e7mac/vision-encoder

1{2  "dim": 4096,3  "ffn_dim_multiplier": 1.3,4  "multiple_of": 1024,5  "n_heads": 32,6  "n_kv_heads": 8,7  "n_layers": 32,8  "norm_eps": 1e-05,9  "rope_theta": 500000.0,10  "use_scaled_rope": true,11  "vision_chunk_size": 448,12  "vision_max_num_chunks": 4,13  "vocab_size": 128256,14  "vision_num_cross_attention_layers": 815}