---
license: apache-2.0
library_name: transformers
---
# Laser-Dolphin-Mixtral-4x7b-dpo

*New version is coming because of chat template issues. The other MoE models in my collection do not have this issue and have been tested more*

![laser_dolphin_image](./dolphin_moe.png)

Credit to Fernando Fernandes and Eric Hartford for their project [laserRMT](https://github.com/cognitivecomputations/laserRMT)

This model is a medium-sized MoE implementation based on [cognitivecomputations/dolphin-2.6-mistral-7b-dpo-laser](https://huggingface.co/cognitivecomputations/dolphin-2.6-mistral-7b-dpo-laser)

The process is outlined in this [notebook](https://github.com/cognitivecomputations/laserRMT/blob/main/examples/laser-dolphin-mixtral-2x7b.ipynb)

#### Notes: 

DPO being used in only a portion of the merge seems to be causing issues with code creation.

## Code Example

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

def generate_response(prompt):
    """
    Generate a response from the model based on the input prompt.

    Args:
    prompt (str): Prompt for the model.

    Returns:
    str: The generated response from the model.
    """
    # Tokenize the input prompt
    inputs = tokenizer(prompt, return_tensors="pt")

    # Generate output tokens
    outputs = model.generate(**inputs, max_new_tokens=256, eos_token_id=tokenizer.eos_token_id, pad_token_id=tokenizer.pad_token_id)

    # Decode the generated tokens to a string
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)

    return response

# Load the model and tokenizer
model_id = "macadeliccc/laser-dolphin-mixtral-4x7b-dpo"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, load_in_4bit=True)

prompt = "Write a quicksort algorithm in python"

# Generate and print responses for each language
print("Response:")
print(generate_response(prompt), "\n")
```

## Example output

can you write me a quicksort algorithm in python?

Sure, here's a quicksort algorithm implemented in Python:

```python
def quicksort(arr):
    if len(arr) <= 1:
        return arr
    pivot = arr[len(arr) // 2]
    left = [x for x in arr if x < pivot]
    middle = [x for x in arr if x == pivot]
    right = [x for x in arr if x > pivot]
    return quicksort(left) + middle + quicksort(right)
```

This implementation uses the median of the array as the pivot. 
It first checks if the array has one or fewer elements, in which case it is already sorted and can be returned as is.
Otherwise, it selects the pivot as the middle element of the array. Then, it partitions the array into three sub-arrays: elements less than the pivot, elements equal to the pivot, and elements greater than the pivot. 
It recursively sorts the left and right sub-arrays and concatenates the results with the middle sub-array to obtain the final sorted array. 


## GGUF

Q4_K_M and Q5_K_M quants are available [here](https://huggingface.co/macadeliccc/laser-dolphin-mixtral-4x7b-dpo-GGUF)

![dolphin-cpp-2](dolphin-cpp-2.png)

## Eval

**Models were evaluated in 4bit due to GPU requirements**

I will evaluate on  colab with an A100 asap

|    Tasks    |Version|Filter|n-shot| Metric |Value |   |Stderr|
|-------------|-------|------|-----:|--------|-----:|---|-----:|
|arc_challenge|Yaml   |none  |     0|acc     |0.5538|±  |0.0145|
|             |       |none  |     0|acc_norm|0.5734|±  |0.0145|
|arc_easy     |Yaml   |none  |     0|acc     |0.8291|±  |0.0077|
|             |       |none  |     0|acc_norm|0.7807|±  |0.0085|
|boolq        |Yaml   |none  |     0|acc     |0.8694|±  |0.0059|
|hellaswag    |Yaml   |none  |     0|acc     |0.6402|±  |0.0048|
|             |       |none  |     0|acc_norm|0.8233|±  |0.0038|
|openbookqa   |Yaml   |none  |     0|acc     |0.3380|±  |0.0212|
|             |       |none  |     0|acc_norm|0.4720|±  |0.0223|
|piqa         |Yaml   |none  |     0|acc     |0.8123|±  |0.0091|
|             |       |none  |     0|acc_norm|0.8221|±  |0.0089|
|winogrande   |Yaml   |none  |     0|acc     |0.7348|±  |0.0124|


## Citations

Fernando Fernandes Neto and Eric Hartford. "Optimizing Large Language Models Using Layer-Selective Rank Reduction and Random Matrix Theory." 2024.

```bibtex
@article{sharma2023truth,
title={The Truth is in There: Improving Reasoning in Language Models with Layer-Selective Rank Reduction},
author={Sharma, Pratyusha and Ash, Jordan T and Misra, Dipendra},
journal={arXiv preprint arXiv:2312.13558},
year={2023} }
```

```bibtex
@article{gao2021framework,
  title={A framework for few-shot language model evaluation},
  author={Gao, Leo and Tow, Jonathan and Biderman, Stella and Black, Sid and DiPofi, Anthony and Foster, Charles and Golding, Laurence and Hsu, Jeffrey and McDonell, Kyle and Muennighoff, Niklas and others},
  journal={Version v0. 0.1. Sept},
  year={2021}
}
```