Downloads · 30 days
0
RichardErkhov/thesven_-_Aether-Qwen2-0.5B-SFT-v0.0.2-exl2
thesven_-_Aether-Qwen2-0.5B-SFT-v0.0.2-exl2 is a machine learning model from RichardErkhov. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Downloads · 30 days
0
Access
Public
Updated Jan 18, 2025
Repo size
3.2 GB
Likes
0
Public
Click a slice to open those files.
.json1.3 MB · 99%
From the Hugging Face model README
Quantization made by Richard Erkhov.
Aether-Qwen2-0.5B-SFT-v0.0.2 - EXL2
| Branch | Bits | Description | | ----- | ---- | ------- | ------ | ------ | ------ | ------ | ------------ | | 8_0 | 8.0 | Maximum quality that ExLlamaV2 can produce, near unquantized performance. | | 6_5 | 6.5 | Very similar to 8.0, good tradeoff of size vs performance, recommended. | | 5_0 | 5.0 | Slightly lower quality vs 6.5, but usable on 8GB cards. | | 4_25 | 4.25 | GPTQ equivalent bits per weight, slightly higher quality. | | 3_5 | 3.5 | Lower quality, only use if you have to. |
With git:
git clone --single-branch --branch 6_5 https://huggingface.co/thesven_-_Aether-Qwen2-0.5B-SFT-v0.0.2-exl2 Aether-Qwen2-0.5B-SFT-v0.0.2-6_5
With huggingface hub:
pip3 install huggingface-hub
To download a specific branch, use the --revision parameter. For example, to download the 6.5 bpw branch:
Linux:
huggingface-cli download thesven_-_Aether-Qwen2-0.5B-SFT-v0.0.2-exl2 --revision 6_5 --local-dir Aether-Qwen2-0.5B-SFT-v0.0.2-6_5 --local-dir-use-symlinks False
Windows (which apparently doesn't like _ in folders sometimes?):
huggingface-cli download thesven_-_Aether-Qwen2-0.5B-SFT-v0.0.2-exl2 --revision 6_5 --local-dir Aether-Qwen2-0.5B-SFT-v0.0.2-6.5 --local-dir-use-symlinks False
library_name: transformers tags:

This model is an iteration of the Qwen2 model, fine-tuned using Supervised Fine-Tuning (SFT) on the AetherCode-v1 dataset specifically for code-related tasks. It combines the advanced capabilities of the base Qwen2 model with specialized training to enhance its performance in software development contexts.
Aether-Qwen2-0.5B-SFT-v0.0.1 is a transformer model from the Hugging Face 🤗 transformers library, designed to facilitate and improve automated coding tasks. This model has been enhanced via Supervised Fine-Tuning (SFT) to better understand and generate code, making it ideal for applications in software development, code review, and automated programming assistance.
This model is ready for direct use in environments where coding assistance is needed, providing capabilities such as code completion, error detection, and suggestions for code optimization.
Further fine-tuning on specific coding languages or frameworks can extend its utility to more specialized software development tasks.
The model should not be used for general natural language processing tasks outside the scope of programming and code analysis.
Users should be cautious about relying solely on the model for critical software development tasks without human oversight, due to potential biases in training data or limitations in understanding complex code contexts.
Ongoing validation and testing on diverse coding datasets are recommended to ensure the model remains effective and unbiased.
Use the code below to get started with the model.
from transformers import AutoModel
model = AutoModel.from_pretrained("thesven/Aether-Qwen2-0.5B-SFT-v0.0.2")
or with a pipeline:
from transformers import pipeline
messages = [
{"role": "system", "content": "You are a helpful software development assistant"},
{"role": "user", "content": "can you write a python function that adds 3 numbers together?"},
]
pipe = pipeline("text-generation", model="thesven/Aether-Qwen2-0.5B-SFT-v0.0.2")
print(pipe(messages))
<|im_start|>system
{system}<|im_end|>
<|im_start|>user
{user}<|im_end|>
<|im_start|>assistant
{assistant}
The model was trained using the 5star split from the AetherCode-v1 dataset, designed for enhancing coding-related AI capabilities.
Training regime: The model was trained for 3 epochs on an RTX 4500 using Supervised Fine-Tuning (SFT)
Standard preprocessing techniques were applied to prepare the code data for training.