Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,78 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
language:
|
3 |
+
- en
|
4 |
+
tags:
|
5 |
+
- text-generation
|
6 |
+
- finetuned
|
7 |
+
widget:
|
8 |
+
- text: Is it raining in Hawaii?
|
9 |
+
example_title: Is it raining in Hawaii?
|
10 |
+
- text: Turn off music.
|
11 |
+
example_title: Turn off music.
|
12 |
+
- text: Is it snowing in New York?
|
13 |
+
example_title: Is it snowing in New York?
|
14 |
+
datasets:
|
15 |
+
- commonsense_qa
|
16 |
+
license: apache-2.0
|
17 |
+
pipeline_tag: text-generation
|
18 |
+
---
|
19 |
+
|
20 |
+
# Commonsense-QA-Mistral-7B
|
21 |
+
|
22 |
+
This is a finetuned model of [mistralai/Mistral-7B-Instruct-v0.1](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1)
|
23 |
+
with [commonsense_qa](https://huggingface.co/datasets/commonsense_qa) dataset.
|
24 |
+
|
25 |
+
The model is loaded in 4-bit and fine-tuned with LoRA.
|
26 |
+
|
27 |
+
## Usage
|
28 |
+
|
29 |
+
### Loading of model:
|
30 |
+
```python
|
31 |
+
# Load model directly
|
32 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
33 |
+
|
34 |
+
model = AutoModelForCausalLM.from_pretrained(
|
35 |
+
"rvv-karma/Commonsense-QA-Mistral-7B",
|
36 |
+
low_cpu_mem_usage=True,
|
37 |
+
return_dict=True,
|
38 |
+
torch_dtype=torch.bfloat16,
|
39 |
+
device_map="auto",
|
40 |
+
)
|
41 |
+
|
42 |
+
tokenizer = AutoTokenizer.from_pretrained("rvv-karma/Commonsense-QA-Mistral-7B", trust_remote_code=True)
|
43 |
+
tokenizer.pad_token = tokenizer.eos_token
|
44 |
+
tokenizer.padding_side = "left"
|
45 |
+
```
|
46 |
+
|
47 |
+
### Sample:
|
48 |
+
```python
|
49 |
+
pipe = pipeline(
|
50 |
+
task="text-generation",
|
51 |
+
model=model,
|
52 |
+
tokenizer=tokenizer,
|
53 |
+
return_full_text=False,
|
54 |
+
pad_token_id=tokenizer.pad_token_id,
|
55 |
+
eos_token_id=13,
|
56 |
+
max_new_tokens=8
|
57 |
+
)
|
58 |
+
|
59 |
+
prompt = """<s>
|
60 |
+
QUESTION:
|
61 |
+
The sensor would just the distance then set off an alarm, the installation expert explained it was called a what kind of sensor?
|
62 |
+
|
63 |
+
OPTIONS:
|
64 |
+
["near", "closeness", "here", "proximity", "this"]
|
65 |
+
|
66 |
+
ANSWER:
|
67 |
+
"""
|
68 |
+
result = pipe(prompt)
|
69 |
+
generated = result[0]['generated_text']
|
70 |
+
print(generated)
|
71 |
+
|
72 |
+
# Output: proximity
|
73 |
+
```
|
74 |
+
|
75 |
+
|
76 |
+
## Fine-tuning script
|
77 |
+
|
78 |
+
[Kaggle Notebook](https://www.kaggle.com/rvkarma/commonsense-qa-mistral-7b)
|