domenicrosati
commited on
Commit
•
f13bee7
1
Parent(s):
922bffd
Update README.md
Browse files
README.md
CHANGED
@@ -24,17 +24,41 @@ It achieves the following results on the evaluation set:
|
|
24 |
- Rougelsum: 85.4236
|
25 |
- Bleu: 72.1080
|
26 |
|
|
|
|
|
27 |
## Model description
|
28 |
|
29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
|
31 |
## Intended uses & limitations
|
32 |
|
33 |
-
|
34 |
|
35 |
## Training and evaluation data
|
36 |
|
37 |
-
|
38 |
|
39 |
## Training procedure
|
40 |
|
|
|
24 |
- Rougelsum: 85.4236
|
25 |
- Bleu: 72.1080
|
26 |
|
27 |
+
See: [https://wandb.ai/domenicrosati/huggingface/runs/n1yallpe](https://wandb.ai/domenicrosati/huggingface/runs/n1yallpe)
|
28 |
+
|
29 |
## Model description
|
30 |
|
31 |
+
A t5-model model to convert questions, answer pairs into statements.
|
32 |
+
|
33 |
+
The input should be all lower case and punctuation removed.
|
34 |
+
Use with `. ` as the seperator between question and answer.
|
35 |
+
> "where in the world is carmen. abruzzo"
|
36 |
+
> Output: "carmen is in abruzzo"
|
37 |
+
|
38 |
+
```
|
39 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
40 |
+
|
41 |
+
tokenizer = AutoTokenizer.from_pretrained('domenicrosati/QA2D-t5-small')
|
42 |
+
model = AutoModelForSeq2SeqLM.from_pretrained('domenicrosati/QA2D-t5-small')
|
43 |
+
|
44 |
+
question = "where in the world is carmen sandiego"
|
45 |
+
answer = "she is in abruzzo"
|
46 |
+
SEP = ". "
|
47 |
+
|
48 |
+
prompt = f'{question}{SEP}{answer}'
|
49 |
+
input_ids = tokenizer(prompt, return_tensors='pt').input_ids
|
50 |
+
output_ids = model.generate(input_ids)
|
51 |
+
responses = tokenizer.batch_decode(output_ids, skip_special_tokens=True)
|
52 |
+
# ['carmen sandiego is in abruzzo']
|
53 |
+
```
|
54 |
|
55 |
## Intended uses & limitations
|
56 |
|
57 |
+
To convert questions, answer pairs into statements.
|
58 |
|
59 |
## Training and evaluation data
|
60 |
|
61 |
+
Uses [QA2D](https://huggingface.co/datasets/domenicrosati/QA2D).
|
62 |
|
63 |
## Training procedure
|
64 |
|