Xenova's picture
Xenova HF Staff
sync 6fdf6301e2bb
ca6e3a9 verified
Raw History Blame
6.28 kB
{
"name": "ai.onnx.QLinearMatMul",
"id": "_ai_onnx_qlinearmatmul_webgpu_3bb6b41",
"version": 1,
"license": "Apache-2.0",
"backend": { "type": "webgpu" },
"digest": {
"algorithm": "sha256",
"files": {
"bench.json": "fGmaDuxGofpqiA6cnDrWN9keK+mHui3tirYrV3mqJDY=",
"manifest.json": "84TvG2SeIe5iec6Df/TeGU4lM2kD6Qt25flPuwBnc+Y=",
"qlinear-matmul-accumulate.wgsl.jinja": "mQnZ5o/24le9tAjwm2YaR/VGNgyXPMOw1yYGJncP0lo=",
"qlinear-matmul-requantize.wgsl.jinja": "2Cy1+5R3n9G1TyXgx95i9mRQQ+7RbODyDyOQ+Eddxrw=",
"quant-dp4a-matmul.wgsl.jinja": "yZJNy7Y/CiYnS0HpVGYn0vXEENLU2T0x6V5OLaL0lUM=",
"quant-exact-matrix.wgsl.jinja": "1JvqPPP/RDHvb9K0R1zGpDMhhZgBHxJ1KGLO+zRbRlA=",
"quant-exact-portable.wgsl.jinja": "rjHHWnakekgA3pcIUha56rvJo3KV+nszg5Yo80Fd8zg=",
"quant-exact-prepare.wgsl.jinja": "eKTD4Qm5W9I6QnPYp8NDCLYCOwoq4qHwrAHQpmxEZmA=",
"quant-matmul-accumulate-rank4.wgsl.jinja": "OtFM589bZdZKx2rAZc95apyjZuKT9/ch+EYsP60Yndk=",
"test.json": "0ksQHz3OY+nD0YzYoMZ1mxSD9plQaHNLwerBCRni3K4="
}
},
"provenance": { "kernel": { "sha": "6fdf6301e2bbcc2f03bf1eaf493b7ad55ef33afc", "dirty": false } },
"webgpu": {
"manifestSpec": "2.1",
"variants": {
"sgmat_precast_scalar_f16": ["quant-exact-matrix.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"portable_precast_scalar_f16": ["quant-exact-portable.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"sgmat_precast_scalar": ["quant-exact-matrix.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"portable_precast_scalar": ["quant-exact-portable.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"sgmat_precast_column_f16": ["quant-exact-matrix.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"portable_precast_column_f16": ["quant-exact-portable.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"sgmat_precast_column": ["quant-exact-matrix.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"portable_precast_column": ["quant-exact-portable.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"sgmat_precast_column_scale_f16": ["quant-exact-matrix.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"portable_precast_column_scale_f16": ["quant-exact-portable.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"sgmat_precast_column_scale": ["quant-exact-matrix.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"portable_precast_column_scale": ["quant-exact-portable.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"sgmat_precast_column_zero_f16": ["quant-exact-matrix.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"portable_precast_column_zero_f16": ["quant-exact-portable.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"sgmat_precast_column_zero": ["quant-exact-matrix.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"portable_precast_column_zero": ["quant-exact-portable.wgsl.jinja", "quant-exact-prepare.wgsl.jinja"],
"sgmat_exact_scalar": ["quant-exact-matrix.wgsl.jinja"],
"sgmat_exact_column": ["quant-exact-matrix.wgsl.jinja"],
"sgmat_exact_column_scale": ["quant-exact-matrix.wgsl.jinja"],
"sgmat_exact_column_zero": ["quant-exact-matrix.wgsl.jinja"],
"rank3_per_column_quantization": ["qlinear-matmul-accumulate.wgsl.jinja", "qlinear-matmul-requantize.wgsl.jinja"],
"rank4_scalar_quantization": ["qlinear-matmul-requantize.wgsl.jinja", "quant-matmul-accumulate-rank4.wgsl.jinja"],
"rank2_per_column_scale_scalar_zp_quantization": ["qlinear-matmul-accumulate.wgsl.jinja", "qlinear-matmul-requantize.wgsl.jinja"],
"rank2_scalar_scale_per_column_zp_quantization": ["qlinear-matmul-accumulate.wgsl.jinja", "qlinear-matmul-requantize.wgsl.jinja"],
"rank2_scalar_quantization": ["qlinear-matmul-accumulate.wgsl.jinja", "qlinear-matmul-requantize.wgsl.jinja"],
"rank2_per_column_quantization": ["qlinear-matmul-accumulate.wgsl.jinja", "qlinear-matmul-requantize.wgsl.jinja"],
"dp4a_rank2_scalar_quantization": ["quant-dp4a-matmul.wgsl.jinja"],
"dp4a_rank2_per_column_quantization": ["quant-dp4a-matmul.wgsl.jinja"],
"dp4a_rank2_per_column_scale_scalar_zp": ["quant-dp4a-matmul.wgsl.jinja"],
"dp4a_rank2_scalar_scale_per_column_zp": ["quant-dp4a-matmul.wgsl.jinja"],
"dp4a_rank2_scalar_quantization_tail": ["quant-dp4a-matmul.wgsl.jinja"],
"dp4a_rank2_per_column_scale_scalar_zp_tail": ["quant-dp4a-matmul.wgsl.jinja"],
"dp4a_rank2_scalar_scale_per_column_zp_tail": ["quant-dp4a-matmul.wgsl.jinja"],
"dp4a_rank3_scalar_quantization": ["quant-dp4a-matmul.wgsl.jinja"],
"dp4a_rank4_scalar_quantization": ["quant-dp4a-matmul.wgsl.jinja"],
"rank3_scalar_quantization": ["qlinear-matmul-accumulate.wgsl.jinja", "qlinear-matmul-requantize.wgsl.jinja"],
"rank2_by_rank3_scalar_quantization": ["qlinear-matmul-accumulate.wgsl.jinja", "qlinear-matmul-requantize.wgsl.jinja"],
"rank3_by_rank2_scalar_quantization": ["qlinear-matmul-accumulate.wgsl.jinja", "qlinear-matmul-requantize.wgsl.jinja"],
"portable_accurate_dp4a_rank2_scalar_quantization": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"],
"portable_accurate_dp4a_rank2_per_column_quantization": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"],
"portable_accurate_dp4a_rank2_per_column_scale_scalar_zp": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"],
"portable_accurate_dp4a_rank2_scalar_quantization_tail": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"],
"portable_accurate_dp4a_rank2_per_column_scale_scalar_zp_tail": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"],
"portable_accurate_dp4a_rank3_scalar_quantization": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"],
"portable_accurate_dp4a_rank4_scalar_quantization": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"],
"portable_accurate_dp4a_rank2_scalar_scale_per_column_zp": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"],
"portable_accurate_dp4a_rank2_scalar_scale_per_column_zp_tail": ["qlinear-matmul-requantize.wgsl.jinja", "quant-dp4a-matmul.wgsl.jinja"]
}
}
}