Search results

S01 run.py Bug Report: int8 used for init when vocab size is 256 gives an overflow error

Status: Open.

#7 In rwitten/HighPerfLLMs2024;

· sshkhr opened on Apr 9, 2025

s4p44: all-reduce implementation

Status: Open.

#6 In rwitten/HighPerfLLMs2024;

· alanwaketan opened on Jul 16, 2024

s3p30: Our latency will go down linearly with our number of TP shards.

Status: Open.

#5 In rwitten/HighPerfLLMs2024;

· alanwaketan opened on Jul 15, 2024

s2p9: what's fissions?

Status: Open.

#4 In rwitten/HighPerfLLMs2024;

· alanwaketan opened on Jul 15, 2024

The formula for softmax seems wrong in s1p20

Status: Open.

#3 In rwitten/HighPerfLLMs2024;

· alanwaketan opened on Jul 15, 2024