Megatron-LM: Nvidia's 8.3B Parameter GPT-2 Transformer

Posted by thats_legit on Tuesday August 13, 2019 @05:54PM
from the blog.exxactcorp.com dept.

( 4 comments | 9 points | on HN )

Re:Megatron-LM: Nvidia's 8.3B Parameter GPT-2 Transformer by ribeyes on Tuesday August 13, 2019 @06:40PM (#20690988)

Sounds compelling. Where can we try it out? We need to compare the results with GPT-2 to see how it performs.

[ Parent | on HN ]

Re:Megatron-LM: Nvidia's 8.3B Parameter GPT-2 Transformer by thats_legit on Tuesday August 13, 2019 @06:47PM (#20691048)

[ Parent | on HN ]

Re:Megatron-LM: Nvidia's 8.3B Parameter GPT-2 Transformer by the_good_guy on Tuesday August 13, 2019 @06:49PM (#20691054)

Pytorch implementation

[ Parent | on HN ]

Re:Megatron-LM: Nvidia's 8.3B Parameter GPT-2 Transformer by ptrenko on Wednesday August 14, 2019 @02:03AM (#20693067)

Jesus. Will anybody release a public model??

[ Parent | on HN ]