Download build/webgpu/bench.json from webgpu-kernels/ai.onnx.Gemm: direct link, hf CLI and curl.
- Browser
- Download file 58 kB
-
https://huggingface.co/kernels/webgpu-kernels/ai.onnx.Gemm/resolve/v1/build/webgpu/bench.json
- Command line
-
hf download hf://webgpu-kernels/ai.onnx.Gemm@v1/build/webgpu/bench.json
-
curl -L -o bench.json https://huggingface.co/kernels/webgpu-kernels/ai.onnx.Gemm/resolve/v1/build/webgpu/bench.json
58 kB
| { | |
| "tunableSpace": { "TILE": [4, 8, 16], "GEMV_WORKGROUP": [32, 64, 128, 256], "sgmatConfig": [0, 2, 7] }, | |
| "cases": [ | |
| { | |
| "name": "qwen-image-mlp-gate-m4110-k4096-n12288-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 4110, "K": 4096, "N": 12288 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 4110, "inFeatures": 4096, "outFeatures": 12288 }, | |
| "tunableSpace": { "sgmatConfig": [0, 1, 2, 3, 4, 5, 6, 7, 10, 13] }, | |
| "inputs": { | |
| "a": { "shape": [4110, 4096], "dtype": "float32", "dist": "normal", "seed": 421, "scale": 0.2 }, | |
| "b": { "shape": [12288, 4096], "dtype": "float32", "dist": "normal", "seed": 422, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [4110, 12288], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen-image-mlp-down-m4110-k12288-n4096-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 4110, "K": 12288, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 4110, "inFeatures": 12288, "outFeatures": 4096 }, | |
| "tunableSpace": { "sgmatConfig": [0, 1, 2, 3, 4, 5, 6, 7, 10, 13] }, | |
| "inputs": { | |
| "a": { "shape": [4110, 12288], "dtype": "float32", "dist": "normal", "seed": 423, "scale": 0.2 }, | |
| "b": { "shape": [4096, 12288], "dtype": "float32", "dist": "normal", "seed": 424, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [4110, 4096], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen-image-512-mlp-gate-m1055-k4096-n12288-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 1055, "K": 4096, "N": 12288 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 1055, "inFeatures": 4096, "outFeatures": 12288 }, | |
| "inputs": { | |
| "a": { "shape": [1055, 4096], "dtype": "float32", "dist": "normal", "seed": 431, "scale": 0.2 }, | |
| "b": { "shape": [12288, 4096], "dtype": "float32", "dist": "normal", "seed": 432, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [1055, 12288], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen-image-512-mlp-down-m1055-k12288-n4096-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 1055, "K": 12288, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 1055, "inFeatures": 12288, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [1055, 12288], "dtype": "float32", "dist": "normal", "seed": 433, "scale": 0.2 }, | |
| "b": { "shape": [4096, 12288], "dtype": "float32", "dist": "normal", "seed": 434, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [1055, 4096], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen-image-512-attn-proj-m1055-k4096-n4096-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 1055, "K": 4096, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 1055, "inFeatures": 4096, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [1055, 4096], "dtype": "float32", "dist": "normal", "seed": 435, "scale": 0.2 }, | |
| "b": { "shape": [4096, 4096], "dtype": "float32", "dist": "normal", "seed": 436, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [1055, 4096], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen-image-512-mlp-gate-m1024-k4096-n12288-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 1024, "K": 4096, "N": 12288 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 1024, "inFeatures": 4096, "outFeatures": 12288 }, | |
| "inputs": { | |
| "a": { "shape": [1024, 4096], "dtype": "float32", "dist": "normal", "seed": 437, "scale": 0.2 }, | |
| "b": { "shape": [12288, 4096], "dtype": "float32", "dist": "normal", "seed": 438, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [1024, 12288], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen-image-512-mlp-down-m1024-k12288-n4096-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 1024, "K": 12288, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 1024, "inFeatures": 12288, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [1024, 12288], "dtype": "float32", "dist": "normal", "seed": 439, "scale": 0.2 }, | |
| "b": { "shape": [4096, 12288], "dtype": "float32", "dist": "normal", "seed": 440, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [1024, 4096], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen-image-512-attn-proj-m1024-k4096-n4096-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 1024, "K": 4096, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 1024, "inFeatures": 4096, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [1024, 4096], "dtype": "float32", "dist": "normal", "seed": 441, "scale": 0.2 }, | |
| "b": { "shape": [4096, 4096], "dtype": "float32", "dist": "normal", "seed": 442, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [1024, 4096], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-f32-64x128x128", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float32", "M": 64, "K": 128, "N": 128 }, | |
| "attrs": { "alpha": 1, "beta": 1 }, | |
| "inputs": { | |
| "a": { "shape": [64, 128], "dtype": "float32", "dist": "normal", "seed": 110, "scale": 0.2 }, | |
| "b": { "shape": [128, 128], "dtype": "float32", "dist": "normal", "seed": 111, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [64, 128], "dtype": "float32" } }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-f32-bias-64x128x128", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float32", "M": 64, "K": 128, "N": 128 }, | |
| "attrs": { "alpha": 0.75, "beta": 0.5 }, | |
| "inputs": { | |
| "a": { "shape": [64, 128], "dtype": "float32", "dist": "normal", "seed": 111, "scale": 0.2 }, | |
| "b": { "shape": [128, 128], "dtype": "float32", "dist": "normal", "seed": 112, "scale": 0.2 }, | |
| "bias": { "shape": [128], "dtype": "float32", "dist": "normal", "seed": 113, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "shape": [64, 128], "dtype": "float32" } }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-transb-f32-64x128x128", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float32", "M": 64, "K": 128, "N": 128 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 128, "outFeatures": 128 }, | |
| "inputs": { | |
| "a": { "shape": [64, 128], "dtype": "float32", "dist": "normal", "seed": 114, "scale": 0.2 }, | |
| "b": { "shape": [128, 128], "dtype": "float32", "dist": "normal", "seed": 115, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [64, 128], "dtype": "float32" } }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-transb-f16-256x256x256", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float16", "M": 256, "K": 256, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 256, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [256, 256], "dtype": "float16", "dist": "normal", "seed": 116, "scale": 0.2 }, | |
| "b": { "shape": [256, 256], "dtype": "float16", "dist": "normal", "seed": 117, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 256], "dtype": "float16" } }, | |
| "bench": { "primary": false, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-f16-tiled-256x1024x1024", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float16", "M": 256, "K": 1024, "N": 1024 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 1024, "outFeatures": 1024 }, | |
| "inputs": { | |
| "a": { "shape": [256, 1024], "dtype": "float16", "dist": "normal", "seed": 410, "scale": 0.5 }, | |
| "b": { "shape": [1024, 1024], "dtype": "float16", "dist": "normal", "seed": 411, "scale": 0.5 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 1024], "dtype": "float16" } }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-transb-f16-m256-k512-n512", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float16", "M": 256, "K": 512, "N": 512 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 512, "outFeatures": 512 }, | |
| "inputs": { | |
| "a": { "shape": [256, 512], "dtype": "float16", "dist": "normal", "seed": 412, "scale": 0.5 }, | |
| "b": { "shape": [512, 512], "dtype": "float16", "dist": "normal", "seed": 413, "scale": 0.5 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 512], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-transb-f16-m256-k1024-n512", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float16", "M": 256, "K": 1024, "N": 512 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 1024, "outFeatures": 512 }, | |
| "inputs": { | |
| "a": { "shape": [256, 1024], "dtype": "float16", "dist": "normal", "seed": 414, "scale": 0.5 }, | |
| "b": { "shape": [512, 1024], "dtype": "float16", "dist": "normal", "seed": 415, "scale": 0.5 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 512], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-transb-f16-m256-k512-n1024", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float16", "M": 256, "K": 512, "N": 1024 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 512, "outFeatures": 1024 }, | |
| "inputs": { | |
| "a": { "shape": [256, 512], "dtype": "float16", "dist": "normal", "seed": 416, "scale": 0.5 }, | |
| "b": { "shape": [1024, 512], "dtype": "float16", "dist": "normal", "seed": 417, "scale": 0.5 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 1024], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-transb-bias-f16-256x256x256", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float16", "M": 256, "K": 256, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 256, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [256, 256], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 256], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 256], "dtype": "float16" } }, | |
| "bench": { "primary": false, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "flux2-ff-linear-in-m4608-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 4608, "K": 3072, "N": 18432 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 4608, "inFeatures": 3072, "outFeatures": 18432 }, | |
| "tunableSpace": { "sgmatConfig": [0, 1, 2, 3, 4, 5, 6, 7, 10, 13] }, | |
| "inputs": { | |
| "a": { "shape": [4608, 3072], "dtype": "float32", "dist": "normal", "seed": 320, "scale": 0.2 }, | |
| "b": { "shape": [18432, 3072], "dtype": "float32", "dist": "normal", "seed": 321, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [4608, 18432], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen3-mlp-gate-prefill-m512-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 512, "K": 2560, "N": 9728 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 512, "inFeatures": 2560, "outFeatures": 9728 }, | |
| "tunableSpace": { "sgmatConfig": [0, 1, 2, 3, 4, 5] }, | |
| "inputs": { | |
| "a": { "shape": [512, 2560], "dtype": "float32", "dist": "normal", "seed": 322, "scale": 0.2 }, | |
| "b": { "shape": [9728, 2560], "dtype": "float32", "dist": "normal", "seed": 323, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [512, 9728], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen3-mlp-gate-prefill-m128-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 128, "K": 2560, "N": 9728 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 128, "inFeatures": 2560, "outFeatures": 9728 }, | |
| "tunableSpace": { "sgmatConfig": [0, 1, 2, 3, 4, 5] }, | |
| "inputs": { | |
| "a": { "shape": [128, 2560], "dtype": "float32", "dist": "normal", "seed": 326, "scale": 0.2 }, | |
| "b": { "shape": [9728, 2560], "dtype": "float32", "dist": "normal", "seed": 327, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [128, 9728], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen3-mlp-gate-prefill-m64-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 64, "K": 2560, "N": 9728 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 2560, "outFeatures": 9728 }, | |
| "tunableSpace": { "sgmatConfig": [0, 1, 2, 3, 4, 5] }, | |
| "inputs": { | |
| "a": { "shape": [64, 2560], "dtype": "float32", "dist": "normal", "seed": 328, "scale": 0.2 }, | |
| "b": { "shape": [9728, 2560], "dtype": "float32", "dist": "normal", "seed": 329, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [64, 9728], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "qwen3-qkv-decode-m1-f32", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 1, "K": 2560, "N": 6144 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 1, "inFeatures": 2560, "outFeatures": 6144 }, | |
| "inputs": { | |
| "a": { "shape": [1, 2560], "dtype": "float32", "dist": "normal", "seed": 324, "scale": 0.2 }, | |
| "b": { "shape": [6144, 2560], "dtype": "float32", "dist": "normal", "seed": 325, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [1, 6144], "dtype": "float32" } }, | |
| "bench": { | |
| "metrics": [ | |
| { | |
| "type": "bandwidth", | |
| "value": "(args.M * args.K + args.K * args.N + args.M * args.N) * dtypeBytes(args.dtype)" | |
| } | |
| ] | |
| } | |
| }, | |
| { | |
| "name": "gemm-model-scalar-unaligned-n2050-cliff-m256x2048x2050-f32", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float32", "M": 256, "K": 2048, "N": 2050 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 2048, "outFeatures": 2050 }, | |
| "inputs": { | |
| "a": { "shape": [256, 2048], "dtype": "float32", "dist": "normal", "seed": 330, "scale": 0.2 }, | |
| "b": { "shape": [2050, 2048], "dtype": "float32", "dist": "normal", "seed": 331, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 2050], "dtype": "float32" } }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-subgroup-aligned-n2048-control-m256x2048x2048-f32", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float32", "M": 256, "K": 2048, "N": 2048 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 2048, "outFeatures": 2048 }, | |
| "inputs": { | |
| "a": { "shape": [256, 2048], "dtype": "float32", "dist": "normal", "seed": 332, "scale": 0.2 }, | |
| "b": { "shape": [2048, 2048], "dtype": "float32", "dist": "normal", "seed": 333, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 2048], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-scalar-unaligned-n9700-cliff-m512x2560x9700-f32", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float32", "M": 512, "K": 2560, "N": 9700 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 512, "inFeatures": 2560, "outFeatures": 9700 }, | |
| "inputs": { | |
| "a": { "shape": [512, 2560], "dtype": "float32", "dist": "normal", "seed": 334, "scale": 0.2 }, | |
| "b": { "shape": [9700, 2560], "dtype": "float32", "dist": "normal", "seed": 335, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [512, 9700], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-m1-vec4-decode-f16-m1x2560x6144", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float16", "M": 1, "K": 2560, "N": 6144 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 1, "inFeatures": 2560, "outFeatures": 6144 }, | |
| "inputs": { | |
| "a": { "shape": [1, 2560], "dtype": "float16", "dist": "normal", "seed": 336, "scale": 0.2 }, | |
| "b": { "shape": [6144, 2560], "dtype": "float16", "dist": "normal", "seed": 337, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [1, 6144], "dtype": "float16" } }, | |
| "bench": { | |
| "metrics": [ | |
| { | |
| "type": "bandwidth", | |
| "value": "(args.M * args.K + args.K * args.N + args.M * args.N) * dtypeBytes(args.dtype)" | |
| } | |
| ] | |
| } | |
| }, | |
| { | |
| "name": "gemm-generic-no-trans-naive-tile-m512x2560x2048-f32", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float32", "M": 512, "K": 2560, "N": 2048 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 0 }, | |
| "inputs": { | |
| "a": { "shape": [512, 2560], "dtype": "float32", "dist": "normal", "seed": 338, "scale": 0.2 }, | |
| "b": { "shape": [2560, 2048], "dtype": "float32", "dist": "normal", "seed": 339, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [512, 2048], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-transA-naive-scalar-tile-m2048x2048x2048-f32", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 2048, "K": 2048, "N": 2048 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 1, "transB": 0 }, | |
| "inputs": { | |
| "a": { "shape": [2048, 2048], "dtype": "float32", "dist": "normal", "seed": 340, "scale": 0.2 }, | |
| "b": { "shape": [2048, 2048], "dtype": "float32", "dist": "normal", "seed": 341, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [2048, 2048], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-transA-tail-m2050x2050x2050-f32", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 2050, "K": 2050, "N": 2050 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 1, "transB": 0 }, | |
| "inputs": { | |
| "a": { "shape": [2050, 2050], "dtype": "float32", "dist": "normal", "seed": 344, "scale": 0.2 }, | |
| "b": { "shape": [2050, 2050], "dtype": "float32", "dist": "normal", "seed": 345, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [2050, 2050], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-scalar-kmisaligned-gemv-m64x2050x2048-f32", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 64, "K": 2050, "N": 2048 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 2050, "outFeatures": 2048 }, | |
| "inputs": { | |
| "a": { "shape": [64, 2050], "dtype": "float32", "dist": "normal", "seed": 342, "scale": 0.2 }, | |
| "b": { "shape": [2048, 2050], "dtype": "float32", "dist": "normal", "seed": 343, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [64, 2048], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-generic-transb-f32-512x1024x1024", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float32", "M": 512, "K": 1024, "N": 1024 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transB": 1 }, | |
| "inputs": { | |
| "a": { "shape": [512, 1024], "dtype": "float32", "dist": "normal", "seed": 110, "scale": 0.2 }, | |
| "b": { "shape": [1024, 1024], "dtype": "float32", "dist": "normal", "seed": 111, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [512, 1024], "dtype": "float32" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-generic-f16-transb-512x1024x1024", | |
| "preset": "smoke", | |
| "vars": { "dtype": "float16", "M": 512, "K": 1024, "N": 1024 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transB": 1 }, | |
| "inputs": { | |
| "a": { "shape": [512, 1024], "dtype": "float16", "dist": "normal", "seed": 112, "scale": 0.2 }, | |
| "b": { "shape": [1024, 1024], "dtype": "float16", "dist": "normal", "seed": 113, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [512, 1024], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-generic-f16-transb-m256-k1024-n512", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 256, "K": 1024, "N": 512 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transB": 1 }, | |
| "inputs": { | |
| "a": { "shape": [256, 1024], "dtype": "float16", "dist": "normal", "seed": 114, "scale": 0.2 }, | |
| "b": { "shape": [512, 1024], "dtype": "float16", "dist": "normal", "seed": 115, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 512], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-generic-f16-transb-m512-k512-n512", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 512, "K": 512, "N": 512 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transB": 1 }, | |
| "inputs": { | |
| "a": { "shape": [512, 512], "dtype": "float16", "dist": "normal", "seed": 116, "scale": 0.2 }, | |
| "b": { "shape": [512, 512], "dtype": "float16", "dist": "normal", "seed": 117, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [512, 512], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-generic-f16-transb-m1024-k1024-n512", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 1024, "K": 1024, "N": 512 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transB": 1 }, | |
| "inputs": { | |
| "a": { "shape": [1024, 1024], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [512, 1024], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [1024, 512], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-band-transb-f16-m8-k2560-n4096", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 8, "K": 2560, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 8, "inFeatures": 2560, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [8, 2560], "dtype": "float16", "dist": "normal", "seed": 520, "scale": 0.2 }, | |
| "b": { "shape": [4096, 2560], "dtype": "float16", "dist": "normal", "seed": 521, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [8, 4096], "dtype": "float16" } }, | |
| "bench": { | |
| "metrics": [ | |
| { | |
| "type": "bandwidth", | |
| "value": "(args.M * args.inFeatures + args.inFeatures * args.outFeatures + args.M * args.outFeatures) * dtypeBytes(args.dtype)" | |
| } | |
| ] | |
| } | |
| }, | |
| { | |
| "name": "gemm-band-transb-f16-m16-k2560-n4096", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 16, "K": 2560, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 16, "inFeatures": 2560, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [16, 2560], "dtype": "float16", "dist": "normal", "seed": 523, "scale": 0.2 }, | |
| "b": { "shape": [4096, 2560], "dtype": "float16", "dist": "normal", "seed": 524, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [16, 4096], "dtype": "float16" } }, | |
| "bench": { | |
| "metrics": [ | |
| { | |
| "type": "bandwidth", | |
| "value": "(args.M * args.inFeatures + args.inFeatures * args.outFeatures + args.M * args.outFeatures) * dtypeBytes(args.dtype)" | |
| } | |
| ] | |
| } | |
| }, | |
| { | |
| "name": "gemm-band-transb-bias-f16-m16-k2560-n4096", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 16, "K": 2560, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 16, "inFeatures": 2560, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [16, 2560], "dtype": "float16", "dist": "normal", "seed": 526, "scale": 0.2 }, | |
| "b": { "shape": [4096, 2560], "dtype": "float16", "dist": "normal", "seed": 527, "scale": 0.05 }, | |
| "bias": { "shape": [4096], "dtype": "float16", "dist": "normal", "seed": 528, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "shape": [16, 4096], "dtype": "float16" } }, | |
| "bench": { | |
| "metrics": [ | |
| { | |
| "type": "bandwidth", | |
| "value": "(args.M * args.inFeatures + args.inFeatures * args.outFeatures + args.M * args.outFeatures) * dtypeBytes(args.dtype)" | |
| } | |
| ] | |
| } | |
| }, | |
| { | |
| "name": "gemm-band-transb-f32-m16-k2560-n4096", | |
| "preset": "model", | |
| "vars": { "dtype": "float32", "M": 16, "K": 2560, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 16, "inFeatures": 2560, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [16, 2560], "dtype": "float32", "dist": "normal", "seed": 529, "scale": 0.2 }, | |
| "b": { "shape": [4096, 2560], "dtype": "float32", "dist": "normal", "seed": 530, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [16, 4096], "dtype": "float32" } }, | |
| "bench": { | |
| "metrics": [ | |
| { | |
| "type": "bandwidth", | |
| "value": "(args.M * args.inFeatures + args.inFeatures * args.outFeatures + args.M * args.outFeatures) * dtypeBytes(args.dtype)" | |
| } | |
| ] | |
| } | |
| }, | |
| { | |
| "name": "gemm-band-transb-f16-m32-k2560-n4096", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 32, "K": 2560, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 2560, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [32, 2560], "dtype": "float16", "dist": "normal", "seed": 532, "scale": 0.2 }, | |
| "b": { "shape": [4096, 2560], "dtype": "float16", "dist": "normal", "seed": 533, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [32, 4096], "dtype": "float16" } }, | |
| "bench": { | |
| "metrics": [ | |
| { | |
| "type": "bandwidth", | |
| "value": "(args.M * args.inFeatures + args.inFeatures * args.outFeatures + args.M * args.outFeatures) * dtypeBytes(args.dtype)" | |
| } | |
| ] | |
| } | |
| }, | |
| { | |
| "name": "gemm-band-transb-f16-m16-k8192-n256-narrow", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 16, "K": 8192, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 16, "inFeatures": 8192, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [16, 8192], "dtype": "float16", "dist": "normal", "seed": 535, "scale": 0.2 }, | |
| "b": { "shape": [256, 8192], "dtype": "float16", "dist": "normal", "seed": 536, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [16, 256], "dtype": "float16" } }, | |
| "bench": { | |
| "metrics": [ | |
| { | |
| "type": "bandwidth", | |
| "value": "(args.M * args.inFeatures + args.inFeatures * args.outFeatures + args.M * args.outFeatures) * dtypeBytes(args.dtype)" | |
| } | |
| ] | |
| } | |
| }, | |
| { | |
| "name": "gemm-model-transb-f16-m256-k256-n4096-tile-probe", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 256, "K": 256, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 256, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [256, 256], "dtype": "float16", "dist": "normal", "seed": 560, "scale": 0.2 }, | |
| "b": { "shape": [4096, 256], "dtype": "float16", "dist": "normal", "seed": 561, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 4096], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-model-transb-f16-m256-k512-n2048-tile-probe2", | |
| "preset": "model", | |
| "vars": { "dtype": "float16", "M": 256, "K": 512, "N": 2048 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 512, "outFeatures": 2048 }, | |
| "inputs": { | |
| "a": { "shape": [256, 512], "dtype": "float16", "dist": "normal", "seed": 562, "scale": 0.2 }, | |
| "b": { "shape": [2048, 512], "dtype": "float16", "dist": "normal", "seed": 563, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "shape": [256, 2048], "dtype": "float16" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "gemm-f32-generic-m1-k2048-n1000-transb", | |
| "preset": "stress", | |
| "provenance": { "notes": "A ResNet-class classifier head at batch one, spelled the way a graph node spells it." }, | |
| "vars": { "dtype": "float32", "M": 1, "K": 2048, "N": 1000 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "inputs": { | |
| "a": { "shape": [1, 2048], "dtype": "float32", "dist": "normal", "seed": 960, "scale": 0.2 }, | |
| "b": { "shape": [1000, 2048], "dtype": "float32", "dist": "normal", "seed": 961, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [1, 1000], "dtype": "float32" } } | |
| }, | |
| { | |
| "name": "gemm-f32-generic-m1-k4096-n4096-transb", | |
| "preset": "stress", | |
| "provenance": { "notes": "A decode-shaped projection with no packed-layout arguments." }, | |
| "vars": { "dtype": "float32", "M": 1, "K": 4096, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "inputs": { | |
| "a": { "shape": [1, 4096], "dtype": "float32", "dist": "normal", "seed": 960, "scale": 0.2 }, | |
| "b": { "shape": [4096, 4096], "dtype": "float32", "dist": "normal", "seed": 961, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [1, 4096], "dtype": "float32" } } | |
| }, | |
| { | |
| "name": "gemm-f32-generic-m1-k2048-n1000-transb-bias", | |
| "preset": "stress", | |
| "provenance": { | |
| "notes": "The same classifier head carrying its bias, which is how batch-norm folding leaves it." | |
| }, | |
| "vars": { "dtype": "float32", "M": 1, "K": 2048, "N": 1000 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "inputs": { | |
| "a": { "shape": [1, 2048], "dtype": "float32", "dist": "normal", "seed": 970, "scale": 0.2 }, | |
| "b": { "shape": [1000, 2048], "dtype": "float32", "dist": "normal", "seed": 971, "scale": 0.2 }, | |
| "bias": { "shape": [1000], "dtype": "float32", "dist": "normal", "seed": 972, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "shape": [1, 1000], "dtype": "float32" } } | |
| }, | |
| { | |
| "name": "gemm-transA_sgmat_preferred_floor_m512_k128_n1024", | |
| "preset": "model", | |
| "attrs": { "transA": 1, "transB": 0, "alpha": 1, "beta": 1 }, | |
| "inputs": { | |
| "a": { "dtype": "float32", "shape": [128, 512], "dist": "normal", "seed": 9403, "scale": 0.05 }, | |
| "b": { "dtype": "float32", "shape": [128, 1024], "dist": "normal", "seed": 9404, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [512, 1024], "dist": "empty" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": 134217728 }] } | |
| }, | |
| { | |
| "name": "gemm-transA_sgmat_preferred_high_aspect_m128_k512_n4096", | |
| "preset": "model", | |
| "attrs": { "transA": 1, "transB": 0, "alpha": 1, "beta": 1 }, | |
| "inputs": { | |
| "a": { "dtype": "float32", "shape": [512, 128], "dist": "normal", "seed": 9405, "scale": 0.05 }, | |
| "b": { "dtype": "float32", "shape": [512, 4096], "dist": "normal", "seed": 9406, "scale": 0.05 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [128, 4096], "dist": "empty" } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": 536870912 }] } | |
| }, | |
| { | |
| "name": "model_direct_a_boundary_m63_k96_n128", | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 63, "inFeatures": 96, "outFeatures": 128 }, | |
| "inputs": { | |
| "a": { "dtype": "float32", "shape": [63, 96], "dist": "normal", "seed": 11200, "scale": 0.2 }, | |
| "b": { "dtype": "float32", "shape": [128, 96], "dist": "normal", "seed": 11201, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [63, 128] } }, | |
| "provenance": { | |
| "notes": "Complete 64-row bands and adjacent row tails exercise direct activation loads versus guarded staging. Reduction depth spans one and three 32-element tiles; nonuniform signed operands exercise accumulation." | |
| }, | |
| "vars": { "M": 63, "K": 96, "N": 128 }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "model_direct_a_boundary_m64_k96_n128", | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 96, "outFeatures": 128 }, | |
| "inputs": { | |
| "a": { "dtype": "float32", "shape": [64, 96], "dist": "normal", "seed": 11202, "scale": 0.2 }, | |
| "b": { "dtype": "float32", "shape": [128, 96], "dist": "normal", "seed": 11203, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [64, 128] } }, | |
| "provenance": { | |
| "notes": "Complete 64-row bands and adjacent row tails exercise direct activation loads versus guarded staging. Reduction depth spans one and three 32-element tiles; nonuniform signed operands exercise accumulation." | |
| }, | |
| "vars": { "M": 64, "K": 96, "N": 128 }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "model_direct_a_boundary_m65_k96_n128", | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 65, "inFeatures": 96, "outFeatures": 128 }, | |
| "inputs": { | |
| "a": { "dtype": "float32", "shape": [65, 96], "dist": "normal", "seed": 11204, "scale": 0.2 }, | |
| "b": { "dtype": "float32", "shape": [128, 96], "dist": "normal", "seed": 11205, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [65, 128] } }, | |
| "provenance": { | |
| "notes": "Complete 64-row bands and adjacent row tails exercise direct activation loads versus guarded staging. Reduction depth spans one and three 32-element tiles; nonuniform signed operands exercise accumulation." | |
| }, | |
| "vars": { "M": 65, "K": 96, "N": 128 }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "model_direct_a_boundary_m128_k32_n64", | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 128, "inFeatures": 32, "outFeatures": 64 }, | |
| "inputs": { | |
| "a": { "dtype": "float32", "shape": [128, 32], "dist": "normal", "seed": 11206, "scale": 0.2 }, | |
| "b": { "dtype": "float32", "shape": [64, 32], "dist": "normal", "seed": 11207, "scale": 0.2 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [128, 64] } }, | |
| "provenance": { | |
| "notes": "Complete 64-row bands and adjacent row tails exercise direct activation loads versus guarded staging. Reduction depth spans one and three 32-element tiles; nonuniform signed operands exercise accumulation." | |
| }, | |
| "vars": { "M": 128, "K": 32, "N": 64 }, | |
| "bench": { "primary": true, "metrics": [{ "type": "gflops", "value": "2 * args.M * args.K * args.N" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m32_k512_n1024", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 32, "K": 512, "N": 1024 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 512, "outFeatures": 1024 }, | |
| "inputs": { | |
| "a": { "shape": [32, 512], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [1024, 512], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [1024], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [32, 1024] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m64_k256_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 64, "K": 256, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 256, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [64, 256], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 256], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [64, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m65_k256_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 65, "K": 256, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 65, "inFeatures": 256, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [65, 256], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 256], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [65, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m32_k512_n4032", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 32, "K": 512, "N": 4032 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 512, "outFeatures": 4032 }, | |
| "inputs": { | |
| "a": { "shape": [32, 512], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4032, 512], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4032], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [32, 4032] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m32_k512_n4096", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 32, "K": 512, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 512, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [32, 512], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4096, 512], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4096], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [32, 4096] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m32_k512_n4160", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 32, "K": 512, "N": 4160 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 512, "outFeatures": 4160 }, | |
| "inputs": { | |
| "a": { "shape": [32, 512], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4160, 512], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4160], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [32, 4160] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m63_k512_n4032", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 63, "K": 512, "N": 4032 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 63, "inFeatures": 512, "outFeatures": 4032 }, | |
| "inputs": { | |
| "a": { "shape": [63, 512], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4032, 512], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4032], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [63, 4032] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m64_k512_n4096", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 64, "K": 512, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 512, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [64, 512], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4096, 512], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4096], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [64, 4096] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m256_k2048_n1024", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 256, "K": 2048, "N": 1024 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 2048, "outFeatures": 1024 }, | |
| "inputs": { | |
| "a": { "shape": [256, 2048], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [1024, 2048], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [1024], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [256, 1024] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m64_k992_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 64, "K": 992, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 992, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [64, 992], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 992], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [64, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m64_k1024_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 64, "K": 1024, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 1024, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [64, 1024], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 1024], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [64, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float32_m64_k1056_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float32", "M": 64, "K": 1056, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 1056, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [64, 1056], "dtype": "float32", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 1056], "dtype": "float32", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float32", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float32", "shape": [64, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m32_k512_n1024", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 32, "K": 512, "N": 1024 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 512, "outFeatures": 1024 }, | |
| "inputs": { | |
| "a": { "shape": [32, 512], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [1024, 512], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [1024], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [32, 1024] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m64_k256_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 64, "K": 256, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 256, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [64, 256], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 256], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [64, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m65_k256_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 65, "K": 256, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 65, "inFeatures": 256, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [65, 256], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 256], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [65, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m32_k512_n4032", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 32, "K": 512, "N": 4032 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 512, "outFeatures": 4032 }, | |
| "inputs": { | |
| "a": { "shape": [32, 512], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4032, 512], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4032], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [32, 4032] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m32_k512_n4096", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 32, "K": 512, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 512, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [32, 512], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4096, 512], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4096], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [32, 4096] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m32_k512_n4160", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 32, "K": 512, "N": 4160 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 32, "inFeatures": 512, "outFeatures": 4160 }, | |
| "inputs": { | |
| "a": { "shape": [32, 512], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4160, 512], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4160], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [32, 4160] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m63_k512_n4032", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 63, "K": 512, "N": 4032 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 63, "inFeatures": 512, "outFeatures": 4032 }, | |
| "inputs": { | |
| "a": { "shape": [63, 512], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4032, 512], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4032], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [63, 4032] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m64_k512_n4096", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 64, "K": 512, "N": 4096 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 512, "outFeatures": 4096 }, | |
| "inputs": { | |
| "a": { "shape": [64, 512], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [4096, 512], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [4096], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [64, 4096] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m256_k2048_n1024", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 256, "K": 2048, "N": 1024 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 256, "inFeatures": 2048, "outFeatures": 1024 }, | |
| "inputs": { | |
| "a": { "shape": [256, 2048], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [1024, 2048], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [1024], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [256, 1024] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m64_k992_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 64, "K": 992, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 992, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [64, 992], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 992], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [64, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m64_k1024_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 64, "K": 1024, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 1024, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [64, 1024], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 1024], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [64, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| }, | |
| { | |
| "name": "bias_tile_grid_float16_m64_k1056_n256", | |
| "preset": "stress", | |
| "vars": { "dtype": "float16", "M": 64, "K": 1056, "N": 256 }, | |
| "attrs": { "alpha": 1, "beta": 1, "transA": 0, "transB": 1 }, | |
| "args": { "M": 64, "inFeatures": 1056, "outFeatures": 256 }, | |
| "inputs": { | |
| "a": { "shape": [64, 1056], "dtype": "float16", "dist": "normal", "seed": 118, "scale": 0.2 }, | |
| "b": { "shape": [256, 1056], "dtype": "float16", "dist": "normal", "seed": 119, "scale": 0.05 }, | |
| "bias": { "shape": [256], "dtype": "float16", "dist": "normal", "seed": 120, "scale": 0.1 } | |
| }, | |
| "outputs": { "y": { "dtype": "float16", "shape": [64, 256] } }, | |
| "bench": { "metrics": [{ "type": "gflops", "value": "2 * args.M * args.inFeatures * args.outFeatures" }] } | |
| } | |
| ] | |
| } | |