Evaluation in Transformers (#218)

* 2serious

* load/save

* fixing GPU

* added DEBUG

* needs BatchNorm or doesn't learn anything

* old file not needed

* added conv biases

* added extra/training.py and checkpoint

* assert in test only

* save

* padding

* num_classes

* checkpoint

* checkpoints for padding

* training was broken

* merge

* rotation augmentation

* more aug

* needs testing

* streamline augment, augment is fast thus bicubic

* tidying up

* transformer eval
This commit is contained in:
Marcel Bischoff
2020-12-28 09:24:51 -05:00
committed by GitHub
parent 65b07d2f4f
commit ffff98db78
2 changed files with 6 additions and 10 deletions

View File

@@ -87,19 +87,15 @@ class Transformer:
x = t(x)
x = x.reshape(shape=(-1, x.shape[-1])).dot(self.final).logsoftmax()
return x.reshape(shape=(bs, -1, x.shape[-1]))
from tinygrad.optim import Adam
if __name__ == "__main__":
model = Transformer(10, 6, 2, 128, 4)
#in1 = Tensor.zeros(20, 6, 128)
#ret = model.forward(in1)
#print(ret.shape)
X_train, Y_train, X_test, Y_test = make_dataset()
optim = Adam(get_parameters(model), lr=0.001)
train(model, X_train, Y_train, optim, 100)
train(model, X_train, Y_train, optim, 500)
evaluate(model, X_test, Y_test, num_classes=10)