Struct TorchOpEvent

pub struct TorchOpEvent { /* private fields */ }

A pytorch operation

Implementations

impl TorchOpEvent

pub fn new( op: String, done_tx: OncePortRef<()>, mailbox: Mailbox, args_string: String, kwargs_string: String, worker_actor_id: ActorId, ) -> Box<Self>

Creates a new TorchOpEvent.

Trait Implementations

impl Debug for TorchOpEvent

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Event for TorchOpEvent

fn handle<'life0, 'async_trait>( &'life0 mut self, ) -> Pin<Box<dyn Future<Output = Result<(), SimNetError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait,

This is the method that will be called when the simulator fires the event at a particular time instant. Examples: For messages, it will be delivering the message to the dst’s receiver queue. For a proc spawn, it will be creating the proc object and instantiating it. For any event that manipulates the network (like adding/removing nodes etc.) implement handle_network().

fn handle_network<'life0, 'life1, 'async_trait>( &'life0 mut self, _simnet: &'life1 SimNet, ) -> Pin<Box<dyn Future<Output = Result<(), SimNetError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait,

This is the method that will be called when the simulator fires the event Unless you need to make changes to the network, you do not have to implement this. Only implement handle() method for all non-simnet requirements.

fn duration(&self) -> Duration

The latency of the event. This could be network latency, induced latency (sleep), or GPU work latency.

fn summary(&self) -> String

A user-friendly summary of the event