|
|
--- |
|
|
license: mit |
|
|
base_model: gpt2 |
|
|
tags: |
|
|
- generated_from_trainer |
|
|
model-index: |
|
|
- name: fine_tuned_gpt2_clm-model |
|
|
results: [] |
|
|
datasets: |
|
|
- eli5 |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- perplexity |
|
|
pipeline_tag: text-generation |
|
|
--- |
|
|
|
|
|
<!-- This model card has been generated automatically according to the information the Trainer had access to. You |
|
|
should probably proofread and complete it, then remove this comment. --> |
|
|
|
|
|
# fine_tuned_gpt2_clm-model |
|
|
|
|
|
This model is a fine-tuned version of [gpt2](https://huggingface.co/gpt2) on the eli5 dataset. |
|
|
It achieves the following results on the evaluation set: |
|
|
- Loss: 3.3066 |
|
|
|
|
|
|
|
|
### Training hyperparameters |
|
|
|
|
|
The following hyperparameters were used during training: |
|
|
- learning_rate: 2e-05 |
|
|
- train_batch_size: 16 |
|
|
- eval_batch_size: 16 |
|
|
- seed: 42 |
|
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
|
- lr_scheduler_type: linear |
|
|
- num_epochs: 15 |
|
|
|
|
|
### Training results |
|
|
|
|
|
| Training Loss | Epoch | Step | Validation Loss | |
|
|
|:-------------:|:-----:|:----:|:---------------:| |
|
|
| No log | 1.0 | 142 | 3.3422 | |
|
|
| No log | 2.0 | 284 | 3.3226 | |
|
|
| No log | 3.0 | 426 | 3.3148 | |
|
|
| 3.4352 | 4.0 | 568 | 3.3095 | |
|
|
| 3.4352 | 5.0 | 710 | 3.3074 | |
|
|
| 3.4352 | 6.0 | 852 | 3.3066 | |
|
|
| 3.4352 | 7.0 | 994 | 3.3046 | |
|
|
| 3.3068 | 8.0 | 1136 | 3.3049 | |
|
|
| 3.3068 | 9.0 | 1278 | 3.3048 | |
|
|
| 3.3068 | 10.0 | 1420 | 3.3050 | |
|
|
| 3.2433 | 11.0 | 1562 | 3.3062 | |
|
|
| 3.2433 | 12.0 | 1704 | 3.3059 | |
|
|
| 3.2433 | 13.0 | 1846 | 3.3062 | |
|
|
| 3.2433 | 14.0 | 1988 | 3.3065 | |
|
|
| 3.2113 | 15.0 | 2130 | 3.3066 | |
|
|
|
|
|
### Inference: |
|
|
|
|
|
- prompt = "dna phosphorylation is the process of" |
|
|
|
|
|
- generated Text: dna phosphorylation is the process of forming the deoxygenated product. For example, in a protein phosphorylation inhibitor, it occurs to deoxygenate the phosphorylated protein by binding a phosphate molecule and preventing it from being destroyed by a nonenzymatic process. |
|
|
|
|
|
In a phosphorylation inhibitor like dna, the product is phosphorylated by the phosphocreatine, a phosphorylated phosphocreatine molecule that can bind to other phosphocreatine molecules that bind to phosphocreatine. This interaction helps to separate the phosphocreatine molecule that is phosphorylated from the phosphocreatine-phosphocreatine-phosphocreatine-phosphocreatine-glucose molecule that is phosphocreatine-phosphocreatine-glucose-phosphocreatine-phosphocreatine-glucose. |
|
|
|
|
|
In anoxidase inhibitors like dna, they are a bit more specific, more specific, and have a more complicated interaction with the phosphocreatine molecule that can bind to phosphocreatine molecules. |
|
|
|
|
|
I would argue that both dna-and phosphocreatine-phosphocreatine-glucose will not be able to bind to phosphocreatine because the phosphocreatine-phosphocreatine-phosphocreatine-glucose-phosphocreatine molecule that was phosphocreatine-phosphocreatine-phosphocreatine-phosphocreatine-phosphocreatine-phosphocreatine-phosphocreatine-glucose, is phosphocreatine. |
|
|
|
|
|
That is, dna-and phosphocreatine-glucose will be able to bind to phosphocreatine because the phosphocreatine molecule that was phosphocreatine-glucose will not be phosphocreatine because the phosphocreatine-phosphocreatine-glucose molecule that was phosphocreatine-phosphocreatine-phosphocreatine-phosphocreatine-glucose, is phosphocreatine. |
|
|
|
|
|
Edit: Added: The final point is that it can't bind phosphocreatine because that phosphocreatine molecule (a phosphocreatine-phosphocreatine-phosphocreatine-phosphocreatine molecule) can not be phosphoc |
|
|
|
|
|
### Evaluation metric: |
|
|
|
|
|
Perplexity: 27.29 |
|
|
|
|
|
### GPU: |
|
|
|
|
|
- CUDA Version: 12.1 |
|
|
- 4x Tesla T4 |
|
|
|
|
|
### Framework versions |
|
|
|
|
|
- Transformers 4.34.0 |
|
|
- Pytorch 2.1.0+cu121 |
|
|
- Datasets 2.14.5 |
|
|
- Tokenizers 0.14.1 |
