Hugging Face's logo

Gabriel
/
bart-base-cnn-swe

Summarization
Transformers
PyTorch
TensorBoard
Swedish
bart
text2text-generation
Eval Results
Inference Endpoints
Model card Files Metrics Community
3
Gabriel commited on
Commit
c287791
•
1 Parent(s): 7ca68fe

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +31 -52
README.md CHANGED
@@ -1,67 +1,39 @@
1
  ---
2
- language: sv
3
  license: mit
4
- datasets:
5
- - Gabriel/cnn_daily_swe
6
  tags:
7
- - summarization
8
-
9
- widget:
10
- - text: "En kronologi av bombningar och försök bombattacker i det brittiska fastlandet sedan 1970-talet:. Polisen stänger gatorna runt Haymarket, i Londons livliga teaterdistrikt. 29 juni 2007: Polisen desarmerar en bomb bestående av 200 liter bränsle, gasflaskor och spikar som hittats i en övergiven bil i Haymarket i centrala London. En andra bil fylld med gas och spikar befanns senare ha parkerats bara några hundra meter från den första, innan den bogserades bort av trafikvakter i början av fredagen för att bryta parkeringsrestriktioner. Polisen säger att två fordon är tydligt kopplade. 21 juli 2005: Två veckor efter de dödliga 7/7 bombningarna påstås fyra män ha försökt genomföra en andra våg av attacker mot Londons transportnät vid tre tunnelbanestationer i London och ombord på en buss. Men deras påstådda ryggsäcksbomber exploderar inte. 7 juli 2005: Fyra självmordsbombare detonerar sig själva ombord på tre underjordiska tåg och en buss i en morgon rusningstid attack mot Londons transportnät, döda 52 människor och skada omkring 700 fler. Al-Qaida tar på sig ansvaret i ett videouttalande."
11
-
12
  model-index:
13
  - name: bart-base-cnn-swe
14
- results:
15
- - task:
16
- type: summarization
17
- name: summarization
18
- dataset:
19
- name: Gabriel/cnn_daily_swe
20
- type: Gabriel/cnn_daily_swe
21
- split: validation
22
- metrics:
23
- - name: Validation ROGUE-1
24
- type: rouge-1
25
- value: 21.7291
26
- verified: true
27
- - name: Validation ROGUE-2
28
- type: rouge-2
29
- value: 10.0209
30
- verified: true
31
- - name: Validation ROGUE-L
32
- type: rouge-l
33
- value: 17.775
34
- verified: true
35
  ---
36
 
37
- # bart-base-cnn-swe
38
-
39
- WORK IN PROGRESS!
40
- - 1. Further fine-tune on checkpoint with cnn-daily-swe.
41
- - 2. Further fine-tune on xsum-swe.
42
- - 3. Lastly fine-tune on smaller domain dataset.
43
 
44
- This model is a fine-tuned version of [KBLab/bart-base-swedish-cased](https://huggingface.co/KBLab/bart-base-swedish-cased) on the [Gabriel/bart-base-cnn-swe](https://huggingface.co/datasets/Gabriel/cnn_daily_swe) dataset.
45
 
 
46
  It achieves the following results on the evaluation set:
47
- - Loss: 2.1656
48
- - Rouge1: 21.7291
49
- - Rouge2: 10.0209
50
- - Rougel: 17.775
51
- - Rougelsum: 20.429
52
- - Gen Len: 19.9931
53
 
54
  ## Model description
55
 
56
- WORK IN PROGRESS!
57
 
58
  ## Intended uses & limitations
59
 
60
- WORK IN PROGRESS!
61
 
62
  ## Training and evaluation data
63
 
64
- WORK IN PROGRESS!
65
 
66
  ## Training procedure
67
 
@@ -69,20 +41,27 @@ WORK IN PROGRESS!
69
 
70
  The following hyperparameters were used during training:
71
  - learning_rate: 2e-05
72
- - train_batch_size: 1
73
- - eval_batch_size: 1
74
  - seed: 42
75
- - gradient_accumulation_steps: 16
76
  - total_train_batch_size: 16
77
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
78
  - lr_scheduler_type: linear
79
- - num_epochs: 1
80
  - mixed_precision_training: Native AMP
81
 
82
  ### Training results
83
 
84
- | Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
85
- |:-------------:|:-----:|:-----:|:---------------:|:-------:|:-------:|:------:|:---------:|:-------:|
86
- | 2.3512 | 1.0 | 17944 | 2.1656 | 21.7291 | 10.0209 | 17.775 | 20.429 | 19.9931 |
 
 
87
 
 
88
 
 
 
 
 
 
1
  ---
 
2
  license: mit
 
 
3
  tags:
4
+ - generated_from_trainer
5
+ metrics:
6
+ - rouge
 
 
7
  model-index:
8
  - name: bart-base-cnn-swe
9
+ results: []
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  ---
11
 
12
+ <!-- This model card has been generated automatically according to the information the Trainer had access to. You
13
+ should probably proofread and complete it, then remove this comment. -->
 
 
 
 
14
 
15
+ # bart-base-cnn-swe
16
 
17
+ This model is a fine-tuned version of [Gabriel/bart-base-cnn-swe](https://huggingface.co/Gabriel/bart-base-cnn-swe) on the None dataset.
18
  It achieves the following results on the evaluation set:
19
+ - Loss: 2.0253
20
+ - Rouge1: 22.0568
21
+ - Rouge2: 10.3302
22
+ - Rougel: 18.0648
23
+ - Rougelsum: 20.7482
24
+ - Gen Len: 19.9996
25
 
26
  ## Model description
27
 
28
+ More information needed
29
 
30
  ## Intended uses & limitations
31
 
32
+ More information needed
33
 
34
  ## Training and evaluation data
35
 
36
+ More information needed
37
 
38
  ## Training procedure
39
 
 
41
 
42
  The following hyperparameters were used during training:
43
  - learning_rate: 2e-05
44
+ - train_batch_size: 8
45
+ - eval_batch_size: 8
46
  - seed: 42
47
+ - gradient_accumulation_steps: 2
48
  - total_train_batch_size: 16
49
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
  - lr_scheduler_type: linear
51
+ - num_epochs: 2
52
  - mixed_precision_training: Native AMP
53
 
54
  ### Training results
55
 
56
+ | Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
57
+ |:-------------:|:-----:|:-----:|:---------------:|:-------:|:-------:|:-------:|:---------:|:-------:|
58
+ | 2.2349 | 1.0 | 17944 | 2.0643 | 21.9564 | 10.2133 | 17.9958 | 20.6502 | 19.9992 |
59
+ | 2.0726 | 2.0 | 35888 | 2.0253 | 22.0568 | 10.3302 | 18.0648 | 20.7482 | 19.9996 |
60
+
61
 
62
+ ### Framework versions
63
 
64
+ - Transformers 4.22.0
65
+ - Pytorch 1.12.1+cu113
66
+ - Datasets 2.4.0
67
+ - Tokenizers 0.12.1