For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Python module
max.nn.comm
Multicast pools
MulticastPool | Symmetric multicast pool backing NVSwitch (NVLS) allreduce. |
|---|
Expert parallelism
EPBatchManager | Batch manager for Expert Parallelism (EP). |
|---|---|
EPCommInitializer | Helper class for initializing buffers for Expert Parallelism (EP). |
EPConfig | Configuration for Expert Parallelism (EP) communication. |
calculate_ep_max_tokens_per_rank | Calculate the maximum number of tokens per rank for EP communication. |
|---|