| { | |
| "version": 1, | |
| "config": { | |
| "codebooks": 8, | |
| "codebook_size": 1024, | |
| "hidden": 256, | |
| "channels": 1024, | |
| "blocks": [ | |
| { | |
| "stride": 8, | |
| "pad": 4, | |
| "cin": 1024, | |
| "cout": 512, | |
| "dilations": [ | |
| 1, | |
| 3, | |
| 9 | |
| ] | |
| }, | |
| { | |
| "stride": 5, | |
| "pad": 3, | |
| "cin": 512, | |
| "cout": 256, | |
| "dilations": [ | |
| 1, | |
| 3, | |
| 9 | |
| ] | |
| }, | |
| { | |
| "stride": 4, | |
| "pad": 2, | |
| "cin": 256, | |
| "cout": 128, | |
| "dilations": [ | |
| 1, | |
| 3, | |
| 9 | |
| ] | |
| }, | |
| { | |
| "stride": 2, | |
| "pad": 1, | |
| "cin": 128, | |
| "cout": 64, | |
| "dilations": [ | |
| 1, | |
| 3, | |
| 9 | |
| ] | |
| }, | |
| { | |
| "stride": 3, | |
| "pad": 2, | |
| "cin": 64, | |
| "cout": 32, | |
| "dilations": [ | |
| 1, | |
| 3, | |
| 9 | |
| ] | |
| } | |
| ], | |
| "hop": 960, | |
| "sample_rate": 24000 | |
| }, | |
| "tensors": { | |
| "tables": { | |
| "offset": 0, | |
| "bytes": 8388608, | |
| "shape": [ | |
| 8, | |
| 1024, | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "in_bias": { | |
| "offset": 8388608, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "conv1.w": { | |
| "offset": 8389632, | |
| "bytes": 7340032, | |
| "shape": [ | |
| 1792, | |
| 1024 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "conv1.b": { | |
| "offset": 15729664, | |
| "bytes": 4096, | |
| "shape": [ | |
| 1024 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.a": { | |
| "offset": 15733760, | |
| "bytes": 4096, | |
| "shape": [ | |
| 1024 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.t.w": { | |
| "offset": 15737856, | |
| "bytes": 33554432, | |
| "shape": [ | |
| 8, | |
| 2048, | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.t.b": { | |
| "offset": 49292288, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u0.a1": { | |
| "offset": 49294336, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u0.w1": { | |
| "offset": 49296384, | |
| "bytes": 7340032, | |
| "shape": [ | |
| 3584, | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u0.b1": { | |
| "offset": 56636416, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u0.a2": { | |
| "offset": 56638464, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u0.w2": { | |
| "offset": 56640512, | |
| "bytes": 1048576, | |
| "shape": [ | |
| 512, | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u0.b2": { | |
| "offset": 57689088, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u1.a1": { | |
| "offset": 57691136, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u1.w1": { | |
| "offset": 57693184, | |
| "bytes": 7340032, | |
| "shape": [ | |
| 3584, | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u1.b1": { | |
| "offset": 65033216, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u1.a2": { | |
| "offset": 65035264, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u1.w2": { | |
| "offset": 65037312, | |
| "bytes": 1048576, | |
| "shape": [ | |
| 512, | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u1.b2": { | |
| "offset": 66085888, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u2.a1": { | |
| "offset": 66087936, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u2.w1": { | |
| "offset": 66089984, | |
| "bytes": 7340032, | |
| "shape": [ | |
| 3584, | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u2.b1": { | |
| "offset": 73430016, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u2.a2": { | |
| "offset": 73432064, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u2.w2": { | |
| "offset": 73434112, | |
| "bytes": 1048576, | |
| "shape": [ | |
| 512, | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b0.u2.b2": { | |
| "offset": 74482688, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.a": { | |
| "offset": 74484736, | |
| "bytes": 2048, | |
| "shape": [ | |
| 512 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.t.w": { | |
| "offset": 74486784, | |
| "bytes": 5242880, | |
| "shape": [ | |
| 5, | |
| 1024, | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.t.b": { | |
| "offset": 79729664, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u0.a1": { | |
| "offset": 79730688, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u0.w1": { | |
| "offset": 79731712, | |
| "bytes": 1835008, | |
| "shape": [ | |
| 1792, | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u0.b1": { | |
| "offset": 81566720, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u0.a2": { | |
| "offset": 81567744, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u0.w2": { | |
| "offset": 81568768, | |
| "bytes": 262144, | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u0.b2": { | |
| "offset": 81830912, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u1.a1": { | |
| "offset": 81831936, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u1.w1": { | |
| "offset": 81832960, | |
| "bytes": 1835008, | |
| "shape": [ | |
| 1792, | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u1.b1": { | |
| "offset": 83667968, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u1.a2": { | |
| "offset": 83668992, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u1.w2": { | |
| "offset": 83670016, | |
| "bytes": 262144, | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u1.b2": { | |
| "offset": 83932160, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u2.a1": { | |
| "offset": 83933184, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u2.w1": { | |
| "offset": 83934208, | |
| "bytes": 1835008, | |
| "shape": [ | |
| 1792, | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u2.b1": { | |
| "offset": 85769216, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u2.a2": { | |
| "offset": 85770240, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u2.w2": { | |
| "offset": 85771264, | |
| "bytes": 262144, | |
| "shape": [ | |
| 256, | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b1.u2.b2": { | |
| "offset": 86033408, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.a": { | |
| "offset": 86034432, | |
| "bytes": 1024, | |
| "shape": [ | |
| 256 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.t.w": { | |
| "offset": 86035456, | |
| "bytes": 1048576, | |
| "shape": [ | |
| 4, | |
| 512, | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.t.b": { | |
| "offset": 87084032, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u0.a1": { | |
| "offset": 87084544, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u0.w1": { | |
| "offset": 87085056, | |
| "bytes": 458752, | |
| "shape": [ | |
| 896, | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u0.b1": { | |
| "offset": 87543808, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u0.a2": { | |
| "offset": 87544320, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u0.w2": { | |
| "offset": 87544832, | |
| "bytes": 65536, | |
| "shape": [ | |
| 128, | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u0.b2": { | |
| "offset": 87610368, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u1.a1": { | |
| "offset": 87610880, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u1.w1": { | |
| "offset": 87611392, | |
| "bytes": 458752, | |
| "shape": [ | |
| 896, | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u1.b1": { | |
| "offset": 88070144, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u1.a2": { | |
| "offset": 88070656, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u1.w2": { | |
| "offset": 88071168, | |
| "bytes": 65536, | |
| "shape": [ | |
| 128, | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u1.b2": { | |
| "offset": 88136704, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u2.a1": { | |
| "offset": 88137216, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u2.w1": { | |
| "offset": 88137728, | |
| "bytes": 458752, | |
| "shape": [ | |
| 896, | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u2.b1": { | |
| "offset": 88596480, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u2.a2": { | |
| "offset": 88596992, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u2.w2": { | |
| "offset": 88597504, | |
| "bytes": 65536, | |
| "shape": [ | |
| 128, | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b2.u2.b2": { | |
| "offset": 88663040, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.a": { | |
| "offset": 88663552, | |
| "bytes": 512, | |
| "shape": [ | |
| 128 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.t.w": { | |
| "offset": 88664064, | |
| "bytes": 131072, | |
| "shape": [ | |
| 2, | |
| 256, | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.t.b": { | |
| "offset": 88795136, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u0.a1": { | |
| "offset": 88795392, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u0.w1": { | |
| "offset": 88795648, | |
| "bytes": 114688, | |
| "shape": [ | |
| 448, | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u0.b1": { | |
| "offset": 88910336, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u0.a2": { | |
| "offset": 88910592, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u0.w2": { | |
| "offset": 88910848, | |
| "bytes": 16384, | |
| "shape": [ | |
| 64, | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u0.b2": { | |
| "offset": 88927232, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u1.a1": { | |
| "offset": 88927488, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u1.w1": { | |
| "offset": 88927744, | |
| "bytes": 114688, | |
| "shape": [ | |
| 448, | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u1.b1": { | |
| "offset": 89042432, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u1.a2": { | |
| "offset": 89042688, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u1.w2": { | |
| "offset": 89042944, | |
| "bytes": 16384, | |
| "shape": [ | |
| 64, | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u1.b2": { | |
| "offset": 89059328, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u2.a1": { | |
| "offset": 89059584, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u2.w1": { | |
| "offset": 89059840, | |
| "bytes": 114688, | |
| "shape": [ | |
| 448, | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u2.b1": { | |
| "offset": 89174528, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u2.a2": { | |
| "offset": 89174784, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u2.w2": { | |
| "offset": 89175040, | |
| "bytes": 16384, | |
| "shape": [ | |
| 64, | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b3.u2.b2": { | |
| "offset": 89191424, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.a": { | |
| "offset": 89191680, | |
| "bytes": 256, | |
| "shape": [ | |
| 64 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.t.w": { | |
| "offset": 89191936, | |
| "bytes": 49152, | |
| "shape": [ | |
| 3, | |
| 128, | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.t.b": { | |
| "offset": 89241088, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u0.a1": { | |
| "offset": 89241344, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u0.w1": { | |
| "offset": 89241600, | |
| "bytes": 28672, | |
| "shape": [ | |
| 224, | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u0.b1": { | |
| "offset": 89270272, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u0.a2": { | |
| "offset": 89270528, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u0.w2": { | |
| "offset": 89270784, | |
| "bytes": 4096, | |
| "shape": [ | |
| 32, | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u0.b2": { | |
| "offset": 89274880, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u1.a1": { | |
| "offset": 89275136, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u1.w1": { | |
| "offset": 89275392, | |
| "bytes": 28672, | |
| "shape": [ | |
| 224, | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u1.b1": { | |
| "offset": 89304064, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u1.a2": { | |
| "offset": 89304320, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u1.w2": { | |
| "offset": 89304576, | |
| "bytes": 4096, | |
| "shape": [ | |
| 32, | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u1.b2": { | |
| "offset": 89308672, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u2.a1": { | |
| "offset": 89308928, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u2.w1": { | |
| "offset": 89309184, | |
| "bytes": 28672, | |
| "shape": [ | |
| 224, | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u2.b1": { | |
| "offset": 89337856, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u2.a2": { | |
| "offset": 89338112, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u2.w2": { | |
| "offset": 89338368, | |
| "bytes": 4096, | |
| "shape": [ | |
| 32, | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "b4.u2.b2": { | |
| "offset": 89342464, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "out.a": { | |
| "offset": 89342720, | |
| "bytes": 128, | |
| "shape": [ | |
| 32 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "out.w": { | |
| "offset": 89342976, | |
| "bytes": 896, | |
| "shape": [ | |
| 224, | |
| 1 | |
| ], | |
| "dtype": "f32" | |
| }, | |
| "out.b": { | |
| "offset": 89344000, | |
| "bytes": 4, | |
| "shape": [ | |
| 1 | |
| ], | |
| "dtype": "f32" | |
| } | |
| } | |
| } |