Struct Active 

struct Active<'a> {

    prepared: PreparedRequest,
    phase: Phase,
    checkpoint: Option<PrefixState>,
    reservation: usize,
    ticket: Arc<Ticket>,
    session: Option<Turn>,
    output: Output,
    text_decoder: Box<dyn FnMut(u32) -> Result<Option<String>> + 'a>,
    text: GeneratedText,
    usage: UsageStats,
    config_generation: u64,
    response_bytes: usize,
    last_chunk: Option<(usize, f64)>,
}

Fields

prepared: PreparedRequest

phase: Phase

checkpoint: Option<PrefixState>

reservation: usize

ticket: Arc<Ticket>

session: Option<Turn>

output: Output

text_decoder: Box<dyn FnMut(u32) -> Result<Option<String>> + 'a>

text: GeneratedText

usage: UsageStats

config_generation: u64

response_bytes: usize

last_chunk: Option<(usize, f64)>

Tokens and elapsed seconds from the last step, if eligible for pacing.

Implementations

impl Active<'_>

fn stats(&self) -> ActiveRequest

fn phase_name(&self) -> &'static str

fn generated(&self) -> &[u32]

fn report(&self, gpu: &Gpu<'_>)

Per-request throughput and context/KV-cache fullness, reported to the server console when the request completes.

fn step( &mut self, gpu: &mut Gpu<'_>, cache: &mut PrefixCache, tok: &ChatTokenizer, state: &State, quantum: usize, ) -> Result<bool>

fn finish(self, tok: &ChatTokenizer) -> Result<()>

fn finish_response(&mut self, tok: &ChatTokenizer) -> Result<()>