---
library_name: peft
tags:
- generated_from_trainer
base_model: mistralai/Mixtral-8x7B-v0.1
model-index:
- name: limarp-zloss-mixtral-8x7b-qlora
  results: []
datasets:
- lemonilia/LimaRP
language:
- en
license: apache-2.0
---

[<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)
# limarp-zloss-mixtral-8x7b-qlora

Experimental limarp qlora trained at 10k ctx length (greater than size of the longest limarp sample when tokenized via mistral's tokenizer) on [mistralai/Mixtral-8x7B-v0.1](https://huggingface.co/mistralai/Mixtral-8x7B-v0.1) using [Charles Goddard](https://huggingface.co/chargoddard)'s ZLoss and Megablocks-based fork of transformers.

## Model description

The intended prompt format is the Alpaca instruction format of LimaRP v3:
```
### Instruction:
Character's Persona: {bot character description}

User's Persona: {user character description}

Scenario: {what happens in the story}

Play the role of Character. Taking the above information into consideration, you must engage in a roleplaying chat with User below this line. Do not write dialogues and narration for User.

### Input:
User: {utterance}

### Response:
Character: {utterance}

### Input:
User: {utterance}

### Response:
Character: {utterance}

(etc.)
```
Inspired by the previously named "Roleplay" preset in SillyTavern, with this version of LimaRP it is possible to append a length modifier to the response instruction sequence, like this:

```
### Input
User: {utterance}

### Response: (length = medium)
Character: {utterance}
```

This has an immediately noticeable effect on bot responses. The lengths using during training are:
`micro`, `tiny`, `short`, `medium`, `long`, `massive`, `huge`, `enormous`, `humongous`, `unlimited`.
**The recommended starting length is medium**. Keep in mind that the AI can ramble or impersonate
the user with very long messages.

The length control effect is reproducible, but the messages will not necessarily follow
lengths very precisely, rather follow certain ranges on average, as seen in this table
with data from tests made with one reply at the beginning of the conversation:

![lengths](https://i.imgur.com/2WXGgaV.png)

Response length control appears to work well also deep into the conversation. **By omitting
the modifier, the model will choose the most appropriate response length** (although it might
not necessarily be what the user desires).

## Intended uses & limitations

The model will show biases similar to those observed in niche roleplaying forums on the Internet, besides those exhibited by the base model.

## Training and evaluation data

For more details about LimaRP, see the dataset page.

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 4
- eval_batch_size: 4
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 8
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 10
- num_epochs: 3

### Framework versions

- Transformers 4.37.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.16.1
- Tokenizers 0.15.0
## Training procedure


The following `bitsandbytes` quantization config was used during training:
- quant_method: bitsandbytes
- load_in_8bit: False
- load_in_4bit: True
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: nf4
- bnb_4bit_use_double_quant: True
- bnb_4bit_compute_dtype: bfloat16

### Framework versions


- PEFT 0.6.0