torch

• torch.load and torch.save
• torch.gather
• DDP
  • Initialization
• TorchScript
  • doxygen doc
  • Hello
    • torch.jit.script as a decorator
    • torch.jit.script as a function
    • torchscript a module
    • trace a module
      • Export and ignore methods
  • Load in C++
  • ArrayRef
    • Constructors
      • Data members
      • Default constructed
      • From a single element
      • From an initializer list
      • Other types of constructors
  • ScalarType
    • Members
      • Some aliases
    • ScalarType to CPP type
    • CPP type to ScalarType
  • TypeMeta
    • Constructors
    • Operations with ScalarType
  • torch::Device
    • DeviceType
    • Device
  • TensorOptions
    • Constructors (not recommended)
    • Constructors (Recommended)
    • Methods
  • Tensor Creation
    • TensorDataContainer
    • From std::vector
    • From scalar
    • From initializer list
    • From ArrayRef
  • Tensor
    • Common methods
      • slice
      • topk
      • floor_divide
      • div
      • remainder
      • empty
      • stack
      • unbind
      • full
      • split
      • zeros
      • cat
      • division
      • default constructed
      • copy
      • default addmm
      • elementwise operation
      • torch.roll
      • torch.mean
      • torch.slice
      • torch.as_strided
      • torch.argmax
      • torch.index
      • torch.nn.functional.pad
      • torch.index_put
      • torch.nonzero
  • intrusive_ptr
  • optional
  • PackedSequence
    • pack_padded_sequence
  • ivalue
  • method
  • type
  • trace
  • Node
  • symbol
  • graph
  • passes
    • _jit_pass_fuse_add_relu
      • in_place relu_
    • _jit_pass_fuse_linear
  • ops
    • aten::add
    • aten::add_
    • aten::_add_relu
    • aten::add_relu_
    • aten::_add_relu_
    • aten::relu
  • Add extra files
• Logical operations
• Note
• Quantization
  • Internals
    • QScheme
    • PerTensorAffineQuantizer
  • torch.quantize_per_tensor
    • Compress ration
  • quantize_per_tensor_dynamic
  • torch.quantize_per_channel
  • Observer
  • Hello
  • References
    • Links
• android
  • References
• onnx
  • Install
    • API references
  • Simple OPs
    • linear
  • Hello
  • Multiple models
  • References
  • Quantization
    • TO READ
    • 2020-Integer quantization for deep learning inference: Principles and empirical evaluation
    • 2022-A survey of quantization methods for efficient neural network inference
    • 1998-Quantization
  • Quantization
  • TODOs
• nn.LSTM
• torchvision
• audograd
  • _saved_self, _saved_other, _saved_result, _saved_exponent
  • Example 1
  • Where is Mulbackward0 defined
• seed
• float16
  • IEEE 754 float16
  • bfloat16 (brain float16)
• Dynamo
• random
• compile
  • TODO
• benchmark
• cv2
• einsum