Commit
•
53a21a8
1
Parent(s):
2309c24
Update README.md
Browse files
README.md
CHANGED
@@ -7,13 +7,14 @@ datasets:
|
|
7 |
|
8 |
# Phi-2-ORPO
|
9 |
|
10 |
-
**
|
11 |
preference dataset using **Odds Ratio Preference Optimization (ORPO)**. The model has been trained for 1 epoch.
|
12 |
|
13 |
## LazyORPO
|
14 |
|
15 |
This model has been trained using **[LazyORPO](https://colab.research.google.com/drive/19ci5XIcJDxDVPY2xC1ftZ5z1kc2ah_rx?usp=sharing)**. A colab notebook that makes the training
|
16 |
-
process much easier. Based on [ORPO paper](https://colab.research.google.com/corgiredirector?site=https%3A%2F%2Fhuggingface.co%2Fpapers%2F2403.07691)
|
|
|
17 |
|
18 |
#### What is ORPO?
|
19 |
|
@@ -37,7 +38,7 @@ torch.set_default_device("cuda")
|
|
37 |
model = AutoModelForCausalLM.from_pretrained("abideen/phi2-pro", torch_dtype="auto", trust_remote_code=True)
|
38 |
tokenizer = AutoTokenizer.from_pretrained("abideen/phi2-pro", trust_remote_code=True)
|
39 |
|
40 |
-
inputs = tokenizer('''
|
41 |
"""
|
42 |
Write a detailed analogy between mathematics and a lighthouse.
|
43 |
"""''', return_tensors="pt", return_attention_mask=False)
|
|
|
7 |
|
8 |
# Phi-2-ORPO
|
9 |
|
10 |
+
**Phi-2-ORPO** is a fine-tuned version of **[microsoft/phi-2](https://huggingface.co/microsoft/phi-2)** on **[argilla/dpo-mix-7k](https://huggingface.co/datasets/argilla/dpo-mix-7k)**
|
11 |
preference dataset using **Odds Ratio Preference Optimization (ORPO)**. The model has been trained for 1 epoch.
|
12 |
|
13 |
## LazyORPO
|
14 |
|
15 |
This model has been trained using **[LazyORPO](https://colab.research.google.com/drive/19ci5XIcJDxDVPY2xC1ftZ5z1kc2ah_rx?usp=sharing)**. A colab notebook that makes the training
|
16 |
+
process much easier. Based on [ORPO paper](https://colab.research.google.com/corgiredirector?site=https%3A%2F%2Fhuggingface.co%2Fpapers%2F2403.07691). This notebook has been created by
|
17 |
+
**[Zain Ul Abideen](https://huggingface.co/abideen)**
|
18 |
|
19 |
#### What is ORPO?
|
20 |
|
|
|
38 |
model = AutoModelForCausalLM.from_pretrained("abideen/phi2-pro", torch_dtype="auto", trust_remote_code=True)
|
39 |
tokenizer = AutoTokenizer.from_pretrained("abideen/phi2-pro", trust_remote_code=True)
|
40 |
|
41 |
+
inputs = tokenizer('''
|
42 |
"""
|
43 |
Write a detailed analogy between mathematics and a lighthouse.
|
44 |
"""''', return_tensors="pt", return_attention_mask=False)
|