Extend training after loading the model in fast-ai which was saved locally?

fastai
1

I want to further train my save model in fast-ai (which is a classifier for healthy and unhealthy potato leaves), I trained and model using (resenet34) in fast-ai, then saved using lerner.save(“classification_fastai_model_1”) . It was successfully saved and now I want to train more using new images for the same kind of classification.

from fastai import *
from fastai.vision import *
path = "PlantVillage2"

There are 2 subfolders in this folder one is for healthy and other is for unhealthy

data = ImageDataBunch.from_folder(path,
                                 ds_tfms=get_transforms(do_flip=False),
                                 bs=32,
                                 size=224,
                                 valid_pct=0.26)
data.show_batch(row=3, figsize=(8,9))
data.normalize(imagenet_stats)
learner = cnn_learner(data, models.resnet34, metrics=[accuracy, error_rate])
learner.load("classification_fastai_model_1")
learner.fit_one_cycle(3)

I am getting this error now; I have already converted all the images into RGB using Photoshop Batch processing.

RuntimeError: The size of tensor a (2) must match the size of tensor b (3) at non-singleton dimension 0

Any help from you will be appreciated, Thanks.

2

You should tell us which of the line threw an error and copy the whole stack trace here, as it would be easier to help you with this.

1 Like