File size: 2,637 Bytes
c56018a
 
 
 
 
 
 
 
 
 
 
 
 
 
c4b5010
c56018a
871fcf9
d31f196
c634156
 
 
d31f196
b80485a
 
 
c56018a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: other
license_name: tongyi-qianwen
license_link: https://huggingface.co/Qwen/Qwen2-72B-Instruct/blob/main/LICENSE
language:
- en
- zh
pipeline_tag: text-generation
tags:
- chat
---

# Roleplay Quantization in EXL2 format for Magnum v1

Quantized using the [cleaned PIPPA](https://huggingface.co/datasets/royallab/PIPPA-cleaned) roleplay dataset. Uploading as I didn't see anyone else do this one yet.

[4.0bpw8h quants](https://huggingface.co/luigi86/magnum-72b-v1-exl2-rpcal/tree/4.0bpw8h) (tested and working on two 3090s with Q4 cache at 32k context)

[8.0bpw8h quants](https://huggingface.co/luigi86/magnum-72b-v1-exl2-rpcal/tree/8.0bpw8h)



See [original model](https://huggingface.co/alpindale/magnum-72b-v1) for further details.


# Original Model card

![](https://files.catbox.moe/ngqnb1.png)

This is the first in a series of models designed to replicate the prose quality of the Claude 3 models, specifically Sonnet and Opus. This model is fine-tuned on top of [Qwen-2 72B Instruct](https://huggingface.co/Qwen/Qwen2-72B-Instruct).


## Prompting
Model has been Instruct tuned with the ChatML formatting. A typical input would look like this:

```py
"""<|im_start|>user
Hi there!<|im_end|>
<|im_start|>assistant
Nice to meet you!<|im_end|>
<|im_start|>user
Can I ask a question?<|im_end|>
<|im_start|>assistant
"""
```

## Credits

This model has been a team effort, credits go to:

- [Sao10K](https://huggingface.co/Sao10K) for help with (and cleaning up!) the dataset.
- [alpindale](https://huggingface.co/alpindale) for the training.
- [kalomaze](https://huggingface.co/kalomaze) for helping with the hyperparameter tuning.
- Various other people for their continued help as we tuned the parameters, restarted failed runs. In no particular order: [Doctor Shotgun](https://huggingface.co/Doctor-Shotgun), [Lucy](https://huggingface.co/lucyknada), [Nopm](https://huggingface.co/nopm), [Mango](https://huggingface.co/MangoMango69420), and the rest of the Silly Tilly.

And last but not least, we'd like to thank [Kearm](https://twitter.com/Nottlespike) for sponsoring the compute needed to train this model.

## Training
The training was done with 55 million tokens of high-quality RP data, over 1.5 epochs. We used 8x [AMD Instinct™ MI300X Accelerators](https://www.amd.com/en/products/accelerators/instinct/mi300/mi300x.html) for the full-parameter fine-tuning of the model.

[<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/OpenAccess-AI-Collective/axolotl)

## Safety
...