RE: LeoThread 2024-09-09 11:48

[taskmaster4450le]

1. Model complexity: Multimodal models can be more complex and require more parameters than text-only models. This can lead to increased computational requirements and training times.
2. Training data quality: Multimodal data can be more challenging to collect and preprocess than text data. This can lead to increased costs for data annotation, labeling, and cleaning.