Create config.json
Browse files- config.json +26 -0
config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "nz/RITA_s_HF",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"RITAForCausalLM"
|
| 5 |
+
],
|
| 6 |
+
"auto_map": {
|
| 7 |
+
"AutoConfig": "rita_configuration.RITAConfig",
|
| 8 |
+
"AutoModel": "rita_modeling.RITAForCausalLM"
|
| 9 |
+
},
|
| 10 |
+
"vocab_size": 26,
|
| 11 |
+
"model_dim": 768,
|
| 12 |
+
"num_heads": 12,
|
| 13 |
+
"feedforward_dim": 3072,
|
| 14 |
+
"num_layers": 12,
|
| 15 |
+
"max_seq_len": 1024,
|
| 16 |
+
"attention_dropout": 0.1,
|
| 17 |
+
"mlp_dropout": 0.0,
|
| 18 |
+
"bias": true,
|
| 19 |
+
"activation_function": "RITAGELU",
|
| 20 |
+
"gradient_checkpointing": true,
|
| 21 |
+
"use_cache": true,
|
| 22 |
+
"eos_token_id": 2,
|
| 23 |
+
|
| 24 |
+
"model_type": "rita",
|
| 25 |
+
"transformers_version": "4.18.0",
|
| 26 |
+
}
|