Update config.json
Browse filesRefactor: Remove `_name_or_path` and increase `max_position_embeddings` to 8192
- Removed the `_name_or_path` parameter as it's no longer necessary.
- Updated `max_position_embeddings` from 2048 to 8192 for better model capacity.
- config.json +1 -2
config.json
CHANGED
@@ -1,5 +1,4 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "./models/llama3-8b_sft_full_1e-5_bs8000_3_trainv59_a/checkpoint-1596",
|
3 |
"architectures": [
|
4 |
"LlamaForCausalLM"
|
5 |
],
|
@@ -11,7 +10,7 @@
|
|
11 |
"hidden_size": 4096,
|
12 |
"initializer_range": 0.02,
|
13 |
"intermediate_size": 14336,
|
14 |
-
"max_position_embeddings":
|
15 |
"model_type": "llama",
|
16 |
"num_attention_heads": 32,
|
17 |
"num_hidden_layers": 32,
|
|
|
1 |
{
|
|
|
2 |
"architectures": [
|
3 |
"LlamaForCausalLM"
|
4 |
],
|
|
|
10 |
"hidden_size": 4096,
|
11 |
"initializer_range": 0.02,
|
12 |
"intermediate_size": 14336,
|
13 |
+
"max_position_embeddings": 4096,
|
14 |
"model_type": "llama",
|
15 |
"num_attention_heads": 32,
|
16 |
"num_hidden_layers": 32,
|