(Trained with Unsloth)
Browse files- config.json +3 -2
config.json
CHANGED
|
@@ -41,6 +41,7 @@
|
|
| 41 |
],
|
| 42 |
"max_position_embeddings": 32768,
|
| 43 |
"max_window_layers": 28,
|
|
|
|
| 44 |
"model_type": "qwen2",
|
| 45 |
"num_attention_heads": 28,
|
| 46 |
"num_hidden_layers": 28,
|
|
@@ -51,9 +52,9 @@
|
|
| 51 |
"rope_theta": 1000000.0,
|
| 52 |
"sliding_window": null,
|
| 53 |
"tie_word_embeddings": false,
|
| 54 |
-
"transformers_version": "4.
|
| 55 |
"unsloth_fixed": true,
|
| 56 |
-
"unsloth_version": "2025.10.
|
| 57 |
"use_cache": true,
|
| 58 |
"use_sliding_window": false,
|
| 59 |
"vocab_size": 152064
|
|
|
|
| 41 |
],
|
| 42 |
"max_position_embeddings": 32768,
|
| 43 |
"max_window_layers": 28,
|
| 44 |
+
"model_name": "DidulaThavishaPro/exp_11_0_sft_16bit_vllm",
|
| 45 |
"model_type": "qwen2",
|
| 46 |
"num_attention_heads": 28,
|
| 47 |
"num_hidden_layers": 28,
|
|
|
|
| 52 |
"rope_theta": 1000000.0,
|
| 53 |
"sliding_window": null,
|
| 54 |
"tie_word_embeddings": false,
|
| 55 |
+
"transformers_version": "4.57.1",
|
| 56 |
"unsloth_fixed": true,
|
| 57 |
+
"unsloth_version": "2025.10.11",
|
| 58 |
"use_cache": true,
|
| 59 |
"use_sliding_window": false,
|
| 60 |
"vocab_size": 152064
|