TorchDynamo Update 4: LazyTensor & nvFuser Experiments
|
|
2
|
1943
|
February 25, 2022
|
Prim::autocast_promote operation?
|
|
4
|
639
|
December 31, 2021
|
TorchDynamo Update: 1.48x geomean speedup on TorchBench CPU Inference
|
|
0
|
2371
|
November 12, 2021
|
Next Steps for PyTorch Compilers
|
|
9
|
4800
|
October 21, 2021
|
Loop_tool's lazy frontend - Experimenting with symbolic laziness
|
|
0
|
352
|
October 11, 2021
|
torch-MLIR presentation from Google / nod.ai
|
|
1
|
570
|
October 8, 2021
|
Depthwise conv2d: An NNC Case Study
|
|
0
|
712
|
April 7, 2021
|
Symbolic Shape Inference
|
|
1
|
648
|
March 31, 2021
|
Summary of Issues Found in Adapting Models to use TorchScript
|
|
1
|
438
|
March 16, 2021
|
ExecutionPlan caching in ProfilingGraphExecutorImpl?
|
|
0
|
342
|
March 3, 2021
|
Adapting Models to use TorchScript and Getting them to Produce Fusions
|
|
15
|
2707
|
February 26, 2021
|
JIT scripting & Autocast
|
|
11
|
1553
|
February 23, 2021
|
Single-op fusion benchmarking
|
|
0
|
441
|
February 4, 2021
|
NNC Per-Operator Benchmarks (on CPU)
|
|
5
|
438
|
January 27, 2021
|
TorchScript usability
|
|
2
|
991
|
January 26, 2021
|
Eliminating Framework Overhead by Compiling a Model Directly
|
|
3
|
707
|
January 25, 2021
|
Understanding TorchScript Type System
|
|
2
|
782
|
January 25, 2021
|
Perf counters for fun and profit
|
|
0
|
390
|
January 23, 2021
|
Quick Intro to E-Graphs
|
|
0
|
516
|
January 22, 2021
|