Class MathBackendCPU

tfjs-backend-cpu/src/backend_cpu.ts:37–233

}

export class MathBackendCPU extends KernelBackend {
  public blockSize = 48;

  data: DataStorage<TensorData<DataType>>;
  private firstUse = true;
  private static nextDataId = 0;
  private nextDataId(): number {
    return MathBackendCPU.nextDataId++;
  }

  constructor() {
    super();
    this.data = new DataStorage(this, engine());
  }

  override write(
      values: backend_util.BackendValues, shape: number[],
      dtype: DataType): DataId {
    if (this.firstUse) {
      this.firstUse = false;
      if (env().get('IS_NODE')) {
        backend_util.warn(
            '\n============================\n' +
            'Hi, looks like you are running TensorFlow.js in ' +
            'Node.js. To speed things up dramatically, install our node ' +
            'backend, visit https://github.com/tensorflow/tfjs-node for more details. ' +
            '\n============================');
      }
    }
    const dataId = {id: this.nextDataId()};

    this.data.set(dataId, {values, dtype, refCount: 1});

    return dataId;
  }

  /**
   * Create a data bucket in cpu backend.
   * @param shape Shape of the `TensorInfo`.
   * @param dtype DType of the `TensorInfo`.
   * @param values The value of the `TensorInfo` stored as a flattened array.
   */
  makeTensorInfo(
      shape: number[], dtype: DataType,
      values?: backend_util.BackendValues|string[]): TensorInfo {
    let outId;
    if (dtype === 'string' && values != null && values.length > 0 &&
        util.isString(values[0])) {
      const encodedValues =
          (values as unknown as string[]).map(d => util.encodeString(d));

      outId = this.write(encodedValues, shape, dtype);
    } else {
      outId = this.write(values as TypedArray, shape, dtype);
    }

    return {dataId: outId, shape, dtype};
  }