diff --git a/.gitattributes b/.gitattributes index f23f37e..dacca79 100644 --- a/.gitattributes +++ b/.gitattributes @@ -45,3 +45,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text *.wasm filter=lfs diff=lfs merge=lfs -text *.zst filter=lfs diff=lfs merge=lfs -text *tfevents* filter=lfs diff=lfs merge=lfs -text + +notebook.ipynb filter=lfs diff=lfs merge=lfs -text \ No newline at end of file diff --git a/README.md b/README.md index b29421f..dc38374 100644 --- a/README.md +++ b/README.md @@ -1,5 +1,14 @@ --- license: mit +language: +- en +- zh +base_model: +- zai-org/GLM-4-9B-0414 +pipeline_tag: image-text-to-text +library_name: transformers +tags: +- reasoning --- # GLM-4.1V-9B-Thinking @@ -8,45 +17,56 @@ license: mit

- 📖 查看 GLM-4.1V-9B-Thinking 论文 。 + 📖 View the GLM-4.1V-9B-Thinking paper.
- 📍 在 智谱大模型开放平台 使用 GLM-4.1V-9B-Thinking 的API服务。 + 📍 Using GLM-4.1V-9B-Thinking API at Zhipu Foundation Model Open Platform

-## 模型介绍 -视觉语言大模型（VLM）已经成为智能系统的关键基石。随着真实世界的智能任务越来越复杂，VLM模型也亟需在基本的多模态感知之外， -逐渐增强复杂任务中的推理能力，提升自身的准确性、全面性和智能化程度，使得复杂问题解决、长上下文理解、多模态智能体等智能任务成为可能。 +## Model Introduction -基于 [GLM-4-9B-0414](https://github.com/zai-org/GLM-4) 基座模型，我们推出新版VLM开源模型 **GLM-4.1V-9B-Thinking** -，引入思考范式，通过课程采样强化学习 RLCS（Reinforcement Learning with Curriculum Sampling）全面提升模型能力， -达到 10B 参数级别的视觉语言模型的最强性能，在18个榜单任务中持平甚至超过8倍参数量的 Qwen-2.5-VL-72B。 -我们同步开源基座模型 **GLM-4.1V-9B-Base**，希望能够帮助更多研究者探索视觉语言模型的能力边界。 +Vision-Language Models (VLMs) have become foundational components of intelligent systems. As real-world AI tasks grow +increasingly complex, VLMs must evolve beyond basic multimodal perception to enhance their reasoning capabilities in +complex tasks. This involves improving accuracy, comprehensiveness, and intelligence, enabling applications such as +complex problem solving, long-context understanding, and multimodal agents. + +Based on the [GLM-4-9B-0414](https://github.com/zai-org/GLM-4) foundation model, we present the new open-source VLM model +**GLM-4.1V-9B-Thinking**, designed to explore the upper limits of reasoning in vision-language models. By introducing +a "thinking paradigm" and leveraging reinforcement learning, the model significantly enhances its capabilities. It +achieves state-of-the-art performance among 10B-parameter VLMs, matching or even surpassing the 72B-parameter +Qwen-2.5-VL-72B on 18 benchmark tasks. We are also open-sourcing the base model GLM-4.1V-9B-Base to +support further research into the boundaries of VLM capabilities. ![rl](https://raw.githubusercontent.com/zai-org/GLM-4.1V-Thinking/refs/heads/main/resources/rl.jpeg) -与上一代的 CogVLM2 及 GLM-4V 系列模型相比，**GLM-4.1V-Thinking** 有如下改进： +Compared to the previous generation models CogVLM2 and the GLM-4V series, **GLM-4.1V-Thinking** offers the +following improvements: -1. 系列中首个推理模型，不仅仅停留在数学领域，在多个子领域均达到世界前列的水平。 -2. 支持 **64k** 上下长度。 -3. 支持**任意长宽比**和高达 **4k** 的图像分辨率。 -4. 提供支持**中英文双语**的开源模型版本。 +1. The first reasoning-focused model in the series, achieving world-leading performance not only in mathematics but also + across various sub-domains. +2. Supports **64k** context length. +3. Handles **arbitrary aspect ratios** and up to **4K** image resolution. +4. Provides an open-source version supporting both **Chinese and English bilingual** usage. -## 榜单信息 +## Benchmark Performance -GLM-4.1V-9B-Thinking 通过引入「思维链」（Chain-of-Thought）推理机制，在回答准确性、内容丰富度与可解释性方面， -全面超越传统的非推理式视觉模型。在28项评测任务中有23项达到10B级别模型最佳，甚至有18项任务超过8倍参数量的Qwen-2.5-VL-72B。 +By incorporating the Chain-of-Thought reasoning paradigm, GLM-4.1V-9B-Thinking significantly improves answer accuracy, +richness, and interpretability. It comprehensively surpasses traditional non-reasoning visual models. +Out of 28 benchmark tasks, it achieved the best performance among 10B-level models on 23 tasks, +and even outperformed the 72B-parameter Qwen-2.5-VL-72B on 18 tasks. ![bench](https://raw.githubusercontent.com/zai-org/GLM-4.1V-Thinking/refs/heads/main/resources/bench.jpeg) -## 快速推理 +## Quick Inference + +This is a simple example of running single-image inference using the `transformers` library. +First, install the `transformers` library from source: -这里展现了一个使用`transformers`进行单张图片推理的代码。首先，从源代码安装`transformers`库。 ``` pip install transformers>=4.57.1 ``` -接着按照以下代码运行: +Then, run the following code: ```python from transformers import AutoProcessor, Glm4vForConditionalGeneration @@ -59,7 +79,7 @@ messages = [ "content": [ { "type": "image", - "url": "https://model-demo.oss-cn-hangzhou.aliyuncs.com/Grayscale_8bits_palette_sample_image.png" + "url": "https://upload.wikimedia.org/wikipedia/commons/f/fa/Grayscale_8bits_palette_sample_image.png" }, { "type": "text", @@ -68,13 +88,12 @@ messages = [ ], } ] +processor = AutoProcessor.from_pretrained(MODEL_PATH, use_fast=True) model = Glm4vForConditionalGeneration.from_pretrained( pretrained_model_name_or_path=MODEL_PATH, torch_dtype=torch.bfloat16, device_map="auto", ) -processor = AutoProcessor.from_pretrained(MODEL_PATH, use_fast=True) - inputs = processor.apply_chat_template( messages, tokenize=True, @@ -87,6 +106,5 @@ output_text = processor.decode(generated_ids[0][inputs["input_ids"].shape[1]:], print(output_text) ``` - -视频推理，网页端Demo部署等更代码请查看我们的 [github](https://github.com/zai-org/GLM-V)。 - +For video reasoning, web demo deployment, and more code, please check +our [GitHub](https://github.com/zai-org/GLM-V). \ No newline at end of file diff --git a/notebook.ipynb b/notebook.ipynb new file mode 100644 index 0000000..eb9cc9a --- /dev/null +++ b/notebook.ipynb @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4e773310cc43f99e534f54b458655bc6f70c439304ac11b6624edec476b64df1 +size 2537101