Hugging Face's logo

Multi-Domain-Expert-Learning
/
expert-min-pile-instruct

Text Generation
Transformers
PyTorch
gpt_neox
Generated from Trainer
text-generation-inference
Model card Files
xet
Community
2
P1ayer-1 commited on
Commit
fc749ce
·
1 Parent(s): 51aefc4

update model card README.md

Browse files
Files changed (1) hide show
  1. README.md +41 -34
README.md CHANGED
@@ -1,35 +1,22 @@
1
  ---
2
  tags:
3
  - generated_from_trainer
4
- datasets:
5
- - pile-instruct/
6
  metrics:
7
  - accuracy
8
  model-index:
9
- - name: layer_4,5,6,7,8
10
- results:
11
- - task:
12
- type: text-generation
13
- name: Causal Language Modeling
14
- dataset:
15
- name: pile-instruct/
16
- type: pile-instruct/
17
- split: None
18
- metrics:
19
- - type: accuracy
20
- value: 0.20994595912408442
21
- name: Accuracy
22
  ---
23
 
24
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
25
  should probably proofread and complete it, then remove this comment. -->
26
 
27
- # layer_4,5,6,7,8
28
 
29
- This model is a fine-tuned version of [P1ayer-1/pythia-deduped-1b-chat-base](https://huggingface.co/P1ayer-1/pythia-deduped-1b-chat-base) on the pile-instruct/ dataset.
30
  It achieves the following results on the evaluation set:
31
- - Loss: 6.9437
32
- - Accuracy: 0.2099
33
 
34
  ## Model description
35
 
@@ -49,27 +36,51 @@ More information needed
49
 
50
  The following hyperparameters were used during training:
51
  - learning_rate: 0.0001
52
- - train_batch_size: 1
53
  - eval_batch_size: 8
54
  - seed: 42
55
  - distributed_type: multi-GPU
56
  - num_devices: 8
57
- - gradient_accumulation_steps: 8
58
- - total_train_batch_size: 64
59
  - total_eval_batch_size: 64
60
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
61
  - lr_scheduler_type: linear
62
- - training_steps: 1000
63
 
64
  ### Training results
65
 
66
- | Training Loss | Epoch | Step | Validation Loss | Accuracy |
67
- |:-------------:|:-----:|:----:|:---------------:|:--------:|
68
- | 7.6017 | 0.02 | 200 | 7.5928 | 0.1605 |
69
- | 7.1871 | 0.03 | 400 | 7.2690 | 0.1847 |
70
- | 7.0356 | 0.05 | 600 | 7.0897 | 0.1980 |
71
- | 6.93 | 0.07 | 800 | 6.9870 | 0.2064 |
72
- | 6.9089 | 0.08 | 1000 | 6.9437 | 0.2099 |
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
73
 
74
 
75
  ### Framework versions
@@ -78,7 +89,3 @@ The following hyperparameters were used during training:
78
  - Pytorch 2.0.0+cu117
79
  - Datasets 2.11.0
80
  - Tokenizers 0.13.3
81
-
82
-
83
- ## Wandb Report
84
- https://wandb.ai/ontocord/pythia-1b-deduped-layer-test-min-pile-instruct/runs/6hvfd11h
 
1
  ---
2
  tags:
3
  - generated_from_trainer
 
 
4
  metrics:
5
  - accuracy
6
  model-index:
7
+ - name: expert-min-pile-instruct
8
+ results: []
 
 
 
 
 
 
 
 
 
 
 
9
  ---
10
 
11
  <!-- This model card has been generated automatically according to the information the Trainer had access to. You
12
  should probably proofread and complete it, then remove this comment. -->
13
 
14
+ # expert-min-pile-instruct
15
 
16
+ This model is a fine-tuned version of [P1ayer-1/pythia-deduped-1b-chat-base](https://huggingface.co/P1ayer-1/pythia-deduped-1b-chat-base) on the None dataset.
17
  It achieves the following results on the evaluation set:
18
+ - Loss: 4.9648
19
+ - Accuracy: 0.3842
20
 
21
  ## Model description
22
 
 
36
 
37
  The following hyperparameters were used during training:
38
  - learning_rate: 0.0001
39
+ - train_batch_size: 12
40
  - eval_batch_size: 8
41
  - seed: 42
42
  - distributed_type: multi-GPU
43
  - num_devices: 8
44
+ - total_train_batch_size: 96
 
45
  - total_eval_batch_size: 64
46
  - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
47
  - lr_scheduler_type: linear
48
+ - training_steps: 6000
49
 
50
  ### Training results
51
 
52
+ | Training Loss | Epoch | Step | Accuracy | Validation Loss |
53
+ |:-------------:|:-----:|:----:|:--------:|:---------------:|
54
+ | 7.4574 | 0.1 | 200 | 0.1688 | 7.4961 |
55
+ | 7.0445 | 0.2 | 400 | 0.1997 | 7.0547 |
56
+ | 6.7483 | 0.3 | 600 | 0.2190 | 6.7930 |
57
+ | 6.4568 | 0.4 | 800 | 0.2376 | 6.5703 |
58
+ | 6.2865 | 0.5 | 1000 | 0.2552 | 6.375 |
59
+ | 6.1028 | 0.6 | 1200 | 0.2793 | 6.1484 |
60
+ | 5.8888 | 0.7 | 1400 | 0.2982 | 5.9570 |
61
+ | 5.7362 | 0.8 | 1600 | 0.3121 | 5.8008 |
62
+ | 5.6507 | 0.9 | 1800 | 0.3238 | 5.6797 |
63
+ | 5.565 | 1.0 | 2000 | 0.3318 | 5.5781 |
64
+ | 5.4688 | 1.1 | 2200 | 0.3392 | 5.4961 |
65
+ | 5.4044 | 1.2 | 2400 | 0.3456 | 5.4219 |
66
+ | 5.3323 | 1.3 | 2600 | 0.3516 | 5.3594 |
67
+ | 5.2598 | 1.4 | 2800 | 0.3562 | 5.3047 |
68
+ | 5.2159 | 1.5 | 3000 | 0.3596 | 5.2578 |
69
+ | 5.1992 | 1.6 | 3200 | 0.3638 | 5.2148 |
70
+ | 5.1429 | 1.69 | 3400 | 0.3672 | 5.1797 |
71
+ | 5.095 | 1.79 | 3600 | 0.3696 | 5.1445 |
72
+ | 5.0646 | 1.89 | 3800 | 0.3715 | 5.1172 |
73
+ | 5.059 | 1.99 | 4000 | 0.3742 | 5.0859 |
74
+ | 5.0152 | 2.09 | 4200 | 0.3756 | 5.0664 |
75
+ | 5.0251 | 2.19 | 4400 | 0.3769 | 5.0469 |
76
+ | 5.022 | 2.29 | 4600 | 0.3790 | 5.0273 |
77
+ | 4.9939 | 2.39 | 4800 | 0.3798 | 5.0156 |
78
+ | 4.924 | 2.49 | 5000 | 0.3811 | 5.0 |
79
+ | 4.9335 | 2.59 | 5200 | 0.3821 | 4.9883 |
80
+ | 4.9231 | 2.69 | 5400 | 0.3829 | 4.9805 |
81
+ | 4.8886 | 2.79 | 5600 | 4.9727 | 0.3835 |
82
+ | 4.9419 | 2.89 | 5800 | 4.9648 | 0.3843 |
83
+ | 4.9227 | 2.99 | 6000 | 4.9648 | 0.3842 |
84
 
85
 
86
  ### Framework versions
 
89
  - Pytorch 2.0.0+cu117
90
  - Datasets 2.11.0
91
  - Tokenizers 0.13.3