Caption analysis returns empty output despite successful execution - GeoAI Moondream Analysis Result

[riteta-rimon]

After following all the steps provided in the installation guideline, running the caption analysis completes without errors but returns only an empty/"caption" string with no actual generated text.

[giswqs]

Did other functionality (e.g., detect, object) work?

[riteta-rimon]

I've also tested other functionality and these are the result:

1. Query
   Prompt: How many buildings in the image
   Result:
   Q: How many buildings in the image
   A

2. Detect
   Prompt: Trees
   Results:
   Detected 'trees': 0 object(s)

[giswqs]

That means the model is not working at all. Do you use CPU or GPU?

[riteta-rimon]

PyTorch + Cuda Verification

PyTorch: 2.7.1
CUDA available: True
GPU: NVIDIA GeForce GTX 1650 Ti

System Summary:

OS Name: Microsoft Windows 11 Home
Processor: AMD Ryzen 7 4800H with Radeon Graphics, 2900 Mhz, 8 Core(s), 16 Logical Processor(s)
System Type : x64-based PC
Installed Physical Memory (RAM):16.0 GB
Total Physical Memory: 15.4 GB
Available Physical Memory: 3.77 GB
Total Virtual Memory: 26.1GB
Available Virtual Memory: 3.57 GB