Downloads · 30 days
402
7% of all-time downloads
RichardErkhov/duoqi_-_Nanbeige2-16B-Chat-gguf
duoqi_-_Nanbeige2-16B-Chat-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.
Downloads · 30 days
402
7% of all-time downloads
All-time downloads
5.4K
Public
Repo size
180 GB
Likes
0
Public
Click a slice to open those files.
.gguf208 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
Nanbeige2-16B-Chat - GGUF
| Name | Quant method | Size |
|---|---|---|
| Nanbeige2-16B-Chat.Q2_K.gguf | Q2_K | 5.54GB |
| Nanbeige2-16B-Chat.IQ3_XS.gguf | IQ3_XS | 6.12GB |
| Nanbeige2-16B-Chat.IQ3_S.gguf | IQ3_S | 6.45GB |
| Nanbeige2-16B-Chat.Q3_K_S.gguf | Q3_K_S | 6.45GB |
| Nanbeige2-16B-Chat.IQ3_M.gguf | IQ3_M | 6.81GB |
| Nanbeige2-16B-Chat.Q3_K.gguf | Q3_K | 7.21GB |
| Nanbeige2-16B-Chat.Q3_K_M.gguf | Q3_K_M | 7.21GB |
| Nanbeige2-16B-Chat.Q3_K_L.gguf | Q3_K_L | 7.87GB |
| Nanbeige2-16B-Chat.IQ4_XS.gguf | IQ4_XS | 7.98GB |
| Nanbeige2-16B-Chat.Q4_0.gguf | Q4_0 | 8.37GB |
| Nanbeige2-16B-Chat.IQ4_NL.gguf | IQ4_NL | 8.42GB |
| Nanbeige2-16B-Chat.Q4_K_S.gguf | Q4_K_S | 8.43GB |
| Nanbeige2-16B-Chat.Q4_K.gguf | Q4_K | 8.93GB |
| Nanbeige2-16B-Chat.Q4_K_M.gguf | Q4_K_M | 8.93GB |
| Nanbeige2-16B-Chat.Q4_1.gguf | Q4_1 | 9.27GB |
| Nanbeige2-16B-Chat.Q5_0.gguf | Q5_0 | 10.18GB |
| Nanbeige2-16B-Chat.Q5_K_S.gguf | Q5_K_S | 10.18GB |
| Nanbeige2-16B-Chat.Q5_K.gguf | Q5_K | 10.47GB |
| Nanbeige2-16B-Chat.Q5_K_M.gguf | Q5_K_M | 10.47GB |
| Nanbeige2-16B-Chat.Q5_1.gguf | Q5_1 | 11.08GB |
| Nanbeige2-16B-Chat.Q6_K.gguf | Q6_K | 12.1GB |
| Nanbeige2-16B-Chat.Q8_0.gguf | Q8_0 | 15.67GB |
language:
Introduction
A Llama version for Nanbeige-16B-Chat, which could be loaded by LlamaForCausalLM.
The Nanbeige2-16B-Chat is the latest 16B model developed by the Nanbeige Lab, which utilized 4.5T tokens of high-quality training data during the training phase. During the alignment phase, we initially trained our model using 1 million samples through Supervised Fine-Tuning (SFT). We then engaged in curriculum learning with 400,000 high-quality samples that presented a greater level of difficulty. Subsequently, we incorporated human feedback through the Direct Preference Optimization (DPO), culminating in the development of Nanbeige2-16B-Chat. Nanbeige2-16B-Chat has achieved superior performance across various authoritative benchmark datasets.