为什么运行python server.py后mlx_vlm.server 会加载nanollava-1.5-8bit

[eruca]

System Info / 系統信息

M1 macstudio 32G Sequoia 15.7.3

Who can help? / 谁可以帮助到您？

(.venv-mlx) ➜ glm-ocr mlx_vlm.server --trust-remote-code
INFO: Will watch for changes in these directories: ['/Users/nick/Codings/glm-ocr']
INFO: Uvicorn running on http://0.0.0.0:8080 (Press CTRL+C to quit)
INFO: Started reloader process [27549] using StatReload
INFO: Started server process [27570]
INFO: Waiting for application startup.
INFO: Application startup complete.
Loading model from: mlx-community/nanoLLaVA-1.5-8bit
Downloading (incomplete total...): 0%| | 0.00/1.12G [00:00<?, ?B/s]^CCancellation requested; stopping current tasks. | 0/11 [00:00<?, ?it/s]
Fetching 11 files: 36%|████████████████████████████████████████████████▋ | 4/11 [00:02<00:04, 1.52it/s]
ERROR: Exception in ASGI application

Information / 问题信息

• The official example scripts / 官方的示例脚本
• My own modified scripts / 我自己修改的脚本和任务

Reproduction / 复现过程

是严格按照m1安装教程进行的

Expected behavior / 期待表现

正常进行

[JaredforReal]

@TeHelado WDYT?

[kafkaliu]

Because mlx_vlm.server will load this model by default, see the attached images. Currently, this command does not have parameters to specify the model.

[eruca]

要怎么做呢，这个模型不是目标模型啊？

[kafkaliu]

When performing OCR, it will use the model configured in config.yaml.

[eruca]

是的，我的配置严格按照指令来完成了，所以config.yaml也是正常设置成 model: mlx-community/GLM-OCR-bf16
我检查了这个配置是正确的，而且还是出现了上面下载mlx-community/nanoLLaVA-1.5-8bit

[kafkaliu]

I mean, just let it download mlx-community/nanoLLaVA-1.5-8bit, and it should run normally once it's done.