style: rustfmt server engine files
This commit is contained in:
@@ -396,9 +396,12 @@ fn emit_token(tokenizer: &Tokenizer, seq: &mut Sequence, token_id: u32) {
|
|||||||
if tokenizer.eos_token_id() == Some(token_id) {
|
if tokenizer.eos_token_id() == Some(token_id) {
|
||||||
let tail = tokenizer.flush_decode_stream(&mut seq.decode_buffer);
|
let tail = tokenizer.flush_decode_stream(&mut seq.decode_buffer);
|
||||||
send_token_if_nonempty(seq, tail);
|
send_token_if_nonempty(seq, tail);
|
||||||
try_send_event(seq, GenerateEvent::Done {
|
try_send_event(
|
||||||
|
seq,
|
||||||
|
GenerateEvent::Done {
|
||||||
finish_reason: "stop".to_string(),
|
finish_reason: "stop".to_string(),
|
||||||
});
|
},
|
||||||
|
);
|
||||||
return;
|
return;
|
||||||
}
|
}
|
||||||
|
|
||||||
@@ -407,9 +410,12 @@ fn emit_token(tokenizer: &Tokenizer, seq: &mut Sequence, token_id: u32) {
|
|||||||
let tail = tokenizer.flush_decode_stream(&mut seq.decode_buffer);
|
let tail = tokenizer.flush_decode_stream(&mut seq.decode_buffer);
|
||||||
send_token_if_nonempty(seq, text);
|
send_token_if_nonempty(seq, text);
|
||||||
send_token_if_nonempty(seq, tail);
|
send_token_if_nonempty(seq, tail);
|
||||||
try_send_event(seq, GenerateEvent::Done {
|
try_send_event(
|
||||||
|
seq,
|
||||||
|
GenerateEvent::Done {
|
||||||
finish_reason: "length".to_string(),
|
finish_reason: "length".to_string(),
|
||||||
});
|
},
|
||||||
|
);
|
||||||
} else {
|
} else {
|
||||||
send_token_if_nonempty(seq, text);
|
send_token_if_nonempty(seq, text);
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -318,7 +318,12 @@ pub fn run_pp(
|
|||||||
/// Returns false if the send would block (client too slow) or the client is
|
/// Returns false if the send would block (client too slow) or the client is
|
||||||
/// gone — the caller stops generating so the coordinator thread is free to
|
/// gone — the caller stops generating so the coordinator thread is free to
|
||||||
/// admit the next request instead of blocking on one slow consumer.
|
/// admit the next request instead of blocking on one slow consumer.
|
||||||
fn emit_text(tokenizer: &Tokenizer, req: &GenerateRequest, token_id: u32, buf: &mut Vec<u8>) -> bool {
|
fn emit_text(
|
||||||
|
tokenizer: &Tokenizer,
|
||||||
|
req: &GenerateRequest,
|
||||||
|
token_id: u32,
|
||||||
|
buf: &mut Vec<u8>,
|
||||||
|
) -> bool {
|
||||||
if tokenizer.is_eos(token_id) {
|
if tokenizer.is_eos(token_id) {
|
||||||
return true;
|
return true;
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -346,7 +346,12 @@ pub fn run_tp(
|
|||||||
/// Returns false if the send would block (client too slow) or the client is
|
/// Returns false if the send would block (client too slow) or the client is
|
||||||
/// gone — the caller stops generating so the serial coordinator thread is free
|
/// gone — the caller stops generating so the serial coordinator thread is free
|
||||||
/// to admit the next request instead of blocking on one slow consumer.
|
/// to admit the next request instead of blocking on one slow consumer.
|
||||||
fn emit_text(tokenizer: &Tokenizer, req: &GenerateRequest, token_id: u32, buf: &mut Vec<u8>) -> bool {
|
fn emit_text(
|
||||||
|
tokenizer: &Tokenizer,
|
||||||
|
req: &GenerateRequest,
|
||||||
|
token_id: u32,
|
||||||
|
buf: &mut Vec<u8>,
|
||||||
|
) -> bool {
|
||||||
if tokenizer.is_eos(token_id) {
|
if tokenizer.is_eos(token_id) {
|
||||||
return true;
|
return true;
|
||||||
}
|
}
|
||||||
|
|||||||
Reference in New Issue
Block a user