You can deploy this model and start earning money today!

RichardErkhov/
royallab_-_ZephRP-m7b-gguf

Name: royallab_-_ZephRP-m7b-gguf
Availability: InStock

endpoints_compatible

gguf

template

Views

No views yet

Quantization made by Richard Erkhov.

Github

Discord

Request more models

ZephRP-m7b - GGUF

Model creator: https://huggingface.co/royallab/
Original model: https://huggingface.co/royallab/ZephRP-m7b/

Name	Quant method	Size
ZephRP-m7b.Q2_K.gguf	Q2_K	2.53GB
ZephRP-m7b.IQ3_XS.gguf	IQ3_XS	2.81GB
ZephRP-m7b.IQ3_S.gguf	IQ3_S	2.96GB
ZephRP-m7b.Q3_K_S.gguf	Q3_K_S	2.95GB
ZephRP-m7b.IQ3_M.gguf	IQ3_M	3.06GB
ZephRP-m7b.Q3_K.gguf	Q3_K	3.28GB
ZephRP-m7b.Q3_K_M.gguf	Q3_K_M	3.28GB
ZephRP-m7b.Q3_K_L.gguf	Q3_K_L	3.56GB
ZephRP-m7b.IQ4_XS.gguf	IQ4_XS	3.67GB
ZephRP-m7b.Q4_0.gguf	Q4_0	3.83GB
ZephRP-m7b.IQ4_NL.gguf	IQ4_NL	3.87GB
ZephRP-m7b.Q4_K_S.gguf	Q4_K_S	3.86GB
ZephRP-m7b.Q4_K.gguf	Q4_K	4.07GB
ZephRP-m7b.Q4_K_M.gguf	Q4_K_M	4.07GB
ZephRP-m7b.Q4_1.gguf	Q4_1	4.24GB
ZephRP-m7b.Q5_0.gguf	Q5_0	4.65GB
ZephRP-m7b.Q5_K_S.gguf	Q5_K_S	4.65GB
ZephRP-m7b.Q5_K.gguf	Q5_K	4.78GB
ZephRP-m7b.Q5_K_M.gguf	Q5_K_M	4.78GB
ZephRP-m7b.Q5_1.gguf	Q5_1	5.07GB
ZephRP-m7b.Q6_K.gguf	Q6_K	5.53GB
ZephRP-m7b.Q8_0.gguf	Q8_0	7.17GB

Original model description:

inference: false language:

en library_name: transformers pipeline_tag: text-generation tags:
mistral license: cc-by-nc-4.0

ZephRP-m7b

This is a Mistral-based model consisting of a merge between HuggingFaceH4/zephyr-7b-alpha and PEFT adapter trained using the LimaRP dataset.

The goal was to combine the message length instruction training of LimaRPv3 and additional stylistic elements with the superior knowledge and instruction-following capabilities of the Zephyr model.

Usage:

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. 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.)

Message length control

Due to the inclusion of LimaRP v3, 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 available lengths are: micro, tiny, short, medium, long, massive, huge, enormous, humongous, unlimited. The recommended starting length is medium. Keep in mind that the AI may ramble or impersonate the user with very long messages.

Bias, Risks, and Limitations

The model will show biases similar to those observed in niche roleplaying forums on the Internet, besides those exhibited by the base model. It is not intended for supplying factual information or advice in any form.

Training Details

The LimaRP PEFT adapter was trained as an 8-bit lora using axolotl.

The following hyperparameters were used during training of the adapter on the original mistralai/Mistral-7B-v0.1 model using a single L40 GPU:

learning_rate: 0.00015
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 4
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: 2

RichardErkhov/royallab_-_ZephRP-m7b-gguf