Struct DistributedProfiler 

pub struct DistributedProfiler { /* private fields */ }

Distributed training profiler

Provides advanced profiling capabilities for distributed training including:

• Cross-node communication analysis
• Gradient synchronization profiling
• Load balancing metrics
• Communication bottleneck detection

Implementations

impl DistributedProfiler

pub fn new(config: DistributedProfilerConfig) -> Self

Create a new distributed profiler

Arguments

• config - Profiler configuration

Example

use trustformers_debug::{DistributedProfiler, DistributedProfilerConfig};

let config = DistributedProfilerConfig::default();
let profiler = DistributedProfiler::new(config);

pub fn register_node(&self, node_info: NodeInfo) -> Result<()>

Register a node in the cluster

Arguments

• node_info - Information about the node

pub fn record_communication(&self, event: CommunicationEvent) -> Result<()>

Record a communication event

Arguments

• event - Communication event to record

pub fn record_synchronization(&self, event: SynchronizationEvent) -> Result<()>

Record a synchronization event

Arguments

• event - Synchronization event to record

pub fn record_snapshot(&self, snapshot: NodePerformanceSnapshot) -> Result<()>

Record a performance snapshot for a node

Arguments

• snapshot - Performance snapshot

pub fn generate_report(&self) -> Result<DistributedProfilingReport>

Generate a comprehensive profiling report

Returns

Detailed profiling report with analysis and recommendations

pub fn export_json(&self, path: &Path) -> Result<()>

Export profiling data to JSON

Arguments

• path - Output file path

pub fn get_realtime_stats(&self) -> Result<RealtimeStats>

Get real-time statistics (for dashboards)

Returns

Current profiling statistics

Trait Implementations

impl Debug for DistributedProfiler

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

Formats the value using the given formatter.