Downloads · 30 days
106
2% of all-time downloads
Defetya_-_qwen-4B-saiga-gguf
RichardErkhov/Defetya_-_qwen-4B-saiga-gguf
Defetya_-_qwen-4B-saiga-gguf 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.
All-time downloads
6.1K
Public
Repo size
46.4 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.gguf53.7 GB · 100%
At a glance
- Access
- Public
- Created
- Oct 16, 2024
- Updated
- Oct 17, 2024
- SHA
- 20404d0c
Model card
From the Hugging Face model README
Quantization made by Richard Erkhov.
qwen-4B-saiga - GGUF
- Model creator: https://huggingface.co/Defetya/
- Original model: https://huggingface.co/Defetya/qwen-4B-saiga/
| Name | Quant method | Size |
|---|---|---|
| qwen-4B-saiga.Q2_K.gguf | Q2_K | 1.51GB |
| qwen-4B-saiga.IQ3_XS.gguf | IQ3_XS | 1.66GB |
| qwen-4B-saiga.IQ3_S.gguf | IQ3_S | 1.73GB |
| qwen-4B-saiga.Q3_K_S.gguf | Q3_K_S | 1.73GB |
| qwen-4B-saiga.IQ3_M.gguf | IQ3_M | 1.81GB |
| qwen-4B-saiga.Q3_K.gguf | Q3_K | 1.89GB |
| qwen-4B-saiga.Q3_K_M.gguf | Q3_K_M | 1.89GB |
| qwen-4B-saiga.Q3_K_L.gguf | Q3_K_L | 2.03GB |
| qwen-4B-saiga.IQ4_XS.gguf | IQ4_XS | 2.08GB |
| qwen-4B-saiga.Q4_0.gguf | Q4_0 | 2.17GB |
| qwen-4B-saiga.IQ4_NL.gguf | IQ4_NL | 2.18GB |
| qwen-4B-saiga.Q4_K_S.gguf | Q4_K_S | 2.18GB |
| qwen-4B-saiga.Q4_K.gguf | Q4_K | 2.29GB |
| qwen-4B-saiga.Q4_K_M.gguf | Q4_K_M | 2.29GB |
| qwen-4B-saiga.Q4_1.gguf | Q4_1 | 2.38GB |
| qwen-4B-saiga.Q5_0.gguf | Q5_0 | 2.58GB |
| qwen-4B-saiga.Q5_K_S.gguf | Q5_K_S | 2.58GB |
| qwen-4B-saiga.Q5_K.gguf | Q5_K | 2.64GB |
| qwen-4B-saiga.Q5_K_M.gguf | Q5_K_M | 2.64GB |
| qwen-4B-saiga.Q5_1.gguf | Q5_1 | 2.79GB |
| qwen-4B-saiga.Q6_K.gguf | Q6_K | 3.03GB |
| qwen-4B-saiga.Q8_0.gguf | Q8_0 | 3.92GB |
Original model description:
license: apache-2.0 tags:
- Russian
Qwen 4B chat by Alibaba, SFTuned on Saiga dataset. Finetuned with EasyDeL framework on v3-8 Google TPU, provided by TRC.
Модель Qwen 4B, дообученая на датасете Ильи Гусева. По моему краткому опыту общения с моделью, лучше чем Saiga-mistral. Не ошибается в падежах. Карточка модели будет дополнена после теста на Russian SuperGlue. Возможно, будет DPO
Чтобы использовать модель, необходимо назначить eos токен как <|im_end|>. Рабочий ноутбук на Kaggle: https://www.kaggle.com/code/defdet/smol-chatbot/notebook
- Created
- Oct 16, 2024
- Updated
- Oct 17, 2024