🦦 Otter, a multi-modal model based on OpenFlamingo (open-sourced version of DeepMind's Flamingo), trained on MIMIC-IT and showcasing improved instruction-following and in-context learning ability.
From my understanding "luodian/OTTER-MPT7B-Init" is almost the same as original openflamingo_v2 model, expect for some special tokens.
However, when I load "luodian/OTTER-MPT7B-Init" and try the pipeline/demo/interactive/otter_image.py file with the model set to "luodian/OTTER-MPT7B-Init", it just outputs repeated "\<PAD>\<PAD>\<PAD>". Any thoughts on how to solve this? @Luodian @liuziwei7 @Jingkang50 @gordonhu608
From my understanding "luodian/OTTER-MPT7B-Init" is almost the same as original openflamingo_v2 model, expect for some special tokens.
However, when I load "luodian/OTTER-MPT7B-Init" and try the pipeline/demo/interactive/otter_image.py file with the model set to "luodian/OTTER-MPT7B-Init", it just outputs repeated "\<PAD>\<PAD>\<PAD>". Any thoughts on how to solve this? @Luodian @liuziwei7 @Jingkang50 @gordonhu608
Many thanks!