continual-pretraining
20260703_continual-pretraining-gpt2medium-step34000-nodecayStacking_and_Continual_Pretraining_1024dit_fft_pretraining_v2_lerobot30_seed1000dit_fft_pretraining_v1_lerobot30_seed1000Stacking_and_Continual_Pretraining_1204continual-pretraining-checkpointdit_fft_pretraining_v1_lerobot30_seed1000_testStacking_and_Continual_Pretraining_43
wc-vqa-continual-pretraining
garrykuwanto/wc-vqa-continual-pretraining
English subset of worldcuisines/vqa-v1.1
Task 1 (dish name prediction), packaged for continual pretraining / SFT of
small vision-language models (specifically SmolVLM2-256M).
Splits
split
rows
source
train
27000
task1 / train / lang=en
validation
300
task1 / test_small / lang=en
test
1500
task1 / test_large / lang=en
Schema
field
type
notes
image
Image
bytes from upstream images.tar.gz… See the full description on the dataset page: https://huggingface.co/datasets/garrykuwanto/wc-vqa-continual-pretraining.LLM_Continual_Pretraining_Datajapanese-dedupContinual-Pre-training-Supervised-Datasetcontinual-pretraining