burak commited on
Update config.json
Browse files- config.json +18 -2
config.json
CHANGED
|
@@ -1,9 +1,25 @@
|
|
| 1 |
{
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
"vocab_size": 8000,
|
| 3 |
"n_embd": 64,
|
| 4 |
-
"block_size": 512,
|
| 5 |
"n_layer": 4,
|
| 6 |
"n_head": 4,
|
|
|
|
|
|
|
| 7 |
"dropout": 0.0,
|
| 8 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"model_type": "sykollm",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"SykoLLM"
|
| 5 |
+
],
|
| 6 |
+
"class_name": "SykoLLM",
|
| 7 |
+
"description": "Custom NanoGPT style decoder-only transformer model",
|
| 8 |
"vocab_size": 8000,
|
| 9 |
"n_embd": 64,
|
|
|
|
| 10 |
"n_layer": 4,
|
| 11 |
"n_head": 4,
|
| 12 |
+
"block_size": 512,
|
| 13 |
+
"max_position_embeddings": 512,
|
| 14 |
"dropout": 0.0,
|
| 15 |
+
"bias": true,
|
| 16 |
+
"layer_norm_epsilon": 1e-05,
|
| 17 |
+
"initializer_range": 0.02,
|
| 18 |
+
"bos_token_id": 1,
|
| 19 |
+
"eos_token_id": 2,
|
| 20 |
+
"pad_token_id": 0,
|
| 21 |
+
"unk_token_id": 3,
|
| 22 |
+
"use_cache": true,
|
| 23 |
+
"device": "cuda",
|
| 24 |
+
"version": "1.0"
|
| 25 |
}
|