File size: 2,815 Bytes
c7efe3e
02aed86
 
 
 
5821ea3
02aed86
 
 
5821ea3
 
 
 
 
 
c7efe3e
02aed86
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
479d9bb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
02aed86
 
 
 
 
 
e83efd0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
---
license: apache-2.0
base_model: distilgpt2
tags:
- generated_from_trainer
- gpt
model-index:
- name: gpt2_dolly_lite
  results: []
datasets:
- tatsu-lab/alpaca
language:
- en
metrics:
- accuracy
---

<!-- 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. -->

# gpt2_dolly_lite

This model is a fine-tuned version of [distilgpt2](https://huggingface.co/distilgpt2) on an unknown dataset.
It achieves the following results on the evaluation set:
- Loss: 2.4067

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 0.001
- train_batch_size: 8
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 3

### Training results

| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 2.708         | 1.0   | 1300 | 2.5611          |
| 2.1768        | 2.0   | 2600 | 2.4149          |
| 1.7189        | 3.0   | 3900 | 2.4067          |

### USAGE
```
MODEL = 'distilgpt2'

tokenizer = AutoTokenizer.from_pretrained(MODEL)

tokenizer.pad_token = tokenizer.eos_token

def respond(instruction, generator, _input=None, verbose=False, **options):
    if not _input:
        prompt = f'Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n### Instruction:\n{instruction}\n\n### Response:\n'
    else:
        prompt = f'Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.\n\n### Instruction:\n{instruction}\n\n### Input: {_input}\n\n### Response:\n'
    if verbose:
        print(prompt)
    generated_texts = generator(
        prompt,
        num_return_sequences=3,
        temperature=options.get('temperature', 0.7),
        max_new_tokens=options.get('max_new_tokens', 128)
    )
    for generated_text in generated_texts:
        print(generated_text['generated_text'].split('### Response:\n')[1])
        print('----')

loaded_model = AutoModelForCausalLM.from_pretrained('Andyrasika/gpt2_dolly_lite')

dolly_lite = pipeline('text-generation', model=loaded_model, tokenizer=tokenizer)

respond(
    'Write me an email to my boss, telling her I quit because I made a cool LLM.', dolly_lite
)
```

### Framework versions

- Transformers 4.32.1
- Pytorch 2.0.1+cu118
- Datasets 2.14.4
- Tokenizers 0.13.3