File size: 3,529 Bytes
7c14f0c
da16538
 
5ff3463
da16538
5ff3463
 
64c2314
 
 
 
e238be0
64c2314
da16538
 
 
 
5ff3463
 
 
 
 
 
 
64c2314
5ff3463
 
 
 
 
64c2314
 
 
5ff3463
 
 
64c2314
5ff3463
 
 
64c2314
5ff3463
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
e238be0
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
---
language:
- code
license: openrail
library_name: transformers
tags:
- generated_from_trainer
- code
- codegen
- assembly
datasets:
- bigcode/the-stack-dedup
pipeline_tag: text-generation
base_model: bigcode/santacoder
model-index:
- name: santacoder-finetuned-the-stack-cobol
  results: []
---

<!-- 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. -->

# santacoder-finetuned-the-stack-cobol

This model is a fine-tuned version of [bigcode/santacoder](https://huggingface.co/bigcode/santacoder) on an The Stack [cobol](https://huggingface.co/datasets/bigcode/the-stack-dedup) dataset.
It achieves the following results on the evaluation set:
- Loss: 0.7161

## Model description

The [SantaCoder](https://huggingface.co/bigcode/santacoder) models are a series of 1.1B parameter models trained on the Python, Java, and JavaScript subset of [The Stack (v1.1)](https://huggingface.co/datasets/bigcode/the-stack) (which excluded opt-out requests). 
The main model uses [Multi Query Attention](https://arxiv.org/abs/1911.02150), was trained using near-deduplication and comment-to-code ratio as filtering criteria and using the [Fill-in-the-Middle objective](https://arxiv.org/abs/2207.14255).
In addition, there are several models that were trained on datasets with different filter parameters and with architecture and objective variations. 

## Intended uses & limitations

The predominant language in source is English although other languages are also present. As such the model is capable to generate code snippets provided some context but the generated code is not guaranteed to work as intended. It can be inefficient, contain bugs or exploits.

## Training and evaluation data

The Stack contains over 6TB of permissively-licensed source code files covering 358 programming languages. The dataset was created as part of the [BigCode Project](https://www.bigcode-project.org/), an open scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs). The Stack serves as a pre-training dataset for Code LLMs, i.e., code-generating AI systems which enable the synthesis of programs from natural language descriptions as well as other from code snippets. **This is the near-deduplicated version with 3TB data.**

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 8
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 64
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 100
- training_steps: 1000

### Training results

| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 1.3911        | 0.1   | 100  | 1.1141          |
| 0.9478        | 0.2   | 200  | 0.9735          |
| 0.784         | 0.3   | 300  | 0.8497          |
| 0.4702        | 0.4   | 400  | 0.7686          |
| 0.6133        | 0.5   | 500  | 0.7375          |
| 0.5396        | 0.6   | 600  | 0.7265          |
| 0.3937        | 0.7   | 700  | 0.6952          |
| 0.5691        | 0.8   | 800  | 0.7059          |
| 0.6366        | 0.9   | 900  | 0.7069          |
| 0.3661        | 1.0   | 1000 | 0.7161          |


### Framework versions

- Transformers 4.26.0
- Pytorch 1.13.1+cu116
- Datasets 2.9.0
- Tokenizers 0.13.2