Xenova's picture
Xenova HF Staff
sync 91d990483a17
299d1df verified
Raw
History Blame
6.28 kB
{
"name": "ai.onnx.QLinearMatMul",
"id": "_ai_onnx_qlinearmatmul_webgpu_01c645e",
"version": 1,
"license": "Apache-2.0",
"backend": { "type": "webgpu" },
"digest": {
"algorithm": "sha256",
"files": {
"bench.json": "lNyYkeSCgx7aTrdH8bySfVgtFpShaVV7TmiwfkqMT78=",
"manifest.json": "by08nTulDE6RQaDnK/XTNy72r0iAKsQ23MhPmLfz0/4=",
"qlinear-matmul-accumulate.wgsl.jinja": "7rk2usO99dMU2qTBn4abzKoDaUCslKFba2Bp5sycDzw=",
"qlinear-matmul-requantize.wgsl.jinja": "yq9xtdlRJCjiaL0ntq0ZUYkaSgihrss42xbsSduidPY=",
"quant-dp4a-matmul.wgsl.jinja": "W5vXGi5jYNJzVsrM64Dv60NTBzBjnWwLwx83H6PYwnU=",
"quant-exact-matrix.wgsl.jinja": "Wb81U+WVi3IGnPrSboB7slVryPtLIndIfbmPL02TY48=",
"quant-exact-portable.wgsl.jinja": "src2Cb4EczrGNomwZCeTGYG2ZK+Vn/8sr7VY2GtMJWY=",
"quant-exact-prepare.wgsl.jinja": "Bumszk5Q6V6roGeAeQc7CD7Gl24gCFQmcjhm+qWAUS8=",
"quant-matmul-accumulate-rank4.wgsl.jinja": "n88NaFUVmu5MfVzjeA8am7YXKPwtKZbPo0VPH9Himho=",
"test.json": "0PZK6X10pbz0Zs2XeBiJOlQ2dMTzxxwAs6xxBVWWm8c="
}
},
"provenance": { "kernel": { "sha": "91d990483a174128daf7673f3f37a7c890493ae1", "dirty": false } },
"webgpu": {
"manifestSpec": "2.0",
"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"]
}
}
}