Spaces:
Running on Zero
Running on Zero
File size: 968 Bytes
6cf9641 2a071e1 cc3c1a2 2a071e1 6cf9641 cc3c1a2 2a071e1 cc3c1a2 6cf9641 2a071e1 6cf9641 2a071e1 6cf9641 2a071e1 6cf9641 2a071e1 cc3c1a2 6cf9641 2a071e1 6cf9641 2a071e1 6cf9641 2a071e1 6cf9641 | 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 | # Testing the local model
The Space inference path is:
```text
Custom frontend
-> queued Gradio backend
-> Nemotron OCR v2 for screenshot text
-> app/model_endpoint.py
-> MiniCPM5-1B through Transformers on ZeroGPU
```
Local endpoint tests use MiniCPM5-1B GGUF through `llama-cpp-python`.
## Fast checks
Run tests that do not load the model:
```powershell
python app.py --self-test
python -m unittest
```
Download the configured GGUF:
```powershell
python -m pip install -r requirements-local.txt
python app.py --download-model
```
Run a real text-generation contract test:
```powershell
python app.py --test-endpoint
```
The command fails unless the model returns all required fields:
- `risk_label`
- `simple_explanation`
- `red_flags`
- `safe_next_steps`
- `reply_draft`
The old Modal deployments and request scripts are intentionally preserved under
`experiments/` for comparison and reproducibility. They are not imported by the
application.
|