logo

Software Documentation (Version 1.5.0)

  • Software Release Notes
  • Documentation Updates

Cerebras Basics

  • How Cerebras Works
  • The Cerebras ML Workflow
  • Cerebras Execution Modes

Getting Started

  • Software Requirements and Dependencies
  • Checklist Before You Quickstart
  • TensorFlow Quickstart
  • PyTorch Quickstart
  • Weight Streaming Quickstart

Reference Samples

  • Cerebras Reference Implementations

General

  • Data Formats
  • Dynamic Loss Scaling
  • Performance Optimization Practices
  • Multi-Replica Data Parallel Training

Scripts and Templates

  • The run.py Template
  • Cerebras Command Line Pattern
  • The csrun_cpu Script
  • The csrun_wse Script
  • The cs_input_analyzer Script

Develop with TensorFlow

  • Workflow for TensorFlow on CS
  • Port TensorFlow to Cerebras
    • Keras Model to CerebrasEstimator
    • Using the CerebrasEstimator
    • The CerebrasEstimator Interface
    • TensorFlow Estimator to CerebrasEstimator
    • Limitations of the CerebrasEstimator
  • Prepare Input
    • Multi-worker Input Pipeline
    • Sharding For the CS system
    • The CS_AUTOTUNE
    • Optimize Input Function
  • TensorFlow Dynamic Loss Scaling
  • Compile on CPU
  • Train, Eval and Predict
  • Early Stopping
  • Multiple Models
    • Multi-model Inference
  • TensorFlow Variable Sequence Length
  • Using TensorBoard
  • Best Practices
  • Supported TensorFlow Layers
    • tf package
      • tf.ActivationLayer module
      • tf.AddLayer module
      • tf.AttentionLayer module
      • tf.BahdanauAttention module
      • tf.BaseLayer module
      • tf.Conv2DLayer module
      • tf.Conv2DTransposeLayer module
      • tf.CrossEntropyFromLogitsLayer module
      • tf.DenseLayer module
      • tf.DropoutLayer module
      • tf.EmbeddingLayer module
      • tf.FeedForwardNetwork module
      • tf.GraphAttentionLayer module
      • tf.GraphConvolutionLayer module
      • tf.Input module
      • tf.LSTMCell module
      • tf.LayerNormalizationLayer module
      • tf.MaxPool2DLayer module
      • tf.PoolerLayer module
      • tf.PoolerLayerV2 module
      • tf.PositionEmbeddingLayer module
      • tf.PrePostProcessWrapper module
      • tf.RNNEncoderBlock module
      • tf.RNNLayer module
      • tf.ReshapeLayer module
      • tf.SegmentEmbeddingLayer module
      • tf.SharedWeightsDenseLayer module
      • tf.SoftmaxLayer module
      • tf.SquaredErrorLayer module

Develop with PyTorch

  • Workflow for PyTorch on CS
  • Porting PyTorch Model to CS
  • PyTorch Runners
  • PyTorch Variable Tensor Shape
  • Limitations of PyTorch on Cerebras
  • Supported PyTorch Ops

Compiler Reports

  • Input Function Report
  • Compile Report
  • Incremental Compile

Extensions

  • Adding Custom Packages To cbcore Container
Theme by the Executable Book Project

Port TensorFlow to Cerebras

Port TensorFlow to Cerebras¶

This section describes how to port your TensorFlow code to Cerebras.

  • Keras Model to CerebrasEstimator
    • Using the KerasModelToCerebrasEstimator
  • Using the CerebrasEstimator
    • Calling the CerebrasEstimator
    • Callback input function
    • Callback model function
    • Setting the runtime configuration
  • The CerebrasEstimator Interface
    • Syntax
    • Arguments
    • Methods
  • TensorFlow Estimator to CerebrasEstimator
    • Step 1: Model function
    • Step 2: Input function
    • Step 3: Use CerebrasEstimator
    • Step 4: Edit RunConfig
    • Step 5: Ensure mixed precision
  • Limitations of the CerebrasEstimator
    • Model function limitations
    • Input function differences
    • Input function limitations
    • Config differences
    • Compilation differences

previous

Workflow for TensorFlow on CS

next

Keras Model to CerebrasEstimator

© Copyright 2022, Cerebras Systems.
Last updated on Apr 11, 2023, 11:24:35 PM.