Hugging Face's logo

statking
/
paligemma-vqa

PEFT
TensorBoard
Safetensors
paligemma
Generated from Trainer
Model card Files Metrics Community
statking commited on
Commit
620f719
·
verified ·
1 Parent(s): 3188382

Model save

Browse files
Files changed (3) hide show
  1. README.md +16 -22
  2. adapter_model.safetensors +1 -1
  3. runs/May28_07-41-31_ae63705f58eb/events.out.tfevents.1716882099.ae63705f58eb.96253.0 +2 -2
README.md CHANGED
@@ -1,8 +1,9 @@
1
  ---
2
  license: gemma
3
- base_model: google/paligemma-3b-pt-224
4
  tags:
5
  - generated_from_trainer
 
6
  datasets:
7
  - vq_av2
8
  model-index:
@@ -13,12 +14,11 @@ model-index:
13
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
14
  should probably proofread and complete it, then remove this comment. -->
15
 
16
- [<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="200" height="32"/>](https://wandb.ai/statking/huggingface/runs/xgb0dent)
17
  # paligemma-vqa
18
 
19
  This model is a fine-tuned version of [google/paligemma-3b-pt-224](https://huggingface.co/google/paligemma-3b-pt-224) on the vq_av2 dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 0.0001
22
 
23
  ## Model description
24
 
@@ -37,39 +37,33 @@ More information needed
37
  ### Training hyperparameters
38
 
39
  The following hyperparameters were used during training:
40
- - learning_rate: 0.02
41
  - train_batch_size: 16
42
  - eval_batch_size: 16
43
  - seed: 42
 
 
44
  - gradient_accumulation_steps: 4
45
- - total_train_batch_size: 64
 
46
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
  - lr_scheduler_type: linear
48
- - lr_scheduler_warmup_steps: 1200
49
- - num_epochs: 1
50
 
51
  ### Training results
52
 
53
  | Training Loss | Epoch | Step | Validation Loss |
54
  |:-------------:|:------:|:----:|:---------------:|
55
- | 0.0019 | 0.0736 | 500 | 0.0081 |
56
- | 0.0004 | 0.1472 | 1000 | 0.0002 |
57
- | 0.0003 | 0.2207 | 1500 | 0.0002 |
58
- | 0.0001 | 0.2943 | 2000 | 0.0001 |
59
- | 0.0001 | 0.3679 | 2500 | 0.0001 |
60
- | 0.0001 | 0.4415 | 3000 | 0.0001 |
61
- | 0.0002 | 0.5151 | 3500 | 0.0002 |
62
- | 0.0001 | 0.5886 | 4000 | 0.0001 |
63
- | 0.0001 | 0.6622 | 4500 | 0.0001 |
64
- | 0.0001 | 0.7358 | 5000 | 0.0001 |
65
- | 0.0001 | 0.8094 | 5500 | 0.0001 |
66
- | 0.0001 | 0.8830 | 6000 | 0.0001 |
67
- | 0.0001 | 0.9566 | 6500 | 0.0001 |
68
 
69
 
70
  ### Framework versions
71
 
72
- - Transformers 4.41.0
 
73
  - Pytorch 2.2.0+cu121
74
  - Datasets 2.19.1
75
- - Tokenizers 0.19.1
 
1
  ---
2
  license: gemma
3
+ library_name: peft
4
  tags:
5
  - generated_from_trainer
6
+ base_model: google/paligemma-3b-pt-224
7
  datasets:
8
  - vq_av2
9
  model-index:
 
14
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
15
  should probably proofread and complete it, then remove this comment. -->
16
 
 
17
  # paligemma-vqa
18
 
19
  This model is a fine-tuned version of [google/paligemma-3b-pt-224](https://huggingface.co/google/paligemma-3b-pt-224) on the vq_av2 dataset.
20
  It achieves the following results on the evaluation set:
21
+ - Loss: 0.5071
22
 
23
  ## Model description
24
 
 
37
  ### Training hyperparameters
38
 
39
  The following hyperparameters were used during training:
40
+ - learning_rate: 0.0001
41
  - train_batch_size: 16
42
  - eval_batch_size: 16
43
  - seed: 42
44
+ - distributed_type: multi-GPU
45
+ - num_devices: 4
46
  - gradient_accumulation_steps: 4
47
+ - total_train_batch_size: 256
48
+ - total_eval_batch_size: 64
49
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
50
  - lr_scheduler_type: linear
51
+ - lr_scheduler_warmup_steps: 50
52
+ - num_epochs: 2
53
 
54
  ### Training results
55
 
56
  | Training Loss | Epoch | Step | Validation Loss |
57
  |:-------------:|:------:|:----:|:---------------:|
58
+ | 0.5618 | 0.5886 | 1000 | 0.5531 |
59
+ | 0.5268 | 1.1772 | 2000 | 0.5335 |
60
+ | 0.5099 | 1.7657 | 3000 | 0.5071 |
 
 
 
 
 
 
 
 
 
 
61
 
62
 
63
  ### Framework versions
64
 
65
+ - PEFT 0.11.1
66
+ - Transformers 4.41.1
67
  - Pytorch 2.2.0+cu121
68
  - Datasets 2.19.1
69
+ - Tokenizers 0.19.1
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e477bed6630ea7d3f1d5af070b0c2d7b2440fac4e0fb1bd9907d5e9caf337034
3
  size 45258384
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36baffd20b87b8b1431cfe8c35045dd83c082f3d404894a0dd11bf78dc0ce038
3
  size 45258384
runs/May28_07-41-31_ae63705f58eb/events.out.tfevents.1716882099.ae63705f58eb.96253.0 CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:a052271f8f2ea7195a5cb56bd1ad39aa1429c59c6c6d034cd18c242044bfce94
3
- size 12758
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d891a13fe1e0e0b6927a7fcbf3ba26f10c4c09fa53d46a77c0d962c9c485e215
3
+ size 13745