Update README.md
Browse files
README.md
CHANGED
|
@@ -16,14 +16,13 @@ This model is based on 283 specific terms and definitions of French cuisine.
|
|
| 16 |
|
| 17 |
# Fine Tuning
|
| 18 |
|
| 19 |
-
Fine tuning done efficiently with Unsloth,
|
| 20 |
-
with which I saved processing time on a single T4 GPU (AzureML compute instance).
|
| 21 |
-
|
| 22 |
For this version of the model I experimented a training method with a double fine-tuning, SFT then DPO.
|
| 23 |
I generated two datasets exclusively for this model, with GPT-4o deployed on Azure OpenAI.
|
| 24 |
The challenge was to achieve a consistent alignment between the two fine-tuning methods.
|
| 25 |
SFT to teach the terms and DPO to reinforce the understanding achieved during the first learning.
|
| 26 |
|
|
|
|
|
|
|
| 27 |
# Usage
|
| 28 |
|
| 29 |
The recommended usage is by loading the low-rank adapter using unsloth:
|
|
|
|
| 16 |
|
| 17 |
# Fine Tuning
|
| 18 |
|
|
|
|
|
|
|
|
|
|
| 19 |
For this version of the model I experimented a training method with a double fine-tuning, SFT then DPO.
|
| 20 |
I generated two datasets exclusively for this model, with GPT-4o deployed on Azure OpenAI.
|
| 21 |
The challenge was to achieve a consistent alignment between the two fine-tuning methods.
|
| 22 |
SFT to teach the terms and DPO to reinforce the understanding achieved during the first learning.
|
| 23 |
|
| 24 |
+
Fine tuning done efficiently with Unsloth, with which I saved processing time on a single T4 GPU (AzureML compute instance).
|
| 25 |
+
|
| 26 |
# Usage
|
| 27 |
|
| 28 |
The recommended usage is by loading the low-rank adapter using unsloth:
|