File size: 1,666 Bytes
f4c0211
 
 
 
 
 
 
 
 
 
 
96d0ecb
f4c0211
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
language:
- en
license: apache-2.0
tags:
- text-generation-inference
- transformers
- unsloth
- mistral
- trl
- sft
base_model: mpasila/JP-EN-Translator-2K-steps-7B
datasets:
- NilanE/ParallelFiction-Ja_En-100k
- mpasila/ParallelFiction-Ja_En-100k-alpaca
---
This is an ExLlamaV2 quantized model in 4bpw of [mpasila/JP-EN-Translator-2K-steps-7B](https://huggingface.co/mpasila/JP-EN-Translator-2K-steps-7B) using the default calibration dataset.

# Original Model card

Experimental model, may not perform that well. Dataset used is [a modified](https://huggingface.co/datasets/mpasila/ParallelFiction-Ja_En-100k-alpaca) version of [NilanE/ParallelFiction-Ja_En-100k](https://huggingface.co/datasets/NilanE/ParallelFiction-Ja_En-100k).

After training with an 8k context length it didn't appear to improve performance much at all. Not sure if I should keep training it (which is costly) or if I should fix some issues with the dataset (like it starting with Ch or Chapter) or I go back to finetuning Finnish models.

### Prompt format: Alpaca
```
Below is a translation task, paired with an input that provides further context. Write a response that appropriately completes the request.

### Instruction:
{}

### Input:
{}

### Response:
{}
```

# Uploaded  model

- **Developed by:** mpasila
- **License:** apache-2.0
- **Finetuned from model :** augmxnt/shisa-base-7b-v1

This mistral model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.

[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)