Update README.md

README.md

@@ -5,6 +5,7 @@ datasets:
 - Sao10K/Claude-3-Opus-Instruct-15K
 - Sao10K/Short-Storygen-v2
 - Sao10K/c2-Logs-Filtered
+- Gryphe/Sonnet3.5-SlimOrcaDedupCleaned-20k
 language:
 - en
 license: cc-by-nc-4.0
@@ -15,26 +16,6 @@ tags:
 
 # synk/8B-Stheno-Diverse-v3.2-MLX
 
-The Model [synk/8B-Stheno-Diverse-v3.2-MLX](https://huggingface.co/synk/8B-Stheno-Diverse-v3.2-MLX) was converted to MLX format from [synk/L3-8B-Stheno-v3.2-MLX](https://huggingface.co/synk/L3-8B-Stheno-v3.2-MLX) using mlx-lm version **0.19.3**.
+This model is [Stheno-v3.2 MLX](https://huggingface.co/synk/L3-8B-Stheno-v3.2-MLX) fine tuned on data from Claude Sonnet, where only the most _diverse_ responses were given. This was fine tuned on [Sonnet3.5-SlimOrcaDedupCleaned-20k](https://huggingface.co/datasets/Gryphe/Sonnet3.5-SlimOrcaDedupCleaned-20k) by Gryphe.
 
-## Use with mlx
-
-```bash
-pip install mlx-lm
-```
-
-```python
-from mlx_lm import load, generate
-
-model, tokenizer = load("synk/8B-Stheno-Diverse-v3.2-MLX")
-
-prompt="hello"
-
-if hasattr(tokenizer, "apply_chat_template") and tokenizer.chat_template is not None:
-    messages = [{"role": "user", "content": prompt}]
-    prompt = tokenizer.apply_chat_template(
-        messages, tokenize=False, add_generation_prompt=True
-    )
-
-response = generate(model, tokenizer, prompt=prompt, verbose=True)
-```
+It was fine-tuned using LoRA with MLX for 600 iterations.