Update README.md
Browse files
README.md
CHANGED
@@ -8,20 +8,20 @@ tags:
|
|
8 |
- 量化修复
|
9 |
- vLLM
|
10 |
base_model:
|
11 |
-
-
|
12 |
base_model_relation: quantized
|
13 |
---
|
14 |
# GLM-4.5-GPTQ-Int4-Int8Mix
|
15 |
-
|
16 |
|
17 |
|
18 |
-
### 【
|
19 |
-
<i
|
20 |
```
|
21 |
CONTEXT_LENGTH=32768
|
22 |
|
23 |
vllm serve \
|
24 |
-
|
25 |
--served-model-name GLM-4.5-GPTQ-Int4-Int8Mix \
|
26 |
--enable-expert-parallel \
|
27 |
--swap-space 16 \
|
@@ -34,71 +34,71 @@ vllm serve \
|
|
34 |
--disable-log-requests \
|
35 |
--host 0.0.0.0 \
|
36 |
--port 8000
|
|
|
37 |
```
|
38 |
|
39 |
-
###
|
40 |
|
41 |
```
|
42 |
vllm==0.10.0
|
43 |
```
|
44 |
|
45 |
-
###
|
46 |
```
|
47 |
2025-07-30
|
48 |
-
1.
|
49 |
```
|
50 |
|
51 |
### 【模型列表】
|
52 |
|
53 |
-
|
|
54 |
|---------|--------------|
|
55 |
| `192GB` | `2025-07-30` |
|
56 |
|
57 |
|
58 |
|
59 |
-
###
|
60 |
|
61 |
```python
|
62 |
-
from
|
63 |
-
snapshot_download('
|
64 |
```
|
65 |
|
66 |
|
67 |
-
###
|
68 |
# GLM-4.5
|
69 |
|
70 |
<div align="center">
|
71 |
<img src=https://raw.githubusercontent.com/zai-org/GLM-4.5/refs/heads/main/resources/logo.svg width="15%"/>
|
72 |
</div>
|
73 |
<p align="center">
|
74 |
-
👋
|
75 |
<br>
|
76 |
-
📖
|
77 |
<br>
|
78 |
-
📍
|
79 |
<br>
|
80 |
-
👉
|
81 |
</p>
|
|
|
|
|
82 |
|
83 |
-
|
84 |
-
|
85 |
-
**GLM-4.5** 系列模型是专为智能体设计的基础模型。GLM-4.5拥有 **3550** 亿总参数量,其中 **320** 亿活跃参数;GLM-4.5-Air采用更紧凑的设计,拥有
|
86 |
-
**1060** 亿总参数量,其中 **120** 亿活跃参数。GLM-4.5模型统一了推理、编码和智能体能力,以满足智能体应用的复杂需求。
|
87 |
|
88 |
-
GLM-4.5
|
89 |
|
90 |
-
|
91 |
|
92 |
-
|
93 |
-
。值得注意的是,GLM-4.5-Air在保持优异效率的同时,仍取得了 **59.8** 的竞争性成绩。
|
94 |
|
95 |

|
96 |
|
97 |
-
|
|
|
|
|
98 |
|
99 |
-
|
100 |
-
和 [SGLang](https://github.com/sgl-project/sglang/blob/main/python/sglang/srt/models/glm4_moe.py) 的实现中找到。
|
101 |
|
102 |
-
##
|
103 |
|
104 |
-
|
|
|
8 |
- 量化修复
|
9 |
- vLLM
|
10 |
base_model:
|
11 |
+
- zai-org/GLM-4.5
|
12 |
base_model_relation: quantized
|
13 |
---
|
14 |
# GLM-4.5-GPTQ-Int4-Int8Mix
|
15 |
+
Base model [zai-org/GLM-4.5](https://huggingface.co/zai-org/GLM-4.5)
|
16 |
|
17 |
|
18 |
+
### 【VLLM Launch Command for 8-GPU Single Node】
|
19 |
+
<i>Note: When launching this model on 8 GPUs, you must include --enable-expert-parallel, otherwise expert tensor partitioning will fail due to mismatch. This flag is not required for 4-GPU setups.</i>
|
20 |
```
|
21 |
CONTEXT_LENGTH=32768
|
22 |
|
23 |
vllm serve \
|
24 |
+
QuantTrio/GLM-4.5-GPTQ-Int4-Int8Mix \
|
25 |
--served-model-name GLM-4.5-GPTQ-Int4-Int8Mix \
|
26 |
--enable-expert-parallel \
|
27 |
--swap-space 16 \
|
|
|
34 |
--disable-log-requests \
|
35 |
--host 0.0.0.0 \
|
36 |
--port 8000
|
37 |
+
|
38 |
```
|
39 |
|
40 |
+
### 【Dependencies】
|
41 |
|
42 |
```
|
43 |
vllm==0.10.0
|
44 |
```
|
45 |
|
46 |
+
### 【Model Update】
|
47 |
```
|
48 |
2025-07-30
|
49 |
+
1. fast commit
|
50 |
```
|
51 |
|
52 |
### 【模型列表】
|
53 |
|
54 |
+
| File Size | Last Updated |
|
55 |
|---------|--------------|
|
56 |
| `192GB` | `2025-07-30` |
|
57 |
|
58 |
|
59 |
|
60 |
+
### 【Model Download】
|
61 |
|
62 |
```python
|
63 |
+
from huggingface_hub import snapshot_download
|
64 |
+
snapshot_download('QuantTrio/GLM-4.5-GPTQ-Int4-Int8Mix', cache_dir="your_local_path")
|
65 |
```
|
66 |
|
67 |
|
68 |
+
### 【Overview】
|
69 |
# GLM-4.5
|
70 |
|
71 |
<div align="center">
|
72 |
<img src=https://raw.githubusercontent.com/zai-org/GLM-4.5/refs/heads/main/resources/logo.svg width="15%"/>
|
73 |
</div>
|
74 |
<p align="center">
|
75 |
+
👋 Join our <a href="https://discord.gg/QR7SARHRxK" target="_blank">Discord</a> community.
|
76 |
<br>
|
77 |
+
📖 Check out the GLM-4.5 <a href="https://z.ai/blog/glm-4.5" target="_blank">technical blog</a>.
|
78 |
<br>
|
79 |
+
📍 Use GLM-4.5 API services on <a href="https://docs.z.ai/guides/llm/glm-4.5">Z.ai API Platform (Global)</a> or <br> <a href="https://docs.bigmodel.cn/cn/guide/models/text/glm-4.5">Zhipu AI Open Platform (Mainland China)</a>.
|
80 |
<br>
|
81 |
+
👉 One click to <a href="https://chat.z.ai">GLM-4.5</a>.
|
82 |
</p>
|
83 |
+
|
84 |
+
## Model Introduction
|
85 |
|
86 |
+
The **GLM-4.5** series models are foundation models designed for intelligent agents. GLM-4.5 has **355** billion total parameters with **32** billion active parameters, while GLM-4.5-Air adopts a more compact design with **106** billion total parameters and **12** billion active parameters. GLM-4.5 models unify reasoning, coding, and intelligent agent capabilities to meet the complex demands of intelligent agent applications.
|
|
|
|
|
|
|
87 |
|
88 |
+
Both GLM-4.5 and GLM-4.5-Air are hybrid reasoning models that provide two modes: thinking mode for complex reasoning and tool usage, and non-thinking mode for immediate responses.
|
89 |
|
90 |
+
We have open-sourced the base models, hybrid reasoning models, and FP8 versions of the hybrid reasoning models for both GLM-4.5 and GLM-4.5-Air. They are released under the MIT open-source license and can be used commercially and for secondary development.
|
91 |
|
92 |
+
As demonstrated in our comprehensive evaluation across 12 industry-standard benchmarks, GLM-4.5 achieves exceptional performance with a score of **63.2**, in the **3rd** place among all the proprietary and open-source models. Notably, GLM-4.5-Air delivers competitive results at **59.8** while maintaining superior efficiency.
|
|
|
93 |
|
94 |

|
95 |
|
96 |
+
For more eval results, show cases, and technical details, please visit
|
97 |
+
our [technical blog](https://z.ai/blog/glm-4.5). The technical report will be released soon.
|
98 |
+
|
99 |
|
100 |
+
The model code, tool parser and reasoning parser can be found in the implementation of [transformers](https://github.com/huggingface/transformers/tree/main/src/transformers/models/glm4_moe), [vLLM](https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/models/glm4_moe_mtp.py) and [SGLang](https://github.com/sgl-project/sglang/blob/main/python/sglang/srt/models/glm4_moe.py).
|
|
|
101 |
|
102 |
+
## Quick Start
|
103 |
|
104 |
+
Please refer our [github page](https://github.com/zai-org/GLM-4.5) for more detail.
|