Update config.json
Browse files- config.json +7 -7
config.json
CHANGED
@@ -2,25 +2,25 @@
|
|
2 |
"architectures": [
|
3 |
"AquilaDenseForCausalLM"
|
4 |
],
|
|
|
5 |
"auto_map": {
|
6 |
"AutoConfig": "configuration_aquiladense.AquilaDenseConfig",
|
7 |
"AutoModelForCausalLM": "modeling_aquiladense.AquilaDenseForCausalLM"
|
8 |
},
|
9 |
-
"attention_dropout": 0.0,
|
10 |
"bos_token_id": 151849,
|
11 |
"eos_token_id": 151850,
|
12 |
"hidden_act": "silu",
|
13 |
-
"hidden_size":
|
14 |
"initializer_range": 0.02,
|
15 |
-
"intermediate_size":
|
16 |
-
"rms_norm_eps": 1e-05,
|
17 |
"max_position_embeddings": 4096,
|
18 |
"model_type": "aquiladense",
|
19 |
-
"num_attention_heads":
|
20 |
-
"num_hidden_layers":
|
21 |
"num_key_value_heads": 8,
|
22 |
"pad_token_id": 151643,
|
23 |
"pretraining_tp": 1,
|
|
|
24 |
"rope_scaling": null,
|
25 |
"rope_theta": 1000000.0,
|
26 |
"tie_word_embeddings": false,
|
@@ -29,4 +29,4 @@
|
|
29 |
"_attn_implementation": "flash_attention_2",
|
30 |
"use_cache": true,
|
31 |
"vocab_size": 151851
|
32 |
-
}
|
|
|
2 |
"architectures": [
|
3 |
"AquilaDenseForCausalLM"
|
4 |
],
|
5 |
+
"attention_dropout": 0.0,
|
6 |
"auto_map": {
|
7 |
"AutoConfig": "configuration_aquiladense.AquilaDenseConfig",
|
8 |
"AutoModelForCausalLM": "modeling_aquiladense.AquilaDenseForCausalLM"
|
9 |
},
|
|
|
10 |
"bos_token_id": 151849,
|
11 |
"eos_token_id": 151850,
|
12 |
"hidden_act": "silu",
|
13 |
+
"hidden_size": 4096,
|
14 |
"initializer_range": 0.02,
|
15 |
+
"intermediate_size": 14336,
|
|
|
16 |
"max_position_embeddings": 4096,
|
17 |
"model_type": "aquiladense",
|
18 |
+
"num_attention_heads": 32,
|
19 |
+
"num_hidden_layers": 32,
|
20 |
"num_key_value_heads": 8,
|
21 |
"pad_token_id": 151643,
|
22 |
"pretraining_tp": 1,
|
23 |
+
"rms_norm_eps": 1e-05,
|
24 |
"rope_scaling": null,
|
25 |
"rope_theta": 1000000.0,
|
26 |
"tie_word_embeddings": false,
|
|
|
29 |
"_attn_implementation": "flash_attention_2",
|
30 |
"use_cache": true,
|
31 |
"vocab_size": 151851
|
32 |
+
}
|