Update config.json
Browse filesremoved redundant entry
- config.json +3 -4
config.json
CHANGED
|
@@ -1,10 +1,9 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "/proj/checkpoints/ashishagr/thinking-granite/outputs/tpo_final/8b/exp_10/iter4/dpo",
|
| 3 |
"architectures": [
|
| 4 |
"GraniteForCausalLM"
|
| 5 |
],
|
| 6 |
"attention_bias": false,
|
| 7 |
-
"attention_dropout": 0.
|
| 8 |
"attention_multiplier": 0.0078125,
|
| 9 |
"bos_token_id": 0,
|
| 10 |
"embedding_multiplier": 12.0,
|
|
@@ -27,7 +26,7 @@
|
|
| 27 |
"rope_theta": 10000000.0,
|
| 28 |
"tie_word_embeddings": true,
|
| 29 |
"torch_dtype": "bfloat16",
|
| 30 |
-
"transformers_version": "4.47.0
|
| 31 |
"use_cache": true,
|
| 32 |
"vocab_size": 49155
|
| 33 |
-
}
|
|
|
|
| 1 |
{
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
"GraniteForCausalLM"
|
| 4 |
],
|
| 5 |
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.1,
|
| 7 |
"attention_multiplier": 0.0078125,
|
| 8 |
"bos_token_id": 0,
|
| 9 |
"embedding_multiplier": 12.0,
|
|
|
|
| 26 |
"rope_theta": 10000000.0,
|
| 27 |
"tie_word_embeddings": true,
|
| 28 |
"torch_dtype": "bfloat16",
|
| 29 |
+
"transformers_version": "4.47.0",
|
| 30 |
"use_cache": true,
|
| 31 |
"vocab_size": 49155
|
| 32 |
+
}
|