Compiler collectives

Compiler collectives

Update: 2024-08-04
Share

Description

Compiler collectives are a PT2 feature where by compiler instances across multiple ranks use NCCL collectives to communicate information to other instances. This is used to ensure we consistently decide if inputs or static or dynamic across all ranks. See also PR at https://github.com/pytorch/pytorch/pull/130935

Comments 
loading
In Channel
Compiler collectives

Compiler collectives

2024-08-0416:33

Higher order operators

Higher order operators

2024-04-2117:10

CUDA graph trees

CUDA graph trees

2024-03-2420:50

Min-cut partitioner

Min-cut partitioner

2024-03-1715:56

AOTInductor

AOTInductor

2024-03-0217:30

Compiled autograd

Compiled autograd

2024-02-1918:07

PT2 extension points

PT2 extension points

2024-02-0515:54

Unsigned integers

Unsigned integers

2024-01-1713:07

Inductor - IR

Inductor - IR

2024-01-1618:00

Unbacked SymInts

Unbacked SymInts

2023-02-2121:31

torchdynamo

torchdynamo

2022-12-0625:35

PyTorch 2.0

PyTorch 2.0

2022-12-0417:51

History of functorch

History of functorch

2022-11-0719:10

loading
00:00
00:00
x

0.5x

0.8x

1.0x

1.25x

1.5x

2.0x

3.0x

Sleep Timer

Off

End of Episode

5 Minutes

10 Minutes

15 Minutes

30 Minutes

45 Minutes

60 Minutes

120 Minutes

Compiler collectives

Compiler collectives

PyTorch