Running training jobs across multiple nodes scales really well. A common assumption is that scale inevitably means slowdowns: more GPUs means more synchronization overhead, especially with multiple nodes communicating across a network. But we observed that the performance penalty isn’t as harsh as what you might think. Instead, we found near-linear strong scaling: fixing the global batch size and training on more GPUs led to proportional increases in training throughput. On a 1.3B parameter model, 4 nodes means a 3.9x gain over one node. On 16 nodes, it’s 14.4x. This is largely thanks to the super fast interconnects that major cloud providers have built in: @awscloud EC2 P4d instances provide 400 Gbps networking bandwidth, @Azure provides 1600 Gbps, and @OraclePaaS provides 800 Gbps.
Recent articles
- Weeknotes: datasette-enrichments, datasette-comments, sqlite-chronicle - 8th December 2023
- Datasette Enrichments: a new plugin framework for augmenting your data - 1st December 2023
- llamafile is the new best way to run a LLM on your own computer - 29th November 2023
- Prompt injection explained, November 2023 edition - 27th November 2023
- I'm on the Newsroom Robots podcast, with thoughts on the OpenAI board - 25th November 2023
- Weeknotes: DevDay, GitHub Universe, OpenAI chaos - 22nd November 2023