File size: 1,231 Bytes
dc1633f
 
d5e2d4d
 
dc1633f
 
 
 
 
fbe9fc8
dc1633f
 
 
 
 
 
 
fbe9fc8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
dc1633f
 
 
 
 
 
 
 
 
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
---
license: gpl
language:
- en
---
This is the GPTQ 4Bit Groupsize 128 Pre-Quantized Model, for the full model in fp32, visit https://huggingface.co/NousResearch/gpt4-x-vicuna-13b

As a base model used https://huggingface.co/eachadea/vicuna-13b-1.1

Finetuned on Teknium's GPTeacher dataset, Teknium's unreleased Roleplay v2 dataset, WizardLM Uncensored, GPT-4-LLM Uncensored, and Nous Research Instruct Dataset

Approx 180k instructions, all from GPT-4, all cleaned of any OpenAI censorship/"As an AI Language Model" etc.

Base model still has OpenAI censorship. Soon, a new version will be released with cleaned vicuna from https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltere

Trained on 8 A100-80GB GPUs for 5 epochs following Alpaca deepspeed training code.

Prompt format is Alpaca:  
```
### Instruction:

### Response:
```  

or 

```
### Instruction:

### Input:

### Response:

```  


Nous Research Instruct Dataset will be released soon.

GPTeacher, Roleplay v2 by https://huggingface.co/teknium

Wizard LM by https://github.com/nlpxucan

Nous Research Instruct Dataset by https://huggingface.co/karan4d and https://huggingface.co/huemin

Compute provided by our project sponsor https://redmond.ai/