Upload config.json with huggingface_hub
Browse files- config.json +21 -1
config.json
CHANGED
|
@@ -1,4 +1,7 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
| 2 |
"auto_map": {
|
| 3 |
"AutoConfig": "config.HATArchitectureConfig",
|
| 4 |
"AutoModelForCausalLM": "model.HATForCausalLM"
|
|
@@ -6,6 +9,9 @@
|
|
| 6 |
"backbone_config": {
|
| 7 |
"hidden_size": 4096,
|
| 8 |
"intermediate_size": 14336,
|
|
|
|
|
|
|
|
|
|
| 9 |
"max_position_embeddings": 32900,
|
| 10 |
"mlp_bias": false,
|
| 11 |
"num_attention_heads": 32,
|
|
@@ -31,12 +37,17 @@
|
|
| 31 |
"hidden_size": 1024,
|
| 32 |
"hidden_size_kv": 4096,
|
| 33 |
"hidden_size_q": 1024,
|
|
|
|
|
|
|
| 34 |
"num_attention_heads": 8,
|
| 35 |
"word_window_size": 1
|
| 36 |
},
|
| 37 |
"cross_attn_every_layer": true,
|
| 38 |
"hidden_size": 1024,
|
| 39 |
"intermediate_size": 2816,
|
|
|
|
|
|
|
|
|
|
| 40 |
"max_position_embeddings": 262144,
|
| 41 |
"mlp_bias": false,
|
| 42 |
"num_attention_heads": 8,
|
|
@@ -58,11 +69,16 @@
|
|
| 58 |
"hidden_size": 4096,
|
| 59 |
"hidden_size_kv": 1024,
|
| 60 |
"hidden_size_q": 4096,
|
|
|
|
|
|
|
| 61 |
"num_attention_heads": 32,
|
| 62 |
"word_window_size": 1
|
| 63 |
},
|
| 64 |
"hidden_size": 1024,
|
| 65 |
"intermediate_size": 2816,
|
|
|
|
|
|
|
|
|
|
| 66 |
"max_position_embeddings": 262144,
|
| 67 |
"mlp_bias": false,
|
| 68 |
"num_attention_heads": 8,
|
|
@@ -78,9 +94,13 @@
|
|
| 78 |
"use_cache": true,
|
| 79 |
"vocab_size": 256
|
| 80 |
},
|
|
|
|
|
|
|
| 81 |
"model_type": "hierarchical_autoregressive_transformer",
|
|
|
|
| 82 |
"special_token_dict": {
|
| 83 |
"<|eot_id|>": 192
|
| 84 |
},
|
|
|
|
| 85 |
"transformers_version": "4.46.3"
|
| 86 |
-
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"HATForCausalLM"
|
| 4 |
+
],
|
| 5 |
"auto_map": {
|
| 6 |
"AutoConfig": "config.HATArchitectureConfig",
|
| 7 |
"AutoModelForCausalLM": "model.HATForCausalLM"
|
|
|
|
| 9 |
"backbone_config": {
|
| 10 |
"hidden_size": 4096,
|
| 11 |
"intermediate_size": 14336,
|
| 12 |
+
"is_neox_style": true,
|
| 13 |
+
"key_query_norm": false,
|
| 14 |
+
"key_query_norm_per_head": false,
|
| 15 |
"max_position_embeddings": 32900,
|
| 16 |
"mlp_bias": false,
|
| 17 |
"num_attention_heads": 32,
|
|
|
|
| 37 |
"hidden_size": 1024,
|
| 38 |
"hidden_size_kv": 4096,
|
| 39 |
"hidden_size_q": 1024,
|
| 40 |
+
"key_query_norm": false,
|
| 41 |
+
"key_query_norm_per_head": false,
|
| 42 |
"num_attention_heads": 8,
|
| 43 |
"word_window_size": 1
|
| 44 |
},
|
| 45 |
"cross_attn_every_layer": true,
|
| 46 |
"hidden_size": 1024,
|
| 47 |
"intermediate_size": 2816,
|
| 48 |
+
"is_neox_style": true,
|
| 49 |
+
"key_query_norm": false,
|
| 50 |
+
"key_query_norm_per_head": false,
|
| 51 |
"max_position_embeddings": 262144,
|
| 52 |
"mlp_bias": false,
|
| 53 |
"num_attention_heads": 8,
|
|
|
|
| 69 |
"hidden_size": 4096,
|
| 70 |
"hidden_size_kv": 1024,
|
| 71 |
"hidden_size_q": 4096,
|
| 72 |
+
"key_query_norm": false,
|
| 73 |
+
"key_query_norm_per_head": false,
|
| 74 |
"num_attention_heads": 32,
|
| 75 |
"word_window_size": 1
|
| 76 |
},
|
| 77 |
"hidden_size": 1024,
|
| 78 |
"intermediate_size": 2816,
|
| 79 |
+
"is_neox_style": true,
|
| 80 |
+
"key_query_norm": false,
|
| 81 |
+
"key_query_norm_per_head": false,
|
| 82 |
"max_position_embeddings": 262144,
|
| 83 |
"mlp_bias": false,
|
| 84 |
"num_attention_heads": 8,
|
|
|
|
| 94 |
"use_cache": true,
|
| 95 |
"vocab_size": 256
|
| 96 |
},
|
| 97 |
+
"max_position_embeddings": 262144,
|
| 98 |
+
"max_word_size": 100,
|
| 99 |
"model_type": "hierarchical_autoregressive_transformer",
|
| 100 |
+
"sliding_window": 768,
|
| 101 |
"special_token_dict": {
|
| 102 |
"<|eot_id|>": 192
|
| 103 |
},
|
| 104 |
+
"torch_dtype": "bfloat16",
|
| 105 |
"transformers_version": "4.46.3"
|
| 106 |
+
}
|