| { |
| "schema": "trellis-webgpu-pack", |
| "version": 1, |
| "stage": "litoVoxelDecoder", |
| "upstreamRevision": "apple/ml-lito", |
| "shards": [ |
| { |
| "file": "stage-000.bin", |
| "bytes": 61382144, |
| "sha256": "a9b696c647e3771a3e0ef3ce97e3b43ffcbe0a54010794f6601691cdd7aea8d6" |
| } |
| ], |
| "tensors": { |
| "final_layer.linear.bias": { |
| "shape": [ |
| 8 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 0, |
| "bytes": 16 |
| } |
| }, |
| "final_layer.linear.weight": { |
| "shape": [ |
| 8, |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 256, |
| "bytes": 8192 |
| } |
| }, |
| "final_layer.norm_final.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 8448, |
| "bytes": 1024 |
| } |
| }, |
| "final_layer.norm_final.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 9472, |
| "bytes": 1024 |
| } |
| }, |
| "input_linear.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 10496, |
| "bytes": 1024 |
| } |
| }, |
| "input_linear.weight": { |
| "shape": [ |
| 512, |
| 32 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 11520, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.layernorm_kv.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 44288, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.layernorm_kv.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 45312, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.layernorm_q.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 46336, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.layernorm_q.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47360, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.linear_kv.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 48384, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.linear_kv.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 52480, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 1101056, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1133824, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 1134848, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 1659136, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.linear_q.bias": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 1675520, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.linear_q.weight": { |
| "shape": [ |
| 1024, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 1024, |
| "columns": 512, |
| "padded_rows": 1024, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 1677568, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 2201856, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2218240, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.0.ca_layer.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2220288, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.0.ca_ln.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2222336, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ca_ln.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2223360, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ca_mlp.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 2224384, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.0.ca_mlp.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 2228480, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 3277056, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.0.ca_mlp.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 3309824, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ca_mlp.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 3310848, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 4359424, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.0.ln1_layers.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4392192, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ln1_layers.0.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4393216, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ln1_layers.1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4394240, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ln1_layers.1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4395264, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ln2_layers.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4396288, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ln2_layers.0.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4397312, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ln2_layers.1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4398336, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.ln2_layers.1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4399360, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.mlp_layers.0.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 4400384, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.0.mlp_layers.0.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 4404480, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 5453056, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.0.mlp_layers.0.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 5485824, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.mlp_layers.0.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 5486848, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 6535424, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.0.mlp_layers.1.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 6568192, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.0.mlp_layers.1.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 6572288, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 7620864, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.0.mlp_layers.1.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 7653632, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.mlp_layers.1.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 7654656, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 8703232, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.0.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 8736000, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.0.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 8737024, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 9261312, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.0.linear_qkv.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 9277696, |
| "bytes": 6144 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.0.linear_qkv.weight": { |
| "shape": [ |
| 3072, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 3072, |
| "columns": 512, |
| "padded_rows": 3072, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 9283840, |
| "bytes": 1572864 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 10856704, |
| "bytes": 49152 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.0.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 10905856, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.0.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 10907904, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.1.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 10909952, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.1.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 10910976, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 11435264, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.1.linear_qkv.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 11451648, |
| "bytes": 6144 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.1.linear_qkv.weight": { |
| "shape": [ |
| 3072, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 3072, |
| "columns": 512, |
| "padded_rows": 3072, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 11457792, |
| "bytes": 1572864 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 13030656, |
| "bytes": 49152 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.1.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 13079808, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.0.sa_layers.1.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 13081856, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.layernorm_kv.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 13083904, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.layernorm_kv.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 13084928, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.layernorm_q.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 13085952, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.layernorm_q.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 13086976, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.linear_kv.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 13088000, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.linear_kv.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 13092096, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 14140672, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 14173440, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 14174464, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 14698752, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.linear_q.bias": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 14715136, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.linear_q.weight": { |
| "shape": [ |
| 1024, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 1024, |
| "columns": 512, |
| "padded_rows": 1024, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 14717184, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 15241472, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 15257856, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.1.ca_layer.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 15259904, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.1.ca_ln.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 15261952, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ca_ln.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 15262976, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ca_mlp.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 15264000, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.1.ca_mlp.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 15268096, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 16316672, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.1.ca_mlp.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 16349440, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ca_mlp.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 16350464, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 17399040, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.1.ln1_layers.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17431808, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ln1_layers.0.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17432832, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ln1_layers.1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17433856, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ln1_layers.1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17434880, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ln2_layers.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17435904, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ln2_layers.0.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17436928, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ln2_layers.1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17437952, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.ln2_layers.1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17438976, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.mlp_layers.0.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 17440000, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.1.mlp_layers.0.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 17444096, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 18492672, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.1.mlp_layers.0.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 18525440, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.mlp_layers.0.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 18526464, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 19575040, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.1.mlp_layers.1.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 19607808, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.1.mlp_layers.1.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 19611904, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 20660480, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.1.mlp_layers.1.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 20693248, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.mlp_layers.1.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 20694272, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 21742848, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.0.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 21775616, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.0.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 21776640, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 22300928, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.0.linear_qkv.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 22317312, |
| "bytes": 6144 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.0.linear_qkv.weight": { |
| "shape": [ |
| 3072, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 3072, |
| "columns": 512, |
| "padded_rows": 3072, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 22323456, |
| "bytes": 1572864 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 23896320, |
| "bytes": 49152 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.0.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 23945472, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.0.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 23947520, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.1.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 23949568, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.1.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 23950592, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 24474880, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.1.linear_qkv.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 24491264, |
| "bytes": 6144 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.1.linear_qkv.weight": { |
| "shape": [ |
| 3072, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 3072, |
| "columns": 512, |
| "padded_rows": 3072, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 24497408, |
| "bytes": 1572864 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 26070272, |
| "bytes": 49152 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.1.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 26119424, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.1.sa_layers.1.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 26121472, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.layernorm_kv.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 26123520, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.layernorm_kv.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 26124544, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.layernorm_q.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 26125568, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.layernorm_q.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 26126592, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.linear_kv.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 26127616, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.linear_kv.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 26131712, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 27180288, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 27213056, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 27214080, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 27738368, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.linear_q.bias": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 27754752, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.linear_q.weight": { |
| "shape": [ |
| 1024, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 1024, |
| "columns": 512, |
| "padded_rows": 1024, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 27756800, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 28281088, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28297472, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.2.ca_layer.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28299520, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.2.ca_ln.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28301568, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ca_ln.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28302592, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ca_mlp.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 28303616, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.2.ca_mlp.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 28307712, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 29356288, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.2.ca_mlp.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 29389056, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ca_mlp.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 29390080, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 30438656, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.2.ln1_layers.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30471424, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ln1_layers.0.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30472448, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ln1_layers.1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30473472, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ln1_layers.1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30474496, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ln2_layers.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30475520, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ln2_layers.0.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30476544, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ln2_layers.1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30477568, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.ln2_layers.1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30478592, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.mlp_layers.0.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 30479616, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.2.mlp_layers.0.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 30483712, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 31532288, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.2.mlp_layers.0.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 31565056, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.mlp_layers.0.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 31566080, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 32614656, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.2.mlp_layers.1.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 32647424, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.2.mlp_layers.1.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 32651520, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 33700096, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.2.mlp_layers.1.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 33732864, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.mlp_layers.1.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 33733888, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 34782464, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.0.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 34815232, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.0.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 34816256, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 35340544, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.0.linear_qkv.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 35356928, |
| "bytes": 6144 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.0.linear_qkv.weight": { |
| "shape": [ |
| 3072, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 3072, |
| "columns": 512, |
| "padded_rows": 3072, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 35363072, |
| "bytes": 1572864 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 36935936, |
| "bytes": 49152 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.0.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 36985088, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.0.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 36987136, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.1.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 36989184, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.1.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 36990208, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 37514496, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.1.linear_qkv.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 37530880, |
| "bytes": 6144 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.1.linear_qkv.weight": { |
| "shape": [ |
| 3072, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 3072, |
| "columns": 512, |
| "padded_rows": 3072, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 37537024, |
| "bytes": 1572864 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 39109888, |
| "bytes": 49152 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.1.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 39159040, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.2.sa_layers.1.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 39161088, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.layernorm_kv.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 39163136, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.layernorm_kv.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 39164160, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.layernorm_q.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 39165184, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.layernorm_q.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 39166208, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.linear_kv.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 39167232, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.linear_kv.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 39171328, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 40219904, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 40252672, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 40253696, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 40777984, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.linear_q.bias": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 40794368, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.linear_q.weight": { |
| "shape": [ |
| 1024, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 1024, |
| "columns": 512, |
| "padded_rows": 1024, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 40796416, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 41320704, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 41337088, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.3.ca_layer.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 41339136, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.3.ca_ln.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 41341184, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ca_ln.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 41342208, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ca_mlp.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 41343232, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.3.ca_mlp.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 41347328, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 42395904, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.3.ca_mlp.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 42428672, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ca_mlp.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 42429696, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 43478272, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.3.ln1_layers.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43511040, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ln1_layers.0.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43512064, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ln1_layers.1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43513088, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ln1_layers.1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43514112, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ln2_layers.0.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43515136, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ln2_layers.0.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43516160, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ln2_layers.1.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43517184, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.ln2_layers.1.weight": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43518208, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.mlp_layers.0.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 43519232, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.3.mlp_layers.0.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 43523328, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 44571904, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.3.mlp_layers.0.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 44604672, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.mlp_layers.0.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 44605696, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 45654272, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.3.mlp_layers.1.fc1.bias": { |
| "shape": [ |
| 2048 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 45687040, |
| "bytes": 4096 |
| } |
| }, |
| "net.encoder.blocks.3.mlp_layers.1.fc1.weight": { |
| "shape": [ |
| 2048, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 2048, |
| "columns": 512, |
| "padded_rows": 2048, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 45691136, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 46739712, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.3.mlp_layers.1.fc2.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 46772480, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.mlp_layers.1.fc2.weight": { |
| "shape": [ |
| 512, |
| 2048 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 2048, |
| "padded_rows": 512, |
| "padded_columns": 2048, |
| "data": { |
| "shard": 0, |
| "offset": 46773504, |
| "bytes": 1048576 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 47822080, |
| "bytes": 32768 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.0.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 47854848, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.0.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 47855872, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 48380160, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.0.linear_qkv.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 48396544, |
| "bytes": 6144 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.0.linear_qkv.weight": { |
| "shape": [ |
| 3072, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 3072, |
| "columns": 512, |
| "padded_rows": 3072, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 48402688, |
| "bytes": 1572864 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 49975552, |
| "bytes": 49152 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.0.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 50024704, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.0.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 50026752, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.1.linear_out.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 50028800, |
| "bytes": 1024 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.1.linear_out.weight": { |
| "shape": [ |
| 512, |
| 1024 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 512, |
| "columns": 1024, |
| "padded_rows": 512, |
| "padded_columns": 1024, |
| "data": { |
| "shard": 0, |
| "offset": 50029824, |
| "bytes": 524288 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 50554112, |
| "bytes": 16384 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.1.linear_qkv.bias": { |
| "shape": [ |
| 3072 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 50570496, |
| "bytes": 6144 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.1.linear_qkv.weight": { |
| "shape": [ |
| 3072, |
| 512 |
| ], |
| "encoding": "symmetric-quantized", |
| "bits": 8, |
| "group_size": 64, |
| "output_tile": 16, |
| "rows": 3072, |
| "columns": 512, |
| "padded_rows": 3072, |
| "padded_columns": 512, |
| "data": { |
| "shard": 0, |
| "offset": 50576640, |
| "bytes": 1572864 |
| }, |
| "scales": { |
| "shard": 0, |
| "offset": 52149504, |
| "bytes": 49152 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.1.rmsnorm_k.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 52198656, |
| "bytes": 2048 |
| } |
| }, |
| "net.encoder.blocks.3.sa_layers.1.rmsnorm_q.scale": { |
| "shape": [ |
| 1024 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 52200704, |
| "bytes": 2048 |
| } |
| }, |
| "net.init_query": { |
| "shape": [ |
| 16, |
| 16, |
| 16, |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 52202752, |
| "bytes": 4194304 |
| } |
| }, |
| "net.init_query_linear.bias": { |
| "shape": [ |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 56397056, |
| "bytes": 1024 |
| } |
| }, |
| "net.init_query_linear.weight": { |
| "shape": [ |
| 512, |
| 771 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 56398080, |
| "bytes": 789504 |
| } |
| }, |
| "net.init_query_resolved": { |
| "shape": [ |
| 4096, |
| 512 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 57187584, |
| "bytes": 4194304 |
| } |
| }, |
| "net.zyx_pos_encoder.freq_bands": { |
| "shape": [ |
| 128 |
| ], |
| "encoding": "raw", |
| "dtype": "f16", |
| "data": { |
| "shard": 0, |
| "offset": 61381888, |
| "bytes": 256 |
| } |
| } |
| } |
| } |
|
|
