Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
base_model:
|
| 3 |
+
- Qwen/Qwen2.5-Coder-7B-Instruct
|
| 4 |
+
---
|
| 5 |
+
|
| 6 |
+
## Automated benchmark (due to time constrains)
|
| 7 |
+
This benchmark compares the Qwen2.5-Coder-7B-Instruct model with this servicenow finetune, Qwen QwQ 32B and Quasar-Alpha (Secret new model on Openrouter, coding comparable to DeepSeek V3 and GPT 4o, https://openrouter.ai/openrouter/quasar-alpha).
|
| 8 |
+
DeepSeek R1 evaluated the results of each benchmark question.
|
| 9 |
+
|
| 10 |
+
**Please note:** This process definitly needs some improvements, for a general overview it should be good enough tho
|
| 11 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/675da3e0e84072d430260442/N-UoJgRAHGNTwtC0njLB5.png" width="750" title="" alt=""/>
|
| 12 |
+
|
| 13 |
+
Results were okay but not as good as i wanted, definitly taking another look at the training data and different approaches
|