Inside TensorFlow: Functions, not sessions

[@johtib]

I'm still confused. Lets take a simple example, if I create a custom activation function that I pass to a Keras layer using the activation parameter, what would happen to it "graph-wise" and ought it be wrapped in tf.function for performance reasons? The reason this is hard to reason about is that I don't know what Keras does internally. Does it wrap whatever function calls my activation function in tf.function and thus it doesn't make a difference if I annotated my function?

[@kutilkol]

pyton is a program computre?

[@hxrayful]

Could you please give a tutorial for tensorflow grappler usage?

[@omkarenator]

Hey, can you guys do one on Grappler, please?

[@kutilkol]

I lov grapller!

[@alaindimabuyo]

5 min in the video and I dont get anything unless he said tensors I am so bad lol

[@vsigal]

RFC mentioned here : github.com/tensorflow/community/blob/master/rfcs/20180918-functions-not-sessions-20.md

[@aiwithr]

This is really interesting!

[@liveseribuhikmah]

Ok

[@tunnelsloth5948]

SAMSONO CHANNEL same

[@CryptoZZY]

Anyone do tensorflow for physics in gpu

[@kutilkol]

it's already on git

[@guitarheroprince123]

Ok