| { |
| "format_version": "1.0", |
| "quantization_method": "uniform", |
| "uniform_bits": 8, |
| "quantized_layers": { |
| "model.layers.0.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.0.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.0.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.0.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.0.mlp.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 6848, |
| "has_bias": false, |
| "original_shape": [ |
| 6848, |
| 1280 |
| ] |
| }, |
| "model.layers.0.mlp.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 6848, |
| "has_bias": false, |
| "original_shape": [ |
| 6848, |
| 1280 |
| ] |
| }, |
| "model.layers.0.mlp.down_proj": { |
| "bits": 8, |
| "in_features": 6848, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 6848 |
| ] |
| }, |
| "model.layers.1.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.1.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.1.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.1.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.1.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.1.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.2.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.2.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.2.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.2.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.2.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.2.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.3.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.3.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.3.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.3.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.3.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.3.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.4.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.4.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.4.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.4.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.4.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.4.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.5.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.5.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.5.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.5.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.5.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.5.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.6.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.6.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.6.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.6.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.6.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.6.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.7.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.7.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.7.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.7.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.7.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.7.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.8.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.8.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.8.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.8.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.8.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.8.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.9.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.9.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.9.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.9.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.9.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.9.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.10.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.10.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.10.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.10.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.10.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.10.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.layers.11.self_attn.q_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.11.self_attn.k_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.11.self_attn.v_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.11.self_attn.o_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.0.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.0.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.0.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.1.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.1.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.1.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.2.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.2.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.2.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.3.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.3.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.3.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.4.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.4.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.4.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.5.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.5.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.5.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.6.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.6.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.6.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.7.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.7.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.7.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.8.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.8.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.8.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.9.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.9.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.9.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.10.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.10.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.10.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.11.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.11.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.11.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.12.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.12.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.12.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.13.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.13.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.13.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.14.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.14.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.14.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.15.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.15.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.15.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.16.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.16.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.16.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.17.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.17.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.17.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.18.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.18.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.18.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.19.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.19.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.19.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.20.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.20.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.20.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.21.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.21.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.21.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.22.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.22.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.22.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.23.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.23.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.23.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.24.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.24.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.24.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.25.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.25.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.25.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.26.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.26.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.26.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.27.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.27.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.27.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.28.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.28.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.28.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.29.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.29.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.29.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.30.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.30.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.30.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.31.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.31.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.31.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.32.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.32.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.32.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.33.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.33.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.33.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.34.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.34.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.34.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.35.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.35.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.35.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.36.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.36.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.36.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.37.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.37.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.37.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.38.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.38.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.38.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.39.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.39.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.39.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.40.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.40.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.40.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.41.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.41.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.41.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.42.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.42.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.42.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.43.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.43.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.43.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.44.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.44.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.44.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.45.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.45.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.45.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.46.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.46.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.46.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.47.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.47.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.47.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.48.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.48.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.48.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.49.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.49.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.49.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.50.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.50.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.50.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.51.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.51.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.51.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.52.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.52.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.52.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.53.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.53.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.53.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.54.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.54.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.54.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.55.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.55.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.55.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.56.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.56.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.56.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.57.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.57.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.57.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.58.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.58.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.58.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.59.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.59.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.59.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.60.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.60.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.60.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.61.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.61.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.61.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.62.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.62.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.62.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.experts.63.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.63.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 896, |
| "has_bias": false, |
| "original_shape": [ |
| 896, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.experts.63.down_proj": { |
| "bits": 8, |
| "in_features": 896, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 896 |
| ] |
| }, |
| "model.layers.11.mlp.shared_experts.gate_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.shared_experts.up_proj": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 1792, |
| "has_bias": false, |
| "original_shape": [ |
| 1792, |
| 1280 |
| ] |
| }, |
| "model.layers.11.mlp.shared_experts.down_proj": { |
| "bits": 8, |
| "in_features": 1792, |
| "out_features": 1280, |
| "has_bias": false, |
| "original_shape": [ |
| 1280, |
| 1792 |
| ] |
| }, |
| "model.sam_model.blocks.0.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.0.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.0.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.0.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.1.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.1.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.1.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.1.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.2.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.2.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.2.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.2.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.3.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.3.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.3.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.3.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.4.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.4.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.4.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.4.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.5.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.5.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.5.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.5.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.6.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.6.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.6.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.6.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.7.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.7.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.7.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.7.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.8.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.8.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.8.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.8.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.9.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.9.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.9.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.9.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.10.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.10.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.10.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.10.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.sam_model.blocks.11.attn.qkv": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 2304, |
| "has_bias": true, |
| "original_shape": [ |
| 2304, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.11.attn.proj": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.11.mlp.lin1": { |
| "bits": 8, |
| "in_features": 768, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 768 |
| ] |
| }, |
| "model.sam_model.blocks.11.mlp.lin2": { |
| "bits": 8, |
| "in_features": 3072, |
| "out_features": 768, |
| "has_bias": true, |
| "original_shape": [ |
| 768, |
| 3072 |
| ] |
| }, |
| "model.vision_model.transformer.layers.0.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.0.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.0.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.0.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.1.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.1.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.1.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.1.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.2.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.2.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.2.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.2.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.3.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.3.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.3.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.3.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.4.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.4.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.4.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.4.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.5.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.5.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.5.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.5.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.6.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.6.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.6.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.6.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.7.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.7.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.7.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.7.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.8.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.8.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.8.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.8.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.9.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.9.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.9.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.9.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.10.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.10.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.10.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.10.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.11.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.11.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.11.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.11.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.12.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.12.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.12.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.12.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.13.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.13.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.13.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.13.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.14.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.14.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.14.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.14.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.15.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.15.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.15.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.15.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.16.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.16.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.16.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.16.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.17.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.17.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.17.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.17.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.18.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.18.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.18.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.18.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.19.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.19.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.19.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.19.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.20.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.20.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.20.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.20.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.21.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.21.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.21.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.21.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.22.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.22.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.22.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.22.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.vision_model.transformer.layers.23.self_attn.qkv_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 3072, |
| "has_bias": true, |
| "original_shape": [ |
| 3072, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.23.self_attn.out_proj": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.23.mlp.fc1": { |
| "bits": 8, |
| "in_features": 1024, |
| "out_features": 4096, |
| "has_bias": true, |
| "original_shape": [ |
| 4096, |
| 1024 |
| ] |
| }, |
| "model.vision_model.transformer.layers.23.mlp.fc2": { |
| "bits": 8, |
| "in_features": 4096, |
| "out_features": 1024, |
| "has_bias": true, |
| "original_shape": [ |
| 1024, |
| 4096 |
| ] |
| }, |
| "model.projector.layers": { |
| "bits": 8, |
| "in_features": 2048, |
| "out_features": 1280, |
| "has_bias": true, |
| "original_shape": [ |
| 1280, |
| 2048 |
| ] |
| }, |
| "lm_head": { |
| "bits": 8, |
| "in_features": 1280, |
| "out_features": 129280, |
| "has_bias": false, |
| "original_shape": [ |
| 129280, |
| 1280 |
| ] |
| } |
| }, |
| "stats": { |
| "original_size_mb": 6363.11767578125, |
| "compressed_size_mb": 3351.557418823242, |
| "compression_ratio": 1.898555471568018, |
| "vision_layers": 96, |
| "language_layers": 2197, |
| "total_quantized": 2342 |
| } |
| } |
