Download pretrained models for offline usage

[jgwinnup]

Hi,

I'm attempting to use Docling on a standalone network, working through what the collection of packages is expecting for models and other files, it looks like deepsearch-glm is trying to download models (/deepsearch_glm_v1/nlp/crf_pos_model_en.bin among them)

I'm calling deepsearch.glm.utils.load_pretrained_models on an internet-connected system in an attempt to download the above for offline use but have two questions:

• Where are these models normally cached?
• I set DEEPSEARCH_GLM_RESOURCES to a separate directory, but when I then call load_pretrained_models, it complains about config.json not existing. When the environment variable is not set, they appear to download and I get a python list back: ['part-of-speech', 'reference', ...

Thanks!