mstar.engine

Contents

mstar.engine#

Functions

`apply_dynamo_config`()
`apply_torch_config`()
`recompile_limit`()

mstar.engine.apply_dynamo_config()[source]#

Return type:: None

mstar.engine.apply_torch_config()[source]#

Return type:: None

mstar.engine.recompile_limit()[source]#

Return type:: int

Modules

`accelerator_graph_backend`
`accelerator_graph_config`
`accelerator_graph_runner`
`fp32_precision`	Per-node float32 math precision, applied by the engine around a node's forward.
`node_dtype`	Which dtype each node runs in, resolved once at load.
`resources`	The resource layer's declaration surface.
`windowing`	Windowed (sliding-window) autoregressive generation support.