Upload 2 files
Browse files3 layers to 4 layers
- config.json +3 -3
- model.safetensors +2 -2
    	
        config.json
    CHANGED
    
    | @@ -6,6 +6,7 @@ | |
| 6 | 
             
              "attention_dropout": 0.0,
         | 
| 7 | 
             
              "attention_multiplier": 0.0078125,
         | 
| 8 | 
             
              "bos_token_id": 0,
         | 
|  | |
| 9 | 
             
              "embedding_multiplier": 12.0,
         | 
| 10 | 
             
              "eos_token_id": 0,
         | 
| 11 | 
             
              "hidden_act": "silu",
         | 
| @@ -17,7 +18,7 @@ | |
| 17 | 
             
              "mlp_bias": false,
         | 
| 18 | 
             
              "model_type": "granite",
         | 
| 19 | 
             
              "num_attention_heads": 32,
         | 
| 20 | 
            -
              "num_hidden_layers":  | 
| 21 | 
             
              "num_key_value_heads": 8,
         | 
| 22 | 
             
              "pad_token_id": 0,
         | 
| 23 | 
             
              "residual_multiplier": 0.22,
         | 
| @@ -25,8 +26,7 @@ | |
| 25 | 
             
              "rope_scaling": null,
         | 
| 26 | 
             
              "rope_theta": 10000000.0,
         | 
| 27 | 
             
              "tie_word_embeddings": false,
         | 
| 28 | 
            -
              " | 
| 29 | 
            -
              "transformers_version": "4.52.4",
         | 
| 30 | 
             
              "use_cache": true,
         | 
| 31 | 
             
              "vocab_size": 49159
         | 
| 32 | 
             
            }
         | 
|  | |
| 6 | 
             
              "attention_dropout": 0.0,
         | 
| 7 | 
             
              "attention_multiplier": 0.0078125,
         | 
| 8 | 
             
              "bos_token_id": 0,
         | 
| 9 | 
            +
              "dtype": "bfloat16",
         | 
| 10 | 
             
              "embedding_multiplier": 12.0,
         | 
| 11 | 
             
              "eos_token_id": 0,
         | 
| 12 | 
             
              "hidden_act": "silu",
         | 
|  | |
| 18 | 
             
              "mlp_bias": false,
         | 
| 19 | 
             
              "model_type": "granite",
         | 
| 20 | 
             
              "num_attention_heads": 32,
         | 
| 21 | 
            +
              "num_hidden_layers": 4,
         | 
| 22 | 
             
              "num_key_value_heads": 8,
         | 
| 23 | 
             
              "pad_token_id": 0,
         | 
| 24 | 
             
              "residual_multiplier": 0.22,
         | 
|  | |
| 26 | 
             
              "rope_scaling": null,
         | 
| 27 | 
             
              "rope_theta": 10000000.0,
         | 
| 28 | 
             
              "tie_word_embeddings": false,
         | 
| 29 | 
            +
              "transformers_version": "4.56.1",
         | 
|  | |
| 30 | 
             
              "use_cache": true,
         | 
| 31 | 
             
              "vocab_size": 49159
         | 
| 32 | 
             
            }
         | 
    	
        model.safetensors
    CHANGED
    
    | @@ -1,3 +1,3 @@ | |
| 1 | 
             
            version https://git-lfs.github.com/spec/v1
         | 
| 2 | 
            -
            oid sha256: | 
| 3 | 
            -
            size  | 
|  | |
| 1 | 
             
            version https://git-lfs.github.com/spec/v1
         | 
| 2 | 
            +
            oid sha256:ba34d26650e8891a09dfc0a6fcf3a8f9c27ad0afb0912a9596893285ea85610a
         | 
| 3 | 
            +
            size 2399334792
         | 
