ai.onnx.Trilu / build /webgpu /bench.json
Xenova's picture
Xenova HF Staff
sync 6fdf6301e2bb
138cfda verified
Raw History Blame
13.9 kB
{
"cases": [
{
"name": "causal_mask_1x16x1024x1024",
"preset": "smoke",
"attrs": { "upper": 0 },
"inputs": {
"input": { "dtype": "float32", "shape": [1, 16, 1024, 1024], "dist": "normal", "seed": 711, "scale": 1 }
},
"outputs": { "output": { "dtype": "float32", "shape": [1, 16, 1024, 1024] } },
"bench": { "primary": true, "metrics": [{ "type": "bandwidth", "value": "1 * 16 * 1024 * 1024 * 4 * 2" }] }
},
{
"name": "mask_128",
"attrs": { "upper": 0 },
"inputs": { "input": { "dtype": "float32", "shape": [1, 1, 128, 128] } },
"outputs": { "output": { "dtype": "float32", "shape": [1, 1, 128, 128] } }
},
{
"name": "alignment_cliff_scalar_1x16x1023x1023",
"preset": "smoke",
"attrs": { "upper": 0 },
"inputs": {
"input": { "dtype": "float32", "shape": [1, 16, 1023, 1023], "dist": "normal", "seed": 712, "scale": 1 }
},
"outputs": { "output": { "dtype": "float32", "shape": [1, 16, 1023, 1023] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "f16_vec4_narrow_store_1x16x1024x1024",
"preset": "smoke",
"attrs": { "upper": 0 },
"inputs": {
"input": { "dtype": "float16", "shape": [1, 16, 1024, 1024], "dist": "normal", "seed": 713, "scale": 1 }
},
"outputs": { "output": { "dtype": "float16", "shape": [1, 16, 1024, 1024] } },
"bench": { "primary": true, "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
},
{
"name": "f16_alignment_cliff_scalar_1x16x1023x1023",
"preset": "smoke",
"attrs": { "upper": 0 },
"inputs": {
"input": { "dtype": "float16", "shape": [1, 16, 1023, 1023], "dist": "normal", "seed": 714, "scale": 1 }
},
"outputs": { "output": { "dtype": "float16", "shape": [1, 16, 1023, 1023] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
},
{
"name": "k_input_vec4_perlane_load_1x16x1024x1024",
"preset": "smoke",
"attrs": { "upper": 0 },
"inputs": {
"input": { "dtype": "float32", "shape": [1, 16, 1024, 1024], "dist": "normal", "seed": 715, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [0] } }
},
"outputs": { "output": { "dtype": "float32", "shape": [1, 16, 1024, 1024] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "k_input_scalar_cliff_fold_1x16x1025x1023",
"preset": "stress",
"attrs": { "upper": 0 },
"inputs": {
"input": { "dtype": "float32", "shape": [1, 16, 1025, 1023], "dist": "normal", "seed": 811, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [0] } }
},
"outputs": { "output": { "dtype": "float32", "shape": [1, 16, 1025, 1023] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "f16_scalar_cliff_fold_1x16x1025x1023",
"preset": "stress",
"attrs": { "upper": 0 },
"inputs": {
"input": { "dtype": "float16", "shape": [1, 16, 1025, 1023], "dist": "normal", "seed": 812, "scale": 1 }
},
"outputs": { "output": { "dtype": "float16", "shape": [1, 16, 1025, 1023] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
},
{
"name": "causal-mask-odd-context-f32-1x32x513x513",
"preset": "stress",
"attrs": { "upper": 0 },
"inputs": {
"input": { "dtype": "float32", "shape": [1, 32, 513, 513], "dist": "normal", "seed": 813, "scale": 1 }
},
"outputs": { "output": { "dtype": "float32", "shape": [1, 32, 513, 513] } },
"bench": { "primary": true, "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float32_65536x3x1_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float32", "shape": [65536, 3, 1], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "float32", "shape": [65536, 3, 1] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float32_32768x3x3_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float32", "shape": [32768, 3, 3], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "float32", "shape": [32768, 3, 3] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float32_4x513x511_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float32", "shape": [4, 513, 511], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "float32", "shape": [4, 513, 511] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float16_65536x3x1_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float16", "shape": [65536, 3, 1], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "float16", "shape": [65536, 3, 1] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float16_32768x3x3_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float16", "shape": [32768, 3, 3], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "float16", "shape": [32768, 3, 3] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float16_4x513x511_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float16", "shape": [4, 513, 511], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "float16", "shape": [4, 513, 511] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_int32_65536x3x1_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": {
"dtype": "int32",
"shape": [65536, 3, 1],
"dist": "uniform",
"seed": 901,
"min": -2147483648,
"max": 2147483647
},
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "int32", "shape": [65536, 3, 1] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_int32_32768x3x3_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": {
"dtype": "int32",
"shape": [32768, 3, 3],
"dist": "uniform",
"seed": 901,
"min": -2147483648,
"max": 2147483647
},
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "int32", "shape": [32768, 3, 3] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_int32_4x513x511_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": {
"dtype": "int32",
"shape": [4, 513, 511],
"dist": "uniform",
"seed": 901,
"min": -2147483648,
"max": 2147483647
},
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "int32", "shape": [4, 513, 511] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_uint32_65536x3x1_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": {
"dtype": "uint32",
"shape": [65536, 3, 1],
"dist": "uniform",
"seed": 901,
"min": 0,
"max": 4294967295
},
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "uint32", "shape": [65536, 3, 1] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_uint32_32768x3x3_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": {
"dtype": "uint32",
"shape": [32768, 3, 3],
"dist": "uniform",
"seed": 901,
"min": 0,
"max": 4294967295
},
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "uint32", "shape": [32768, 3, 3] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_uint32_4x513x511_upper1_k-1",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": {
"dtype": "uint32",
"shape": [4, 513, 511],
"dist": "uniform",
"seed": 901,
"min": 0,
"max": 4294967295
},
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-1] } }
},
"outputs": { "output": { "dtype": "uint32", "shape": [4, 513, 511] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float32_4x1024x1024_upper1_k2147483647",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float32", "shape": [4, 1024, 1024], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [2147483647] } }
},
"outputs": { "output": { "dtype": "float32", "shape": [4, 1024, 1024] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float32_4x1024x1024_upper1_k-2147483648",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float32", "shape": [4, 1024, 1024], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-2147483648] } }
},
"outputs": { "output": { "dtype": "float32", "shape": [4, 1024, 1024] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float16_4x1024x1024_upper1_k2147483647",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float16", "shape": [4, 1024, 1024], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [2147483647] } }
},
"outputs": { "output": { "dtype": "float16", "shape": [4, 1024, 1024] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
},
{
"name": "mask_float16_4x1024x1024_upper1_k-2147483648",
"preset": "stress",
"attrs": { "upper": 1 },
"inputs": {
"input": { "dtype": "float16", "shape": [4, 1024, 1024], "dist": "normal", "seed": 901, "scale": 1 },
"k": { "dtype": "int32", "shape": [], "data": { "kind": "values", "values": [-2147483648] } }
},
"outputs": { "output": { "dtype": "float16", "shape": [4, 1024, 1024] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
},
{
"name": "fold_float32_65537x8x8_wg8",
"preset": "stress",
"attrs": { "upper": 0 },
"tunables": { "WORKGROUP_SIZE": 8 },
"inputs": { "input": { "dtype": "float32", "shape": [65537, 8, 8], "dist": "normal", "seed": 902, "scale": 1 } },
"outputs": { "output": { "dtype": "float32", "shape": [65537, 8, 8] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "4 * numel(shapes.input) * 2" }] }
},
{
"name": "fold_float16_65537x8x8_wg8",
"preset": "stress",
"attrs": { "upper": 0 },
"tunables": { "WORKGROUP_SIZE": 8 },
"inputs": { "input": { "dtype": "float16", "shape": [65537, 8, 8], "dist": "normal", "seed": 902, "scale": 1 } },
"outputs": { "output": { "dtype": "float16", "shape": [65537, 8, 8] } },
"bench": { "metrics": [{ "type": "bandwidth", "value": "2 * numel(shapes.input) * 2" }] }
}
]
}