Upload config.json
Browse files- config.json +40 -0
config.json
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": ["BitMarModel"],
|
| 3 |
+
"auto_map": {
|
| 4 |
+
"AutoConfig": "modeling_bitmar.BitMarConfig",
|
| 5 |
+
"AutoModel": "modeling_bitmar.BitMarModel"
|
| 6 |
+
},
|
| 7 |
+
"model_type": "bitmar",
|
| 8 |
+
"vocab_size": 50257,
|
| 9 |
+
"text_encoder_dim": 128,
|
| 10 |
+
"text_encoder_layers": 4,
|
| 11 |
+
"text_encoder_heads": 4,
|
| 12 |
+
"text_decoder_dim": 128,
|
| 13 |
+
"text_decoder_layers": 4,
|
| 14 |
+
"text_decoder_heads": 4,
|
| 15 |
+
"vision_encoder_dim": 768,
|
| 16 |
+
"vision_latent_size": 128,
|
| 17 |
+
"vision_hidden_size": 64,
|
| 18 |
+
"vision_compression_method": "learned_compression",
|
| 19 |
+
"vision_spatial_pooling": true,
|
| 20 |
+
"vision_pool_size": 2,
|
| 21 |
+
"fusion_hidden_size": 128,
|
| 22 |
+
"fusion_num_heads": 4,
|
| 23 |
+
"fusion_num_layers": 2,
|
| 24 |
+
"memory_alpha": 0.2,
|
| 25 |
+
"direct_writing": true,
|
| 26 |
+
"memory_compression": true,
|
| 27 |
+
"max_seq_len": 256,
|
| 28 |
+
"dropout": 0.15,
|
| 29 |
+
"torch_dtype": "float32",
|
| 30 |
+
"transformers_version": "4.36.0",
|
| 31 |
+
"use_cache": true,
|
| 32 |
+
"tie_word_embeddings": true,
|
| 33 |
+
"initializer_range": 0.02,
|
| 34 |
+
"layer_norm_epsilon": 1e-5,
|
| 35 |
+
"pad_token_id": 50256,
|
| 36 |
+
"bos_token_id": 50256,
|
| 37 |
+
"eos_token_id": 50256,
|
| 38 |
+
"sep_token_id": null,
|
| 39 |
+
"decoder_start_token_id": null
|
| 40 |
+
}
|