Hugging Face's logo

DLingo
/
qwen2-2b-instruct-trl-sft-mrg

PEFT
Safetensors
qwen2_vl
trl
sft
Generated from Trainer
Model card Files
xet
Community
DLingo commited on
Commit
3d8dbf5
·
verified ·
1 Parent(s): eb93b4a

Model save

Browse files
Files changed (2) hide show
  1. README.md +23 -23
  2. adapter_model.safetensors +1 -1
README.md CHANGED
@@ -18,7 +18,7 @@ should probably proofread and complete it, then remove this comment. -->
18
 
19
  This model is a fine-tuned version of [Qwen/Qwen2-VL-2B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-2B-Instruct) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
- - Loss: 0.6032
22
 
23
  ## Model description
24
 
@@ -38,10 +38,10 @@ More information needed
38
 
39
  The following hyperparameters were used during training:
40
  - learning_rate: 0.0001
41
- - train_batch_size: 2
42
- - eval_batch_size: 2
43
  - seed: 42
44
- - gradient_accumulation_steps: 16
45
  - total_train_batch_size: 32
46
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
  - lr_scheduler_type: constant
@@ -52,25 +52,25 @@ The following hyperparameters were used during training:
52
 
53
  | Training Loss | Epoch | Step | Validation Loss |
54
  |:-------------:|:-------:|:----:|:---------------:|
55
- | 2.5999 | 0.7729 | 50 | 1.5728 |
56
- | 1.2619 | 1.5459 | 100 | 1.0742 |
57
- | 0.9938 | 2.3188 | 150 | 0.8737 |
58
- | 0.8441 | 3.0918 | 200 | 0.7896 |
59
- | 0.7839 | 3.8647 | 250 | 0.7490 |
60
- | 0.7418 | 4.6377 | 300 | 0.7199 |
61
- | 0.7067 | 5.4106 | 350 | 0.6964 |
62
- | 0.6922 | 6.1836 | 400 | 0.6798 |
63
- | 0.6541 | 6.9565 | 450 | 0.6670 |
64
- | 0.6602 | 7.7295 | 500 | 0.6553 |
65
- | 0.6152 | 8.5024 | 550 | 0.6435 |
66
- | 0.6274 | 9.2754 | 600 | 0.6379 |
67
- | 0.6154 | 10.0483 | 650 | 0.6297 |
68
- | 0.6104 | 10.8213 | 700 | 0.6244 |
69
- | 0.5662 | 11.5942 | 750 | 0.6147 |
70
- | 0.5905 | 12.3671 | 800 | 0.6137 |
71
- | 0.5673 | 13.1401 | 850 | 0.6109 |
72
- | 0.5703 | 13.9130 | 900 | 0.6005 |
73
- | 0.557 | 14.6860 | 950 | 0.6032 |
74
 
75
 
76
  ### Framework versions
 
18
 
19
  This model is a fine-tuned version of [Qwen/Qwen2-VL-2B-Instruct](https://huggingface.co/Qwen/Qwen2-VL-2B-Instruct) on an unknown dataset.
20
  It achieves the following results on the evaluation set:
21
+ - Loss: 1.2108
22
 
23
  ## Model description
24
 
 
38
 
39
  The following hyperparameters were used during training:
40
  - learning_rate: 0.0001
41
+ - train_batch_size: 4
42
+ - eval_batch_size: 4
43
  - seed: 42
44
+ - gradient_accumulation_steps: 8
45
  - total_train_batch_size: 32
46
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
  - lr_scheduler_type: constant
 
52
 
53
  | Training Loss | Epoch | Step | Validation Loss |
54
  |:-------------:|:-------:|:----:|:---------------:|
55
+ | 2.1076 | 0.7722 | 50 | 2.0130 |
56
+ | 1.7006 | 1.5444 | 100 | 1.6928 |
57
+ | 1.5932 | 2.3166 | 150 | 1.5687 |
58
+ | 1.5092 | 3.0888 | 200 | 1.4995 |
59
+ | 1.4633 | 3.8610 | 250 | 1.4468 |
60
+ | 1.3849 | 4.6332 | 300 | 1.4023 |
61
+ | 1.3616 | 5.4054 | 350 | 1.3673 |
62
+ | 1.361 | 6.1776 | 400 | 1.3386 |
63
+ | 1.3253 | 6.9498 | 450 | 1.3159 |
64
+ | 1.3204 | 7.7220 | 500 | 1.2976 |
65
+ | 1.1944 | 8.4942 | 550 | 1.2814 |
66
+ | 1.2286 | 9.2664 | 600 | 1.2703 |
67
+ | 1.3097 | 10.0386 | 650 | 1.2532 |
68
+ | 1.263 | 10.8108 | 700 | 1.2466 |
69
+ | 1.1474 | 11.5830 | 750 | 1.2374 |
70
+ | 1.191 | 12.3552 | 800 | 1.2298 |
71
+ | 1.09 | 13.1274 | 850 | 1.2246 |
72
+ | 1.1622 | 13.8996 | 900 | 1.2130 |
73
+ | 1.1883 | 14.6718 | 950 | 1.2108 |
74
 
75
 
76
  ### Framework versions
adapter_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:460df92e38b2b6cd3f0df685b95dbeba171a6fca99c36b3acaa46fdd386e0255
3
  size 4373064
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d8782e5d43ef2cf59856effcb5935892ccf2a1f568969c95663b76633a42d05
3
  size 4373064