Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

ML/TF Meetup - TensorFlow Dev Summit Extended - Special Edition (#AperiTech)

191 views

Published on

https://www.meetup.com/it-IT/Machine-Learning-Data-Science-Meetup/events/259391928/

Published in: Technology
  • Be the first to comment

  • Be the first to like this

ML/TF Meetup - TensorFlow Dev Summit Extended - Special Edition (#AperiTech)

  1. 1. Meetup Special Edition TensorFlow Dev Summit Viewing Party 19 Marzo 2019, LUISS EnLabs
  2. 2. Agenda 19:00-19:15 Ingresso e Registrazione 19:15-19:30 Le principali novità del TF Dev Summit (Simone Scardapane & Norman Di Palo) 19:30-19:50 Viewing Party 19:50-20:15 In Codice Ratio – Replica! (Elena Nieddu) 20:15-21:00 News, Open Mic, Pizza, Birra & Networking!
  3. 3. TF in Retrospettiva 41,000,000 50,000+ 9,900+ 1,800+ downloads commits pull requests contributors
  4. 4. TF in Retrospettiva
  5. 5. TF oggi
  6. 6. TF 2.0 pip install -U --pre tensorflow
  7. 7. Keras for the win
  8. 8. Eager execution by default data = tf.data.TFRecordDataset(['file1', 'file2’]) data = data.map(parse_fn).batch(32) for row in data.take(3): print(row) >>> (<tf.Tensor: id=38, shape=(32, 28, 28), dtype=float64, numpy= array([[ 0.070588, …, 0.498039]])> ... model.fit(data, epochs=1)
  9. 9. DistributedStrategy strategy = tf.distribute.MirroredStrategy() with strategy.scope(): model = tf.keras.models.Sequential([ tf.keras.layers.Dense(64, input_shape=[10]), tf.keras.layers.Dense(64, activation='relu’), tf.keras.layers.Dense(10, activation='softmax’) ]) model.compile(optimizer='adam’, loss='categorical_crossentropy’, metrics=['accuracy'])
  10. 10. Swift For TensorFlow
  11. 11. Swift For TensorFlow import Python let np = Python.import("numpy") let x = np.array([[1, 2], [3, 4]]) let y = np.array([11, 12]) print(x.dot(y)) // [29 67]
  12. 12. Swift For TensorFlow import TensorFlow struct MyModel: Layer { var conv = Conv2D<Float>(filterShape: (5, 5, 3, 6)) var maxPool = MaxPool2D<Float>(poolSize: (2, 2), strides: (2, 2)) var flatten = Flatten<Float>() var dense = Dense<Float>(inputSize: 16 * 5 * 5, outputSize: 10) @differentiable func applied(to input: Tensor<Float>, in context: Context) -> Tensor<Float> { return input.sequenced(in: context, through: conv, maxPool, flatten, dense) } }
  13. 13. Altre novità • Autograph • TensorFlow Lite • TensorFlow.js 1.0 • TensorFlow datasets! (finalmente) • Novità in TensorBoard • Molte novità per le community (TF SIG) • …
  14. 14. Nuovi corsi
  15. 15. Se volete sperimentare https://iaml.it/blog/novita-tensorflow-2 https://colab.research.google.com/drive/14j1skr2TSh3g4oDaT2DAUxg08jkqBJoo
  16. 16. Viewing Party! https://www.youtube.com/watch?v=b5Rs1ToD9aI&t=1641s
  17. 17. Open Mic Open Mic!
  18. 18. Seguici! https://www.meetup.com/it-IT/Machine-Learning-Data-Science-Meetup/ https://www.facebook.com/groups/datascienceroma/ http://www.iaml.it/member https://www.linkedin.com/company/iaml/ https://www.facebook.com/machinelearningitalia/ https://www.slideshare.net/MeetupDataScienceRoma https://twitter.com/iaml_it

×