7.11: TensorFlow.js Color Classifier: Animating Loss

Показать описание

Рекомендации по теме

Комментарии

I'm creating a match predictor for the FIRST Robotics Competition with using these tutorials! Love your videos sooo much! Always wanted to use TensorFlow JS, but I was always so intimidated, but you make it so much simpler to understand!

BradleyHarker

finally an empty comments i can thank you on. thank you from Kuwait. so you will have fans in every country.

Kubaizi

Poor Dan is gonna be so confused about all the memes in the comment section

reimarpb

Opened the video thinking that commenting "is this loss?" was a funny and original idea.
Guess i was wrong

cicciobombo

hey Dan, great job, keep at it with the tensorflow.js
Btw, (not really) was it you who made the website for the Nature of Code book?

iilugs

you could just do { onBatchEnd: tf.nextframe } or { onBatchEnd: () => tf.nextframe() } if nextframe takes args because the fact you can write an async function means that it's awaited on the tf side

aetheralldev

I have two doubts:
1. My animation is running without using any callbacks, tf.nextFrame() or async
2. Training becomes 2x slower if we add anything in the draw

dharmikjoshi

return await has strictly no effect
the whole async function turns into a promise on execution
good job using const by the way !

aetheralldev

Weird, code works untill I put it inside the train function. Then it stops working and tells me "Uncaught (in promise) TypeError: model.fit is not a function". I declared the xs and ys globally.

meetoobaa