Struct PipelineSequence 

pub struct PipelineSequence {
    pub id: u64,
    pub phase: ExecutionPhase,
    pub input_tokens: Vec<TokenId>,
    pub generated_tokens: Vec<TokenId>,
    pub kv_cache: Option<Arc<dyn KvCacheHandle>>,
    pub sampling_params: SamplingParams,
    pub priority: Priority,
    pub rng: SamplingRng,
    pub prefill_start: Option<Instant>,
    pub first_token_time: Option<Instant>,
    pub completion_time: Option<Instant>,
}

Sequence being processed in the pipeline

Fields

id: u64

Unique sequence ID

phase: ExecutionPhase

Current phase

input_tokens: Vec<TokenId>

Input tokens (for prefill)

generated_tokens: Vec<TokenId>

Generated tokens

kv_cache: Option<Arc<dyn KvCacheHandle>>

KV cache handle

sampling_params: SamplingParams

Sampling parameters

priority: Priority

Priority

rng: SamplingRng

Random number generator

prefill_start: Option<Instant>

Prefill start time

first_token_time: Option<Instant>

First token time (TTFT)

completion_time: Option<Instant>

Total time to complete

Implementations

impl PipelineSequence

pub fn new( id: u64, input_tokens: Vec<TokenId>, sampling_params: SamplingParams, priority: Priority, ) -> Self

Create new sequence

pub fn total_tokens(&self) -> usize

Total tokens in sequence

pub fn should_stop(&self, vocab_size: usize) -> bool

Check if generation should stop

pub fn ttft_ms(&self) -> Option<u64>

Get time to first token in milliseconds

Trait Implementations

impl Clone for PipelineSequence

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PipelineSequence

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

Formats the value using the given formatter.