File size: 4,720 Bytes
e0bc976 f31870a e0bc976 789741e e0bc976 6f9297d e0bc976 4f14824 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 |
---
base_model: Infinigence/Megrez-3B-Instruct
license: apache-2.0
model_creator: Infinigence
model_name: Megrez-3B-Instruct
quantized_by: Second State Inc.
language:
- en
- zh
pipeline_tag: text-generation
---
<!-- header start -->
<!-- 200823 -->
<div style="width: auto; margin-left: auto; margin-right: auto">
<img src="https://github.com/LlamaEdge/LlamaEdge/raw/dev/assets/logo.svg" style="width: 100%; min-width: 400px; display: block; margin: auto;">
</div>
<hr style="margin-top: 1.0em; margin-bottom: 1.0em;">
<!-- header end -->
# Megrez-3B-Instruct-GGUF
## Original Model
[Infinigence/Megrez-3B-Instruct](https://huggingface.co/Infinigence/Megrez-3B-Instruct)
## Run with LlamaEdge
- LlamaEdge version: [v0.16.0](https://github.com/LlamaEdge/LlamaEdge/releases/tag/0.16.0)
- Prompt template
- Prompt type: `megrez`
- Prompt string
```text
<|role_start|>system<|role_end|>{system_message}<|turn_end|><|role_start|>user<|role_end|>{user_message}<|turn_end|><|role_start|>assistant<|role_end|>
```
- Context size: `32000`
- Run as LlamaEdge service
```bash
wasmedge --dir .:. --nn-preload default:GGML:AUTO:Megrez-3B-Instruct-Q5_K_M.gguf \
llama-api-server.wasm \
--model-name Megrez-3B-Instruct \
--prompt-template megrez \
--ctx-size 32000
```
> [!TIP]
> For use cases of conversations or article writing, `temperature=0.7` is strongly recommended.
> For use cases of mathematics or logical reasoning, `temperature=0.2` is strongly recommended.
- Run as LlamaEdge command app
```bash
wasmedge --dir .:. --nn-preload default:GGML:AUTO:Megrez-3B-Instruct-Q5_K_M.gguf \
llama-chat.wasm \
--prompt-template megrez \
--ctx-size 32000
```
> [!TIP]
> For use cases of conversations or article writing, `temperature=0.7` is strongly recommended.
> For use cases of mathematics or logical reasoning, `temperature=0.2` is strongly recommended.
## Quantized GGUF Models
| Name | Quant method | Bits | Size | Use case |
| ---- | ---- | ---- | ---- | ----- |
| [Megrez-3B-Instruct-Q2_K.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q2_K.gguf) | Q2_K | 2 | 1.21 GB| smallest, significant quality loss - not recommended for most purposes |
| [Megrez-3B-Instruct-Q3_K_L.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q3_K_L.gguf) | Q3_K_L | 3 | 1.60 GB| small, substantial quality loss |
| [Megrez-3B-Instruct-Q3_K_M.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q3_K_M.gguf) | Q3_K_M | 3 | 1.50 GB| very small, high quality loss |
| [Megrez-3B-Instruct-Q3_K_S.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q3_K_S.gguf) | Q3_K_S | 3 | 1.38 GB| very small, high quality loss |
| [Megrez-3B-Instruct-Q4_0.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q4_0.gguf) | Q4_0 | 4 | 1.73 GB| legacy; small, very high quality loss - prefer using Q3_K_M |
| [Megrez-3B-Instruct-Q4_K_M.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q4_K_M.gguf) | Q4_K_M | 4 | 1.81 GB| medium, balanced quality - recommended |
| [Megrez-3B-Instruct-Q4_K_S.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q4_K_S.gguf) | Q4_K_S | 4 | 1.74 GB| small, greater quality loss |
| [Megrez-3B-Instruct-Q5_0.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q5_0.gguf) | Q5_0 | 5 | 2.05 GB| legacy; medium, balanced quality - prefer using Q4_K_M |
| [Megrez-3B-Instruct-Q5_K_M.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q5_K_M.gguf) | Q5_K_M | 5 | 2.09 GB| large, very low quality loss - recommended |
| [Megrez-3B-Instruct-Q5_K_S.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q5_K_S.gguf) | Q5_K_S | 5 | 2.05 GB| large, low quality loss - recommended |
| [Megrez-3B-Instruct-Q6_K.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q6_K.gguf) | Q6_K | 6 | 2.40 GB| very large, extremely low quality loss |
| [Megrez-3B-Instruct-Q8_0.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-Q8_0.gguf) | Q8_0 | 8 | 3.10 GB| very large, extremely low quality loss - not recommended |
| [Megrez-3B-Instruct-f16.gguf](https://huggingface.co/second-state/Megrez-3B-Instruct-GGUF/blob/main/Megrez-3B-Instruct-f16.gguf) | f16 | 16 | 5.84 GB| |
*Quantized with llama.cpp b4381* |