swissgeol/lithology
Partial fine-tune vs google-bert/bert-base-multilingual-uncased. Partial change: 5% of the weights changed (attention, other), in 1 of 12 layers.
Tensors changed: Embeddings 0/3, Attention 8/96, other 6/50
Partial fine-tune vs google-bert/bert-base-multilingual-uncased. Partial change: 5% of the weights changed (attention, other), in 1 of 12 layers.
Tensors changed: Embeddings 0/3, Attention 8/96, other 6/50