Commit
•
baa4d28
1
Parent(s):
4aaea00
Update README.md
Browse files
README.md
CHANGED
@@ -121,6 +121,9 @@ pipeline_tag: text-generation
|
|
121 |
This is quantized version of [mlabonne/NeuralDaredevil-8B-abliterated](https://huggingface.co/mlabonne/NeuralDaredevil-8B-abliterated) created using llama.cpp
|
122 |
|
123 |
# Model Description
|
|
|
|
|
|
|
124 |
This is a DPO fine-tune of [mlabonne/Daredevil-8-abliterated](https://huggingface.co/mlabonne/Daredevil-8B-abliterated), trained on one epoch of [mlabonne/orpo-dpo-mix-40k](https://huggingface.co/datasets/mlabonne/orpo-dpo-mix-40k).
|
125 |
The DPO fine-tuning successfully recovers the performance loss due to the abliteration process, making it an excellent uncensored model.
|
126 |
|
@@ -156,28 +159,3 @@ Evaluation performed using [LLM AutoEval](https://github.com/mlabonne/llm-autoev
|
|
156 |
## 🌳 Model family tree
|
157 |
|
158 |
![image/png](https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/ekwRGgnjzEOyprT8sEBFt.png)
|
159 |
-
|
160 |
-
## 💻 Usage
|
161 |
-
|
162 |
-
```python
|
163 |
-
!pip install -qU transformers accelerate
|
164 |
-
|
165 |
-
from transformers import AutoTokenizer
|
166 |
-
import transformers
|
167 |
-
import torch
|
168 |
-
|
169 |
-
model = "mlabonne/Daredevil-8B"
|
170 |
-
messages = [{"role": "user", "content": "What is a large language model?"}]
|
171 |
-
|
172 |
-
tokenizer = AutoTokenizer.from_pretrained(model)
|
173 |
-
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
|
174 |
-
pipeline = transformers.pipeline(
|
175 |
-
"text-generation",
|
176 |
-
model=model,
|
177 |
-
torch_dtype=torch.float16,
|
178 |
-
device_map="auto",
|
179 |
-
)
|
180 |
-
|
181 |
-
outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
|
182 |
-
print(outputs[0]["generated_text"])
|
183 |
-
```
|
|
|
121 |
This is quantized version of [mlabonne/NeuralDaredevil-8B-abliterated](https://huggingface.co/mlabonne/NeuralDaredevil-8B-abliterated) created using llama.cpp
|
122 |
|
123 |
# Model Description
|
124 |
+
|
125 |
+
![image/jpeg](https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/gFEhcIDSKa3AWpkNfH91q.jpeg)
|
126 |
+
|
127 |
This is a DPO fine-tune of [mlabonne/Daredevil-8-abliterated](https://huggingface.co/mlabonne/Daredevil-8B-abliterated), trained on one epoch of [mlabonne/orpo-dpo-mix-40k](https://huggingface.co/datasets/mlabonne/orpo-dpo-mix-40k).
|
128 |
The DPO fine-tuning successfully recovers the performance loss due to the abliteration process, making it an excellent uncensored model.
|
129 |
|
|
|
159 |
## 🌳 Model family tree
|
160 |
|
161 |
![image/png](https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/ekwRGgnjzEOyprT8sEBFt.png)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|