File size: 2,890 Bytes
4bedb9d
 
 
 
8504816
 
4bedb9d
 
 
 
 
 
 
 
 
 
8504816
4bedb9d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: other
tags:
- generated_from_trainer
datasets:
- HiTZ/alpaca_mt
model-index:
- name: alpaca-lora-30b-en-pt-es-ca-eu-gl-at
  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. -->

# alpaca-lora-30b-en-pt-es-ca-eu-gl-at

This model is a fine-tuned version of [decapoda-research/llama-30b-hf](https://huggingface.co/decapoda-research/llama-30b-hf) on the HiTZ/alpaca_mt ['en', 'pt', 'es', 'ca', 'eu', 'gl', 'at'] dataset.
It achieves the following results on the evaluation set:
- Loss: 0.9088

## 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: 0.0003
- train_batch_size: 6
- eval_batch_size: 6
- seed: 42
- distributed_type: multi-GPU
- gradient_accumulation_steps: 21
- total_train_batch_size: 126
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.03
- num_epochs: 1
- mixed_precision_training: Native AMP

### Training results

| Training Loss | Epoch | Step | Validation Loss |
|:-------------:|:-----:|:----:|:---------------:|
| 1.1695        | 0.04  | 100  | 1.1716          |
| 1.1211        | 0.07  | 200  | 1.0964          |
| 1.0591        | 0.11  | 300  | 1.0590          |
| 1.0234        | 0.14  | 400  | 1.0341          |
| 1.0345        | 0.18  | 500  | 1.0165          |
| 0.9932        | 0.22  | 600  | 1.0024          |
| 0.9948        | 0.25  | 700  | 0.9895          |
| 1.01          | 0.29  | 800  | 0.9794          |
| 0.9488        | 0.32  | 900  | 0.9708          |
| 0.9518        | 0.36  | 1000 | 0.9627          |
| 0.9463        | 0.4   | 1100 | 0.9557          |
| 0.956         | 0.43  | 1200 | 0.9498          |
| 0.9521        | 0.47  | 1300 | 0.9437          |
| 0.9345        | 0.51  | 1400 | 0.9385          |
| 0.9469        | 0.54  | 1500 | 0.9337          |
| 0.9466        | 0.58  | 1600 | 0.9297          |
| 0.9403        | 0.61  | 1700 | 0.9257          |
| 0.9179        | 0.65  | 1800 | 0.9219          |
| 0.9468        | 0.69  | 1900 | 0.9190          |
| 0.9173        | 0.72  | 2000 | 0.9163          |
| 0.9172        | 0.76  | 2100 | 0.9142          |
| 0.9351        | 0.79  | 2200 | 0.9124          |
| 0.9238        | 0.83  | 2300 | 0.9110          |
| 0.9057        | 0.87  | 2400 | 0.9099          |
| 0.9309        | 0.9   | 2500 | 0.9093          |
| 0.8893        | 0.94  | 2600 | 0.9090          |
| 0.9095        | 0.97  | 2700 | 0.9088          |


### Framework versions

- Transformers 4.28.0.dev0
- Pytorch 2.0.0+cu117
- Datasets 2.10.1
- Tokenizers 0.13.2