Views
No views yet
| Parameter | Value |
|---|---|
| Base Model | google/flan-t5-small |
| Epochs | 5 |
| Batch Size | 16 (per device) |
| Learning Rate | 3e-4 |
| Optimizer | Adafactor |
| Mixed Precision | fp16 |
| Gradient Accumulation | 4 steps |
| Max Length | 512 tokens |
| Metric | Value |
|---|---|
| ROUGE-1 | 0.3722 |
| ROUGE-2 | 0.1066 |
| ROUGE-L | 0.2794 |
1from transformers import pipeline
2from datasets import load_dataset
3from evaluate import load
4
5summarizer = pipeline('summarization', model='ingu627/finetuned-flan-t5-dialogsum')
6dataset = load_dataset('knkarthick/dialogsum', split='test')
7basic_summarizer = pipeline('summarization', model='google/flan-t5-small')
8
9rouge = load('rouge')
10
11references = []
12predictions = []
13basic_predictions = []
14
15for example in dataset.select(range(50)):
16 generated = summarizer(
17 f"Summarize this dialogue:\n{example['dialogue']}\nSummary:",
18 max_length=135,
19 num_beams=3
20 )[0]['summary_text']
21 basic_generated = basic_summarizer(
22 f"Summarize this dialogue:\n{example['dialogue']}\nSummary:",
23 max_length=135,
24 num_beams=3
25 )[0]['summary_text']
26
27 references.append(example['summary'])
28 predictions.append(generated)
29 basic_predictions.append(basic_generated)
30
31fine_tuned_results = rouge.compute(
32 predictions=predictions,
33 references=references,
34 rouge_types=['rouge1', 'rouge2', 'rougeL'],
35 use_aggregator=True,
36 use_stemmer=True,
37)
38
39print(fine_tuned_results)