Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,32 @@
|
|
1 |
-
---
|
2 |
-
license: apache-2.0
|
3 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
---
|
4 |
+
# Model Card for LoRA-FLAN-T5 large
|
5 |
+
|
6 |
+
This repository contains the LoRA (Low Rank Adapters) of `flan-t5-base` that has been fine-tuned on [`knkarthick/dialogsum`](https://huggingface.co/datasets/knkarthick/dialogsum) dataset.
|
7 |
+
|
8 |
+
## Usage
|
9 |
+
|
10 |
+
Use this adapter with `peft` library
|
11 |
+
|
12 |
+
```python
|
13 |
+
# pip install peft transformers
|
14 |
+
import torch
|
15 |
+
from peft import PeftModel, PeftConfig
|
16 |
+
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
|
17 |
+
|
18 |
+
base_model_id = "google/flan-t5-base"
|
19 |
+
peft_model_id = "Jennny/flan-t5-base-erionis-summarization-exp"
|
20 |
+
|
21 |
+
# Load tokenizer and models
|
22 |
+
tokenizer = AutoTokenizer.from_pretrained(base_model_id)
|
23 |
+
base_model = AutoModelForSeq2SeqLM.from_pretrained(
|
24 |
+
base_model_id, torch_dtype=torch.bfloat16
|
25 |
+
).to(device)
|
26 |
+
model = PeftModel.from_pretrained(
|
27 |
+
base_model_id,
|
28 |
+
peft_model_id,
|
29 |
+
torch_dtype=torch.bfloat16,
|
30 |
+
is_trainable=False,
|
31 |
+
).to(device)
|
32 |
+
```
|