karpathy/neuraltalk2

Only 100 captions saved to vis.json

j4ysen opened this issue · 0 comments

Hi
Upon evaluation of eval.lua, with -num_images provided as a value of -1 or a number more than 100, only 100 captions are saved to vis.json. I can see from the terminal that captions are in fact generated, but not saved. The images are copied to the vis/imgs folder.