|
{ |
|
"_name_or_path": "kawn_idefics3_siglib14_384", |
|
"architectures": [ |
|
"KawnIdefics3ForConditionalGeneration" |
|
], |
|
"ignore_index": -100, |
|
"image_grid_pinpoints": [ |
|
[ |
|
256, |
|
256 |
|
], |
|
[ |
|
256, |
|
512 |
|
], |
|
[ |
|
256, |
|
768 |
|
], |
|
[ |
|
256, |
|
1024 |
|
], |
|
[ |
|
256, |
|
1280 |
|
], |
|
[ |
|
1280, |
|
256 |
|
], |
|
[ |
|
1024, |
|
256 |
|
], |
|
[ |
|
768, |
|
256 |
|
], |
|
[ |
|
512, |
|
256 |
|
], |
|
[ |
|
512, |
|
512 |
|
], |
|
[ |
|
512, |
|
768 |
|
], |
|
[ |
|
512, |
|
1024 |
|
], |
|
[ |
|
512, |
|
1280 |
|
], |
|
[ |
|
1280, |
|
512 |
|
], |
|
[ |
|
1024, |
|
512 |
|
], |
|
[ |
|
768, |
|
512 |
|
], |
|
[ |
|
768, |
|
1024 |
|
], |
|
[ |
|
1024, |
|
768 |
|
], |
|
[ |
|
1280, |
|
768 |
|
], |
|
[ |
|
768, |
|
1280 |
|
], |
|
[ |
|
768, |
|
768 |
|
], |
|
[ |
|
1024, |
|
1024 |
|
], |
|
[ |
|
1024, |
|
1280 |
|
], |
|
[ |
|
1280, |
|
1024 |
|
], |
|
[ |
|
1280, |
|
1280 |
|
] |
|
], |
|
"image_seq_length": 64, |
|
"image_token_id": 255030, |
|
"model_type": "kawn_idefics3", |
|
"scale_factor": 2, |
|
"text_config": { |
|
"_attn_implementation_autoset": true, |
|
"_name_or_path": "CohereForAI/aya-expanse-8b", |
|
"architectures": [ |
|
"CohereForCausalLM" |
|
], |
|
"hidden_size": 4096, |
|
"intermediate_size": 14336, |
|
"logit_scale": 0.125, |
|
"model_type": "cohere", |
|
"num_attention_heads": 32, |
|
"num_hidden_layers": 32, |
|
"num_key_value_heads": 8, |
|
"torch_dtype": "bfloat16" |
|
}, |
|
"torch_dtype": "bfloat16", |
|
"transformers_version": "4.47.1", |
|
"use_cache": true, |
|
"vision_config": { |
|
"_attn_implementation_autoset": true, |
|
"hidden_size": 768, |
|
"image_size": 256, |
|
"intermediate_size": 3072, |
|
"model_type": "siglip_vision_model", |
|
"num_attention_heads": 12, |
|
"num_hidden_layers": 12, |
|
"patch_size": 16 |
|
}, |
|
"vision_feature_layer": null |
|
} |
|
|