Downloads · 30 days
247
5% of all-time downloads
RichardErkhov/Qwen_-_Qwen1.5-MoE-A2.7B-gguf
Qwen_-_Qwen1.5-MoE-A2.7B-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
247
5% of all-time downloads
All-time downloads
5.4K
Public
Repo size
169 GB
Likes
1
Public
Click a slice to open those files.
.gguf197 GB · 100%
From the Hugging Face model README
Quantization made by Richard Erkhov.
Qwen1.5-MoE-A2.7B - GGUF
| Name | Quant method | Size |
|---|---|---|
| Qwen1.5-MoE-A2.7B.Q2_K.gguf | Q2_K | 5.49GB |
| Qwen1.5-MoE-A2.7B.IQ3_XS.gguf | IQ3_XS | 6.07GB |
| Qwen1.5-MoE-A2.7B.IQ3_S.gguf | IQ3_S | 6.37GB |
| Qwen1.5-MoE-A2.7B.Q3_K_S.gguf | Q3_K_S | 6.37GB |
| Qwen1.5-MoE-A2.7B.IQ3_M.gguf | IQ3_M | 6.46GB |
| Qwen1.5-MoE-A2.7B.Q3_K.gguf | Q3_K | 6.93GB |
| Qwen1.5-MoE-A2.7B.Q3_K_M.gguf | Q3_K_M | 6.93GB |
| Qwen1.5-MoE-A2.7B.Q3_K_L.gguf | Q3_K_L | 7.21GB |
| Qwen1.5-MoE-A2.7B.IQ4_XS.gguf | IQ4_XS | 7.4GB |
| Qwen1.5-MoE-A2.7B.Q4_0.gguf | Q4_0 | 7.59GB |
| Qwen1.5-MoE-A2.7B.IQ4_NL.gguf | IQ4_NL | 7.68GB |
| Qwen1.5-MoE-A2.7B.Q4_K_S.gguf | Q4_K_S | 8.11GB |
| Qwen1.5-MoE-A2.7B.Q4_K.gguf | Q4_K | 8.84GB |
| Qwen1.5-MoE-A2.7B.Q4_K_M.gguf | Q4_K_M | 8.84GB |
| Qwen1.5-MoE-A2.7B.Q4_1.gguf | Q4_1 | 8.41GB |
| Qwen1.5-MoE-A2.7B.Q5_0.gguf | Q5_0 | 9.22GB |
| Qwen1.5-MoE-A2.7B.Q5_K_S.gguf | Q5_K_S | 9.46GB |
| Qwen1.5-MoE-A2.7B.Q5_K.gguf | Q5_K | 10.09GB |
| Qwen1.5-MoE-A2.7B.Q5_K_M.gguf | Q5_K_M | 10.09GB |
| Qwen1.5-MoE-A2.7B.Q5_1.gguf | Q5_1 | 10.04GB |
| Qwen1.5-MoE-A2.7B.Q6_K.gguf | Q6_K | 11.89GB |
| Qwen1.5-MoE-A2.7B.Q8_0.gguf | Q8_0 | 14.18GB |
license: other license_name: tongyi-qianwen license_link: >- https://huggingface.co/Qwen/Qwen1.5-MoE-A2.7B/blob/main/LICENSE language:
Qwen1.5-MoE is a transformer-based MoE decoder-only language model pretrained on a large amount of data.
For more details, please refer to our blog post and GitHub repo.
Qwen1.5-MoE employs Mixture of Experts (MoE) architecture, where the models are upcycled from dense language models. For instance, Qwen1.5-MoE-A2.7B is upcycled from Qwen-1.8B. It has 14.3B parameters in total and 2.7B activated parameters during runtime, while achieving comparable performance to Qwen1.5-7B, it only requires 25% of the training resources. We also observed that the inference speed is 1.74 times that of Qwen1.5-7B.
The code of Qwen1.5-MoE has been in the latest Hugging face transformers and we advise you to build from source with command pip install git+https://github.com/huggingface/transformers, or you might encounter the following error:
KeyError: 'qwen2_moe'.
We do not advise you to use base language models for text generation. Instead, you can apply post-training, e.g., SFT, RLHF, continued pretraining, etc., on this model.