Hugging Face's logo

cristianglezm
/
ViT-GPT2-FlowerCaptioner

Image-to-Text
Transformers
Safetensors
English
vision-encoder-decoder
image-text-to-text
image-captioning
Inference Endpoints
Model card Files Community
cristianglezm commited on
Commit
742b953
•
1 Parent(s): c78eb64

new model version with better rouge

Browse files
Files changed (3) hide show
  1. README.md +32 -10
  2. pytorch_model.bin +1 -1
  3. training_args.bin +1 -1
README.md CHANGED
@@ -28,12 +28,12 @@ library_name: transformers
28
 
29
  This model is a fine-tuned version of [nlpconnect/vit-gpt2-image-captioning](https://huggingface.co/nlpconnect/vit-gpt2-image-captioning) on the [FlowerEvolver-dataset](https://huggingface.co/datasets/cristianglezm/FlowerEvolver-Dataset) dataset.
30
  It achieves the following results on the evaluation set:
31
- - Loss: 0.3075
32
- - Rouge1: 66.3702
33
- - Rouge2: 45.5642
34
- - Rougel: 61.401
35
- - Rougelsum: 64.0587
36
- - Gen Len: 49.97
37
 
38
  ## sample running code
39
 
@@ -74,15 +74,37 @@ The following hyperparameters were used during training:
74
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
75
  - lr_scheduler_type: linear
76
  - lr_scheduler_warmup_steps: 500
77
- - num_epochs: 3
78
 
79
  ### Training results
80
 
81
  | Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
82
  |:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|:-------:|:---------:|:-------:|
83
- | 0.6755 | 1.0 | 100 | 0.5339 | 60.9402 | 39.3331 | 54.6889 | 59.45 | 36.75 |
84
- | 0.3666 | 2.0 | 200 | 0.3331 | 65.5149 | 43.0245 | 59.3121 | 62.7329 | 52.82 |
85
- | 0.2983 | 3.0 | 300 | 0.3075 | 66.3702 | 45.5642 | 61.401 | 64.0587 | 49.97 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
86
 
87
 
88
  ### Framework versions
 
28
 
29
  This model is a fine-tuned version of [nlpconnect/vit-gpt2-image-captioning](https://huggingface.co/nlpconnect/vit-gpt2-image-captioning) on the [FlowerEvolver-dataset](https://huggingface.co/datasets/cristianglezm/FlowerEvolver-Dataset) dataset.
30
  It achieves the following results on the evaluation set:
31
+ - Loss: 0.4930
32
+ - Rouge1: 68.3498
33
+ - Rouge2: 46.7534
34
+ - Rougel: 62.3763
35
+ - Rougelsum: 65.9575
36
+ - Gen Len: 49.82
37
 
38
  ## sample running code
39
 
 
74
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
75
  - lr_scheduler_type: linear
76
  - lr_scheduler_warmup_steps: 500
77
+ - num_epochs: 25
78
 
79
  ### Training results
80
 
81
  | Training Loss | Epoch | Step | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
82
  |:-------------:|:-----:|:----:|:---------------:|:-------:|:-------:|:-------:|:---------:|:-------:|
83
+ | 0.6986 | 1.0 | 100 | 0.5339 | 64.9813 | 42.4686 | 58.2586 | 63.3933 | 47.25 |
84
+ | 0.3408 | 2.0 | 200 | 0.3263 | 67.5461 | 46.5219 | 62.7962 | 65.6509 | 47.39 |
85
+ | 0.2797 | 3.0 | 300 | 0.2829 | 65.0704 | 42.0682 | 58.4268 | 63.2368 | 56.8 |
86
+ | 0.2584 | 4.0 | 400 | 0.2588 | 65.5074 | 45.227 | 60.2469 | 63.4253 | 52.25 |
87
+ | 0.2589 | 5.0 | 500 | 0.2607 | 66.7346 | 45.8264 | 61.7373 | 64.8857 | 50.64 |
88
+ | 0.2179 | 6.0 | 600 | 0.2697 | 63.8334 | 42.997 | 58.1585 | 61.7704 | 52.43 |
89
+ | 0.1662 | 7.0 | 700 | 0.2631 | 68.6188 | 48.3329 | 63.9474 | 66.6006 | 46.94 |
90
+ | 0.161 | 8.0 | 800 | 0.2749 | 69.0046 | 48.1421 | 63.7844 | 66.8317 | 49.74 |
91
+ | 0.1207 | 9.0 | 900 | 0.3117 | 70.0357 | 48.9002 | 64.416 | 67.7582 | 48.66 |
92
+ | 0.0909 | 10.0 | 1000 | 0.3408 | 65.9578 | 45.2324 | 60.2838 | 63.7493 | 46.92 |
93
+ | 0.0749 | 11.0 | 1100 | 0.3516 | 67.4244 | 46.1985 | 61.6408 | 65.5371 | 46.61 |
94
+ | 0.0665 | 12.0 | 1200 | 0.3730 | 68.6911 | 47.7089 | 63.0381 | 66.6956 | 47.89 |
95
+ | 0.0522 | 13.0 | 1300 | 0.3891 | 67.2365 | 45.4165 | 61.4063 | 64.857 | 48.91 |
96
+ | 0.0355 | 14.0 | 1400 | 0.4128 | 69.1494 | 47.9278 | 63.3334 | 66.5969 | 50.55 |
97
+ | 0.0309 | 15.0 | 1500 | 0.4221 | 66.2447 | 44.937 | 60.1403 | 63.8541 | 50.71 |
98
+ | 0.0265 | 16.0 | 1600 | 0.4343 | 67.8178 | 46.7084 | 61.8173 | 65.4375 | 50.85 |
99
+ | 0.0158 | 17.0 | 1700 | 0.4577 | 67.9846 | 45.9562 | 61.6353 | 65.7207 | 50.81 |
100
+ | 0.0166 | 18.0 | 1800 | 0.4731 | 69.0971 | 47.7001 | 62.856 | 66.7796 | 50.01 |
101
+ | 0.0121 | 19.0 | 1900 | 0.4657 | 68.1397 | 46.4258 | 62.2696 | 65.9332 | 49.15 |
102
+ | 0.0095 | 20.0 | 2000 | 0.4793 | 68.6497 | 47.9446 | 63.0466 | 66.5409 | 50.96 |
103
+ | 0.0086 | 21.0 | 2100 | 0.4780 | 68.4363 | 46.7296 | 62.359 | 66.2626 | 50.02 |
104
+ | 0.0068 | 22.0 | 2200 | 0.4863 | 67.5415 | 46.0821 | 61.57 | 65.4613 | 49.5 |
105
+ | 0.0061 | 23.0 | 2300 | 0.4892 | 68.1283 | 46.5802 | 62.0832 | 66.0203 | 50.21 |
106
+ | 0.006 | 24.0 | 2400 | 0.4912 | 68.1723 | 46.3239 | 62.2007 | 65.6725 | 49.89 |
107
+ | 0.0057 | 25.0 | 2500 | 0.4930 | 68.3498 | 46.7534 | 62.3763 | 65.9575 | 49.82 |
108
 
109
 
110
  ### Framework versions
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9e93652a34d58078990f9ba2846811b986b9d9e1704f170e4b9df255807b0319
3
  size 956935738
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:567092da7af3b1cfc3086895c23b2acd9fe8bb99dea646d9dd240231840b5886
3
  size 956935738
training_args.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2ff5e250ca1d776218baf4d1a255bdc396495e4c8da8b2c2a5057ab9b2372c48
3
  size 4600
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:048064a76f6a8a675a0f9757d322b6138673afe81dae3d3a8746185b5391d1e6
3
  size 4600