| { |
| "schema": "trellis-webgpu-pack", |
| "version": 1, |
| "stage": "idolVaeEncoder", |
| "upstreamRevision": "26afe3a78bb242c0a8bb181dcc8937bb16e5c66c", |
| "shards": [ |
| { |
| "file": "stage-000-afbba03a3458cba0d0f3675fa0d87df81b6b4fda666174702b60dd9ce71afeaf.bin", |
| "bytes": 64121856, |
| "sha256": "afbba03a3458cba0d0f3675fa0d87df81b6b4fda666174702b60dd9ce71afeaf" |
| }, |
| { |
| "file": "stage-001-6fc7193d287e5c4284ef5c42879cf80b2e667dbc02a72937f2ff22492f25474f.bin", |
| "bytes": 4731136, |
| "sha256": "6fc7193d287e5c4284ef5c42879cf80b2e667dbc02a72937f2ff22492f25474f" |
| } |
| ], |
| "tensors": { |
| "bn.running_mean": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f32", |
| "data": { |
| "shard": 0, |
| "offset": 0, |
| "bytes": 512 |
| } |
| }, |
| "bn.running_var": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f32", |
| "data": { |
| "shard": 0, |
| "offset": 512, |
| "bytes": 512 |
| } |
| }, |
| "encoder.conv_in.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1024, |
| "bytes": 256 |
| } |
| }, |
| "encoder.conv_in.weight": { |
| "shape": [ |
| 128, |
| 3, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1280, |
| "bytes": 6912 |
| } |
| }, |
| "encoder.conv_norm_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 8192, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.conv_norm_out.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 9216, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.conv_out.bias": { |
| "shape": [ |
| 64 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 10240, |
| "bytes": 128 |
| } |
| }, |
| "encoder.conv_out.weight": { |
| "shape": [ |
| 64, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 10496, |
| "bytes": 589824 |
| } |
| }, |
| "encoder.down_blocks.0.downsamplers.0.conv.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 600320, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.downsamplers.0.conv.weight": { |
| "shape": [ |
| 128, |
| 128, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 600576, |
| "bytes": 294912 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.0.conv1.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 895488, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.0.conv1.weight": { |
| "shape": [ |
| 128, |
| 128, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 895744, |
| "bytes": 294912 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.0.conv2.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1190656, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.0.conv2.weight": { |
| "shape": [ |
| 128, |
| 128, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1190912, |
| "bytes": 294912 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.0.norm1.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1485824, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.0.norm1.weight": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1486080, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.0.norm2.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1486336, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.0.norm2.weight": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1486592, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.1.conv1.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1486848, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.1.conv1.weight": { |
| "shape": [ |
| 128, |
| 128, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1487104, |
| "bytes": 294912 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.1.conv2.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1782016, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.1.conv2.weight": { |
| "shape": [ |
| 128, |
| 128, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1782272, |
| "bytes": 294912 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.1.norm1.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2077184, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.1.norm1.weight": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2077440, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.1.norm2.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2077696, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.0.resnets.1.norm2.weight": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2077952, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.1.downsamplers.0.conv.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2078208, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.downsamplers.0.conv.weight": { |
| "shape": [ |
| 256, |
| 256, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2078720, |
| "bytes": 1179648 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.conv1.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 3258368, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.conv1.weight": { |
| "shape": [ |
| 256, |
| 128, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 3258880, |
| "bytes": 589824 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.conv2.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 3848704, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.conv2.weight": { |
| "shape": [ |
| 256, |
| 256, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 3849216, |
| "bytes": 1179648 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.conv_shortcut.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5028864, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.conv_shortcut.weight": { |
| "shape": [ |
| 256, |
| 128, |
| 1, |
| 1 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5029376, |
| "bytes": 65536 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.norm1.bias": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5094912, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.norm1.weight": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5095168, |
| "bytes": 256 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.norm2.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5095424, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.0.norm2.weight": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5095936, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.1.conv1.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5096448, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.1.conv1.weight": { |
| "shape": [ |
| 256, |
| 256, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5096960, |
| "bytes": 1179648 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.1.conv2.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 6276608, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.1.conv2.weight": { |
| "shape": [ |
| 256, |
| 256, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 6277120, |
| "bytes": 1179648 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.1.norm1.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 7456768, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.1.norm1.weight": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 7457280, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.1.norm2.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 7457792, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.1.resnets.1.norm2.weight": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 7458304, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.2.downsamplers.0.conv.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 7458816, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.downsamplers.0.conv.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 7459840, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.conv1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 12178432, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.conv1.weight": { |
| "shape": [ |
| 512, |
| 256, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 12179456, |
| "bytes": 2359296 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.conv2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 14538752, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.conv2.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 14539776, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.conv_shortcut.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19258368, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.conv_shortcut.weight": { |
| "shape": [ |
| 512, |
| 256, |
| 1, |
| 1 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19259392, |
| "bytes": 262144 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.norm1.bias": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19521536, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.norm1.weight": { |
| "shape": [ |
| 256 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19522048, |
| "bytes": 512 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.norm2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19522560, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.0.norm2.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19523584, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.1.conv1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19524608, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.1.conv1.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19525632, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.1.conv2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 24244224, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.1.conv2.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 24245248, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.1.norm1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28963840, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.1.norm1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28964864, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.1.norm2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28965888, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.2.resnets.1.norm2.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28966912, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.0.conv1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28967936, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.0.conv1.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28968960, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.0.conv2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 33687552, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.0.conv2.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 33688576, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.0.norm1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 38407168, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.0.norm1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 38408192, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.0.norm2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 38409216, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.0.norm2.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 38410240, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.1.conv1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 38411264, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.1.conv1.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 38412288, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.1.conv2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43130880, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.1.conv2.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43131904, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.1.norm1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47850496, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.1.norm1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47851520, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.1.norm2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47852544, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.down_blocks.3.resnets.1.norm2.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47853568, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.attentions.0.group_norm.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47854592, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.attentions.0.group_norm.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47855616, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.attentions.0.to_k.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47856640, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.attentions.0.to_k.weight": { |
| "shape": [ |
| 512, |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47857664, |
| "bytes": 524288 |
| } |
| }, |
| "encoder.mid_block.attentions.0.to_out.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 48381952, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.attentions.0.to_out.0.weight": { |
| "shape": [ |
| 512, |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 48382976, |
| "bytes": 524288 |
| } |
| }, |
| "encoder.mid_block.attentions.0.to_q.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 48907264, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.attentions.0.to_q.weight": { |
| "shape": [ |
| 512, |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 48908288, |
| "bytes": 524288 |
| } |
| }, |
| "encoder.mid_block.attentions.0.to_v.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 49432576, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.attentions.0.to_v.weight": { |
| "shape": [ |
| 512, |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 49433600, |
| "bytes": 524288 |
| } |
| }, |
| "encoder.mid_block.resnets.0.conv1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 49957888, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.0.conv1.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 49958912, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.mid_block.resnets.0.conv2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 54677504, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.0.conv2.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 54678528, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.mid_block.resnets.0.norm1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 59397120, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.0.norm1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 59398144, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.0.norm2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 59399168, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.0.norm2.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 59400192, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.1.conv1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 59401216, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.1.conv1.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 59402240, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.mid_block.resnets.1.conv2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 64120832, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.1.conv2.weight": { |
| "shape": [ |
| 512, |
| 512, |
| 3, |
| 3 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 1, |
| "offset": 0, |
| "bytes": 4718592 |
| } |
| }, |
| "encoder.mid_block.resnets.1.norm1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 1, |
| "offset": 4718592, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.1.norm1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 1, |
| "offset": 4719616, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.1.norm2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 1, |
| "offset": 4720640, |
| "bytes": 1024 |
| } |
| }, |
| "encoder.mid_block.resnets.1.norm2.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 1, |
| "offset": 4721664, |
| "bytes": 1024 |
| } |
| }, |
| "quant_conv.bias": { |
| "shape": [ |
| 64 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 1, |
| "offset": 4722688, |
| "bytes": 128 |
| } |
| }, |
| "quant_conv.weight": { |
| "shape": [ |
| 64, |
| 64, |
| 1, |
| 1 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 1, |
| "offset": 4722944, |
| "bytes": 8192 |
| } |
| } |
| } |
| } |
|
|
