Cannot connect to customize ML backend(sometimes) but all thing seems ok

**Describe the bug**

1. I have deployed LS, LS-ml and minio storage in a single docker-compose file with same network(listed in [1]):
2. With all things are well configured(including labeling configuration), then I goto **Setting** -> **Machine Learning**, the default model I specified in command line turned out **'Error'**, but I am sure that sometimes it is ok.
3. I clicked 'Edit' to modified the model backend, but didn't pass the verification


4. I made following checks:
- network connection form LS to LS-ml-backend, seems ok:
- LS-ml-backend log checking, seems ok:
- checking `logs/uwsgi.log`, not found.

Is there any clue?

**To Reproduce**
see [Describe the bug]

**Expected behavior**
1. backend was correctly setup.
2. (or) information with stacktrace was provided to debug the error.

**Screenshots**
see [Describe the bug]

**Environment (please complete the following information):**
- OS: MacOS 13.4 (22F66)
- Label Studio Version: docker heartexlabs/label-studio:1.8.1
- label-studio-ml: 1.0.9
- Werkzeug: 2.0.2
- docker version:
- docker Desktop 4.22.0 (117440)
- Engine: 24.0.5
- Compose: v2.20.2-desktop.1

**Additional context**

1. used docker-compose.yaml
```yaml
version: "3.2"

services:
labeling:
container_name: labeling_container
image: heartexlabs/label-studio:1.8.1
ports:
- 8080:8080
depends_on:
- modeling
volumes:
- ${DATA:-./data}:/label-studio/data
- ./labeled:/label-studio/labeled
- ./labeling/configs:/label-studio/configs
command: >
bash -c "
label-studio init /labeled
--username admin@localhost
--password admin
--label-config=configs/${CONFIG:-labeling-config.xml}
--port 8080
--ml-backends http://modeling_container:9090 &&
label-studio start /labeled
--host labeling_container:8080
--port 8080
--log-level DEBUG"
restart: always
modeling:
container_name: modeling_container
build:
context: ./modeling
ports:
- 9090:9090
environment:
- ACCESS_KEY=some-access-key
- SECRET_KEY=some-secret-key
volumes:
- ./modeling/tools:/annotation_service_modeling/tools
command: >
bash -c "
label-studio-ml init modeling_backend
--script tools/${MODEL:-model.py}
--force &&
label-studio-ml start modeling_backend
--debug
--port 9090"
restart: always

storage:
image: quay.io/minio/minio:RELEASE.2023-08-09T23-30-22Z
container_name: storage_container
ports:
- 9000:9000
- 9001:9001
volumes:
- "./storage/data1:/data1"
command: server --console-address ":9001" /data1
environment:
- MINIO_ROOT_USER=******
- MINIO_ROOT_PASSWORD=******
- MINIO_ACCESS_KEY=******
- MINIO_SECRET_KEY=******
```