File size: 3,814 Bytes
77e20b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e5fd80a
 
 
0a7d026
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
77e20b1
 
9b07196
a917f9e
77e20b1
 
a917f9e
46085ee
a917f9e
00743f3
a917f9e
 
77e20b1
a917f9e
1d0db05
a917f9e
 
 
77e20b1
 
9e45379
 
77e20b1
 
a917f9e
77e20b1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a917f9e
77e20b1
a917f9e
77e20b1
 
 
a917f9e
77e20b1
 
 
 
 
 
 
0a7d026
a917f9e
77e20b1
 
 
 
a917f9e
77e20b1
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
---
language:
- en
tags:
- summarization
datasets:
- scientific_papers
metrics:
- rouge
model-index:
- name: ccdv/lsg-bart-base-16384-arxiv
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

**This model relies on a custom modeling file, you need to add trust_remote_code=True**\
**See [\#13467](https://github.com/huggingface/transformers/pull/13467)**

```python
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline

tokenizer = AutoTokenizer.from_pretrained("ccdv/lsg-bart-base-16384-arxiv", trust_remote_code=True)
model = AutoModelForSeq2SeqLM.from_pretrained("ccdv/lsg-bart-base-16384-arxiv", trust_remote_code=True)

text = "Replace by what you want."
pipe = pipeline("text2text-generation", model=model, tokenizer=tokenizer, device=0)
generated_text = pipe(
  text, 
  truncation=True, 
  max_length=64, 
  no_repeat_ngram_size=7,
  num_beams=2,
  early_stopping=True
  )
```

# ccdv/lsg-bart-base-16384-arxiv

This model is a fine-tuned version of [ccdv/lsg-bart-base-4096-arxiv](https://huggingface.co/ccdv/lsg-bart-base-4096-arxiv) on the [scientific_papers arxiv](https://huggingface.co/datasets/scientific_papers) dataset. \
The model is converted to handle 16384 long sequences and fine-tuned accordingly during 1 epoch. \
It achieves the following results on the test set:

| Length | Global tokens | Fine-tuning | Block Size | Sparsity | Connexions | R1    | R2    | RL    | RLsum |
|:------ |:------------- |:----------- |:---------- |:-------- | :--------- |:----- |:----- |:----- |:----- |
| 16384  | 64            | Full        | 256        | 0        | 768        | 48.74 | 20.88 | 28.50 | 44.23 |
| 16384  | 1             | Full        | 256        | 0        | 768        | 48.66 | 20.92 | 28.50 | 44.18 |
| 16384  | 64            | Global only | 256        | 0        | 768        | 48.08 | 20.42 | 28.00 | 43.65 |
| 16384  | 1             | None        | 256        | 0        | 768        | 47.03 | 20.19 | 28.26 | 42.69 |

Reference model:

| Length | Global tokens | Fine-tuning | Block Size | Sparsity | Connexions | R1    | R2    | RL    | RLsum |
|:------ |:------------- |:----------- |:---------- |:-------- | :--------- |:----- |:----- |:----- |:----- |
| 4096   | 1             | -           | 256        | 0        | 768        | 46.65 | 18.91 | 26.90 | 42.18 |

## Model description
The model relies on Local-Sparse-Global attention to handle long sequences:
![attn](attn.png)

The model has about ~145 millions parameters (6 encoder layers - 6 decoder layers). \
The model is warm started from [ccdv/lsg-bart-base-4096-arxiv](https://huggingface.co/ccdv/lsg-bart-base-4096-arxiv), converted to handle long sequences (encoder only) and fine tuned.

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 8e-05
- train_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 1.0

### Generate hyperparameters

The following hyperparameters were used during generation:
- dataset_name: scientific_papers
- dataset_config_name: arxiv
- eval_batch_size: 4
- eval_samples: 6440
- early_stopping: True
- ignore_pad_token_for_loss: True
- length_penalty: 2.0
- max_length: 320
- min_length: 32
- num_beams: 5
- no_repeat_ngram_size: None
- seed: 123

### Framework versions

- Transformers 4.18.0
- Pytorch 1.10.1+cu102
- Datasets 2.1.0
- Tokenizers 0.11.6