| { |
| "backbone.model.embeddings.tok_embeddings.weight": { |
| "shape": [ |
| 50368, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.embeddings.norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.0.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.0.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.0.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.0.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.0.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.1.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.1.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.1.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.1.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.1.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.1.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.2.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.2.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.2.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.2.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.2.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.2.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.3.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.3.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.3.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.3.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.3.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.3.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.4.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.4.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.4.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.4.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.4.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.4.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.5.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.5.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.5.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.5.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.5.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.5.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.6.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.6.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.6.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.6.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.6.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.6.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.7.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.7.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.7.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.7.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.7.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.7.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.8.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.8.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.8.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.8.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.8.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.8.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.9.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.9.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.9.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.9.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.9.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.9.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.10.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.10.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.10.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.10.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.10.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.10.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.11.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.11.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.11.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.11.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.11.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.11.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.12.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.12.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.12.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.12.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.12.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.12.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.13.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.13.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.13.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.13.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.13.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.13.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.14.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.14.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.14.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.14.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.14.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.14.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.15.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.15.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.15.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.15.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.15.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.15.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.16.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.16.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.16.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.16.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.16.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.16.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.17.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.17.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.17.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.17.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.17.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.17.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.18.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.18.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.18.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.18.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.18.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.18.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.19.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.19.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.19.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.19.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.19.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.19.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.20.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.20.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.20.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.20.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.20.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.20.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.21.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.21.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.21.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.21.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.21.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.21.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.22.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.22.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.22.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.22.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.22.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.22.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.23.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.23.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.23.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.23.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.23.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.23.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.24.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.24.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.24.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.24.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.24.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.24.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.25.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.25.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.25.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.25.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.25.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.25.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.26.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.26.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.26.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.26.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.26.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.26.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.27.attn_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.27.attn.Wqkv.weight": { |
| "shape": [ |
| 3072, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.27.attn.Wo.weight": { |
| "shape": [ |
| 1024, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.27.mlp_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.27.mlp.Wi.weight": { |
| "shape": [ |
| 5248, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.layers.27.mlp.Wo.weight": { |
| "shape": [ |
| 1024, |
| 2624 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "backbone.model.final_norm.weight": { |
| "shape": [ |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_claim_present.weight": { |
| "shape": [ |
| 1, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_claim_present.bias": { |
| "shape": [ |
| 1 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_predicate.weight": { |
| "shape": [ |
| 21, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_predicate.bias": { |
| "shape": [ |
| 21 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_subject_type.weight": { |
| "shape": [ |
| 13, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_subject_type.bias": { |
| "shape": [ |
| 13 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_confidence.weight": { |
| "shape": [ |
| 1, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_confidence.bias": { |
| "shape": [ |
| 1 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_span.weight": { |
| "shape": [ |
| 2, |
| 1024 |
| ], |
| "dtype": "bfloat16" |
| }, |
| "head_span.bias": { |
| "shape": [ |
| 2 |
| ], |
| "dtype": "bfloat16" |
| } |
| } |