Implementation:LaurentMazare Tch rs Tensor Display

Knowledge Sources	LaurentMazare_Tch_rs
Domains	Tensor_Operations, Display
Last Updated	2026-02-08 00:00 GMT

Overview

Concrete tool for pretty-printing tensor values provided by the tch-rs library.

Description

The Tensor_Display module implements Rust's Debug and Display formatting traits for the Tensor type, mirroring PyTorch's Python-side tensor printing logic. It classifies tensor element types into BasicKind categories (Float, Int, Bool, Complex) and formats tensor contents with appropriate precision, summarization for large tensors, and shape/dtype annotations.

Usage

This formatting is invoked automatically when using println!("{:?}", tensor) or println!("{}", tensor) in Rust code. No explicit import is needed beyond having a Tensor value in scope.

Code Reference

Source Location

• Repository: LaurentMazare_Tch_rs
• File: src/tensor/display.rs
• Lines: 1-434

Signature

impl std::fmt::Debug for Tensor {
    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result { ... }
}

impl std::fmt::Display for Tensor {
    fn fmt(&self, f: &mut std::fmt::Formatter) -> std::fmt::Result { ... }
}

Import

use tch::Tensor; // Debug and Display are auto-available

I/O Contract

Inputs

Name	Type	Required	Description
self	&Tensor	Yes	The tensor to format
f	&mut Formatter	Yes	The output formatter

Outputs

Name	Type	Description
Result	fmt::Result	Success or formatting error

Usage Examples

use tch::Tensor;

let t = Tensor::from_slice(&[1.0f32, 2.0, 3.0]);
println!("{:?}", t);  // Debug format with shape info
println!("{}", t);     // Display format with values

let matrix = Tensor::from_slice(&[1, 2, 3, 4, 5, 6]).view([2, 3]);
println!("{}", matrix);
// Output:
//  1  2  3
//  4  5  6
// [ CPULongTensor{2,3} ]

Related Pages

• Principle:LaurentMazare_Tch_rs_Tensor_Pretty_Printing