * 256 + d2, d3), memory_config: (32000, 1, 'tile<32x32, f32>', 'dram') mapping_from: (d0, d1), mapping_to: (d0, d1), memory_config: (1, 1, 'tile<32x32, f32>', 'dram') | transpose_a: False transpose_b: False | tensor<[16,250,250,1,f32]> | mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 64000 + d1 * 256 + d2, d3), memory_config: (32000, 1, 'tile<32x32, f32>', 'dram') | nan | nan |
| ttnn.matmul | tensor<[1,2640,768,3,f32]> tensor<[3,1,f32]> | mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 2027520 + d1 * 768 + d2, d3), memory_config: (63360, 1, 'tile<32x32, f32>', 'dram') mapping_from: (d0, d1), mapping_to: (d0, d1), memory_config: (1, 1, 'tile<32x32, f32>', 'dram') | transpose_a: False transpose_b: False | tensor<[1,2640,768,1,f32]> | mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 2027520 + d1 * 768 + d2, d3), memory_config: (63360, 1, 'tile<32x32, f32>', 'dram') | nan | nan |
| ttnn.matmul | tensor<[1,300,4,3,f32]> tensor<[3,1,f32]> | mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 9600 + d1 * 32 + d2, d3), memory_config: (300, 1, 'tile<32x32, f32>', 'dram') mapping_from: (d0, d1), mapping_to: (d0, d1), memory_config: (1, 1, 'tile<32x32, f32>', 'dram') | transpose_a: False transpose_b: False | tensor<[1,300,4,1,f32]> | mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 9600 + d1 * 32 + d2, d3), memory_config: (300, 1, 'tile<32x32, f32>', 'dram') | nan | nan |
| ttnn.matmul | tensor<[1,300,80,3,f32]> tensor<[3,1,f32]> | mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 28800 + d1 * 96 + d2, d3), memory_config: (900, 1, 'tile<32x32, f32>', 'dram') mapping_from: (d0, d1), mapping_to: (d0, d1), memory_config: (1, 1, 'tile<32x32, f32>', 'dram') | transpose_a: False transpose_b: False | tensor<[1,300,80,1,f32]> | mapping_from: (d0, d1, d2, d3), mapping_to: (d0 * 28800 + d1 * 96 + d2, d3), memory_config: (900, 1, 'tile<32x32, f32>', 'dram') | nan | nan |
| ttnn.matmul | tensor<[1,1,1,1,3,1,6,f32]> tensor<[6,1,f32]> | mapping_from: (d0, d1, d2, d3, d4, d5, d6), mapping_to: (d0 * 96 + d1 * 96 + d2 * 96 + d3 * 96 + d4 * 32 + d5, d6), memory_config: (3, 1, 'tile<32x32, f32>', 'dram') mapping_from: (d0, d1), mapping_to: (d0, d1), memory_config: (1, 1, 'tile<32x32, f32>', 'dram') | transpose_a: False transpose_b: False | tensor<[1,1,1,1,3,1,1,f32]> | mapping_from: (d0, d1, d2, d3, d4, d5, d6), mapping_to: (d0 * 96 + d1 * 96 + d2 * 96 + d3 * 96 + d4 * 32 + d5, d6), memory_config: (3, 1, 'tile<32x32, f32>', 'dram') | nan | nan |
| ttnn.matmul | tensor<[1,256,7,25281,4,f32]> tensor<[4,1,f32]> | mapping_from: (d0, d1, d2, d3, d4), mapping_to: (d0 * 45359104 + d1 * 177184 + d2 * 25312 + d3, d4), memory_config: (1417472, 1, 'tile<32x32, f32>', 'dram') mapping_from: (d0, d1), mapping_to: (d0, d1), memory_config: (1, 1, 'tile<32x32, f32>', 'dram') | transpose_a: False transpose_b: False | tensor<[1,256,7,25281,1,f32]> | mapping_from: (d0, d1, d2, d3, d4), mapping_to: (d0 * 45359104 + d1 * 177184 + d2 * 25312 + d3, d4), memory_config: (1417472, 1, 'tile<32x32, f32>', 'dram') | nan | nan |
| ttnn.matmul | tensor<[1,256,7,25281,4,f32]> tensor<[4,1,f32]> | mapping_from: (d0, d1, d2, d3, d4), mapping_to: (d0 * 45359104 + d1 * 177184 + d2 * 25312 + d3, d4), memory_config: (1417472, 1, 'tile<32x32, f32>', 'dram') mapping_from: (d0, d1), mapping_to: (d0, d1), memory_config: (1, 1, 'tile<32x32, f32>', 'dram') | transpose_a: False transpose_b: False | tensor<[1,256,7,25281,1,f32]> | mapping_from: (d0, d1, d2, d3, d4), mapping_to: (d0 * 45359104 + d1 * 177184 + d2 * 25312 + d3, d4), memory_config: (1417472, 1, 'tile<32x32, f32>', 'dram') | nan | nan |
| ttnn.matmul | tensor<[1,256,7,25281,4,f32]> tensor<[4,1,f32]> | mapping_from: (d0, d1, d2, d3, d4), mapping_to: (d0 * 45359104 + d1 * 177184 + d2 * 25312 + d3, d4), memory_config: (1417472, 1, 'tile<32x32, f32>', 'dram') mapping_from: (d0, d1), mapping_to: (d0, d1), memory_config: (1, 1, 'tile<32x32, f32>', 'dram') | transpose_a: False transpose_b: False | tensor<[1,256,7,25281,1,f32]> | mapping_from: (d0, d1, d2, d3, d4), mapping_to: (d0 * 45359104 + d1 * 177184 + d2 * 25312 + d3, d4), memory_config: (1417472, 1, 'tile<32x32, f32>', 'dram') | nan | nan |