File size: 3,379 Bytes
30d7874
5d1ac2d
 
5c4ece7
8628592
 
5c4ece7
5d1ac2d
8628592
5c4ece7
8628592
30d7874
8628592
 
5c4ece7
8628592
 
 
5c4ece7
 
 
8628592
 
 
5c4ece7
 
 
8628592
 
 
5c4ece7
 
8628592
 
 
5c4ece7
 
8628592
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
language:
- nl
license: cc-by-nc-4.0
datasets:
- BramVanroy/alpaca-dolly-dutch
inference: false
base_model: ybelkada/falcon-7b-sharded-bf16
model-index:
- name: falcon-7b-ft-alpaca-cleaned-dutch
  results: []
---


# falcon-7b-ft-alpaca-dolly-dutch

## Model description

This model is a fine-tuned version of [ybelkada/falcon-7b-sharded-bf16](https://huggingface.co/ybelkada/falcon-7b-sharded-bf16) on the [BramVanroy/alpaca-dolly-dutch](https://huggingface.co/datasets/BramVanroy/alpaca-dolly-dutch) dataset.
See the original [Falcon 7B model](https://huggingface.co/tiiuae/falcon-7b/) for more information, intended use, and biases.


## Intended uses & limitations

This model is intended as a (poor) baseline for Dutch generative LLMs. It by no means aims to provide SOTA performance and is specifically intended for research purposes, and an opportunity for me to test hyperparameters and stability.

Importantly, the original Falcon 7B model was only trained on English and French. Therefore, Dutch generations should be taken with a massive grain of salt.

## Training and evaluation data

Trained on the synthetic [BramVanroy/alpaca-dolly-dutch](https://huggingface.co/datasets/BramVanroy/alpaca-dolly-dutch) instruction dataset. 
Therefore, commercial use of this model is forbidden. The model is intended for research purposes only.

## Training procedure

Trained with LoRA and merged before upload. The adapters are in the `adapters` branch.

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 0.001
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- distributed_type: multi-GPU
- num_devices: 16
- total_train_batch_size: 512
- total_eval_batch_size: 512
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.03
- num_epochs: 20

### Training results

| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 1.8677        | 0.16  | 20   | 1.6766          |
| 1.5635        | 0.32  | 40   | 1.5643          |
| 1.6353        | 0.48  | 60   | 1.4980          |
| 1.5166        | 0.65  | 80   | 1.4516          |
| 1.4287        | 0.81  | 100  | 1.4096          |
| 1.5791        | 0.97  | 120  | 1.3802          |
| 1.3911        | 1.13  | 140  | 1.3633          |
| 1.356         | 1.29  | 160  | 1.3419          |
| 1.2524        | 1.45  | 180  | 1.3263          |
| 1.4224        | 1.61  | 200  | 1.3056          |
| 1.2266        | 1.77  | 220  | 1.2897          |
| 1.3242        | 1.94  | 240  | 1.2785          |
| 1.03          | 2.1   | 260  | 1.2957          |
| 1.1643        | 2.26  | 280  | 1.2970          |
| 1.1492        | 2.42  | 300  | 1.2779          |
| 1.0679        | 2.58  | 320  | 1.2770          |
| 1.2695        | 2.74  | 340  | 1.2658          |
| 1.0439        | 2.9   | 360  | 1.2612          |
| 0.9453        | 3.06  | 380  | 1.3157          |
| 0.8494        | 3.23  | 400  | 1.3189          |
| 1.0745        | 3.39  | 420  | 1.3073          |
| 0.8679        | 3.55  | 440  | 1.3019          |
| 1.0569        | 3.71  | 460  | 1.2955          |
| 1.0186        | 3.87  | 480  | 1.2890          |
| 0.8413        | 4.03  | 500  | 1.3445          |


### Framework versions

- Transformers 4.30.1
- Pytorch 2.0.1+cu117
- Datasets 2.13.1
- Tokenizers 0.13.3