Intel/DeepSeek-V3.1-gguf-q2ks-mixed-AutoRound

Model Details

This model is a mixed gguf q2ks format of deepseek-ai/DeepSeek-V3.1 generated by intel/auto-round algorithm. Embedding layer and lm-head layer are fallback to 8 bits and non expert layers are fallback to 4 bits. Please refer to Section Generate the model for more details.

Please follow the license of the original model.

How To Use

Llama.cpp inference

./llama-cli -hf Intel/DeepSeek-V3.1-gguf-q2ks-mixed-AutoRound

> Write a quick sort algorithm.
Here is a quick implementation of the QuickSort algorithm 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)

# Example usage:
arr = [3, 6, 8, 10, 1, 2, 1]
print(quicksort(arr))
```

This implementation uses a recursive approach and selects the middle element as the pivot. It partitions the array into three parts: elements less than the pivot, elements equal to the pivot, and elements greater than the pivot. It then recursively sorts the left and right partitions and combines them with the middle partition.

Note: This version is not in-place and uses extra memory for the partitions. For an in-place version that uses less memory, you would need to implement partitioning within the original array.

Generate the model

Here is the sample command to reproduce the model

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from auto_round import AutoRound

model_name = "deepseek-ai/DeepSeek-V3.1"

model = AutoModelForCausalLM.from_pretrained(model_name,
                                             device_map="cpu", torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained(model_name)
layer_config = {}
for n, m in model.named_modules():
    if n == "lm_head" or isinstance(m,torch.nn.Embedding):
        layer_config[n] = {"bits": 8}
    elif isinstance(m, torch.nn.Linear) and (not "expert" in n or "shared_experts" in n) and n != "lm_head":
        layer_config[n] = {"bits": 4}

autoround = AutoRound(model, tokenizer, iters=0, layer_config=layer_config, nsamples=512)
autoround.quantize_and_save("tmp_autoround", format="gguf:q2_k_s")

Ethical Considerations and Limitations

The model can produce factually incorrect output, and should not be relied on to produce factually accurate information. Because of the limitations of the pretrained model and the finetuning datasets, it is possible that this model could generate lewd, biased or otherwise offensive outputs.

Therefore, before deploying any applications of the model, developers should perform safety testing.

Caveats and Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model.

Here are a couple of useful links to learn more about Intel's AI software:

• Intel Neural Compressor link

Disclaimer

The license on this model does not constitute legal advice. We are not responsible for the actions of third parties who use this model. Please consult an attorney before using this model for commercial purposes.

Cite

@article{cheng2023optimize, title={Optimize weight rounding via signed gradient descent for the quantization of llms}, author={Cheng, Wenhua and Zhang, Weiwei and Shen, Haihao and Cai, Yiyang and He, Xin and Lv, Kaokao and Liu, Yi}, journal={arXiv preprint arXiv:2309.05516}, year={2023} }

arxiv github