Downloads · 30 days
141
2% of all-time downloads
RichardErkhov/xinlai_-_Qwen2-7B-SFT-Step-DPO-gguf
xinlai_-_Qwen2-7B-SFT-Step-DPO-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
141
2% of all-time downloads
All-time downloads
5.8K
Public
Repo size
88.2 GB
Likes
0
Public
Click a slice to open those files.
.gguf102 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
Qwen2-7B-SFT-Step-DPO - GGUF
| Name | Quant method | Size |
|---|---|---|
| Qwen2-7B-SFT-Step-DPO.Q2_K.gguf | Q2_K | 2.81GB |
| Qwen2-7B-SFT-Step-DPO.IQ3_XS.gguf | IQ3_XS | 3.12GB |
| Qwen2-7B-SFT-Step-DPO.IQ3_S.gguf | IQ3_S | 3.26GB |
| Qwen2-7B-SFT-Step-DPO.Q3_K_S.gguf | Q3_K_S | 3.25GB |
| Qwen2-7B-SFT-Step-DPO.IQ3_M.gguf | IQ3_M | 3.33GB |
| Qwen2-7B-SFT-Step-DPO.Q3_K.gguf | Q3_K | 3.55GB |
| Qwen2-7B-SFT-Step-DPO.Q3_K_M.gguf | Q3_K_M | 3.55GB |
| Qwen2-7B-SFT-Step-DPO.Q3_K_L.gguf | Q3_K_L | 3.81GB |
| Qwen2-7B-SFT-Step-DPO.IQ4_XS.gguf | IQ4_XS | 3.96GB |
| Qwen2-7B-SFT-Step-DPO.Q4_0.gguf | Q4_0 | 4.13GB |
| Qwen2-7B-SFT-Step-DPO.IQ4_NL.gguf | IQ4_NL | 4.16GB |
| Qwen2-7B-SFT-Step-DPO.Q4_K_S.gguf | Q4_K_S | 4.15GB |
| Qwen2-7B-SFT-Step-DPO.Q4_K.gguf | Q4_K | 4.36GB |
| Qwen2-7B-SFT-Step-DPO.Q4_K_M.gguf | Q4_K_M | 4.36GB |
| Qwen2-7B-SFT-Step-DPO.Q4_1.gguf | Q4_1 | 4.54GB |
| Qwen2-7B-SFT-Step-DPO.Q5_0.gguf | Q5_0 | 4.95GB |
| Qwen2-7B-SFT-Step-DPO.Q5_K_S.gguf | Q5_K_S | 4.95GB |
| Qwen2-7B-SFT-Step-DPO.Q5_K.gguf | Q5_K | 5.07GB |
| Qwen2-7B-SFT-Step-DPO.Q5_K_M.gguf | Q5_K_M | 5.07GB |
| Qwen2-7B-SFT-Step-DPO.Q5_1.gguf | Q5_1 | 5.36GB |
| Qwen2-7B-SFT-Step-DPO.Q6_K.gguf | Q6_K | 5.82GB |
| Qwen2-7B-SFT-Step-DPO.Q8_0.gguf | Q8_0 | 7.54GB |
This repo contains the Qwen2-7B-SFT-Step-DPO model. It is obtained by performing Step-DPO on Qwen2-7B-SFT.
Step-DPO is a simple, effective, and data-efficient method for boosting the mathematical reasoning ability of LLMs. Notably, Step-DPO, when applied to Qwen2-72B-Instruct, achieves scores of 70.8% and 94.0% on the test sets of MATH and GSM8K without bells and wistles, respectively, surpassing a series of closed-source models, including GPT-4-1106, Claude-3-Opus, and Gemini-1.5-Pro.