chunked/nltk_chunking/_quantization/chunk_71.txt

This works for any model in any modality, as long as it supports loading with Accelerate and contains torch.nn.Linear layers.