| --- |
| license: apache-2.0 |
| language: |
| - en |
| base_model: |
| - meta-llama/Llama-3.1-8B |
| pipeline_tag: text-generation |
| tags: |
| - art |
| --- |
| Cat0.2 |
|
|
|  |
|
|
| **Cat0.2** is a model further fine-tuned from **Cat0.1** using an iterative dataset approach. The process involves iterating over the dataset three times, using the previous generation as the next iteration's dataset. This model is based on an 8 billion parameter **Llama 3.1** architecture, further quantized to 4 bits for ease of upload. It has been iteratively trained on several hundred examples of data. |
|
|
| ## Iterative Process |
|
|
| 1. **Generate examples with the base model (Llama 3.1 8B).** |
| 2. **Fine-tune on those examples.** |
| 3. **Generate new examples with the updated model.** |
| 4. **Repeat the process.** |
|
|
| > **Note:** The model may exhibit some hallucinations in its output due to the iterative training approach and may be prone to repetition. |
