huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-11 04:14:52 +00:00
Code Issues Packages Projects Releases Wiki Activity

Make it clear that this value is only partially correct.

Browse Source
This commit is contained in:
Nicolas Patry 2024-01-11 17:56:16 +01:00
parent 8e0c538a18
commit ca04548080
1 changed files with 6 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
router/src/infer.rs
View File

@ -115,7 +115,7 @@ impl Infer {
// MPSC channel to communicate with the background batching task // MPSC channel to communicate with the background batching task
let (response_tx, response_rx) = mpsc::unbounded_channel(); let (response_tx, response_rx) = mpsc::unbounded_channel();
let input_length = valid_request.input_length; let _input_length = valid_request.input_length;
// Append the request to the queue // Append the request to the queue
self.queue.append(Entry { self.queue.append(Entry {
@ -202,7 +202,7 @@ impl Infer {
{ {
Ok(InferResponse { Ok(InferResponse {
prefill: result_prefill, prefill: result_prefill,
input_length, _input_length,
tokens: result_tokens, tokens: result_tokens,
generated_text, generated_text,
queued, queued,
@ -643,7 +643,10 @@ pub(crate) enum InferStreamResponse {
#[derive(Debug)] #[derive(Debug)]
pub(crate) struct InferResponse { pub(crate) struct InferResponse {
pub(crate) input_length: u32, /// input_length is the input as perceived by the rust tokenizer in the
/// validation pathway. It is redundant with prefill.len() but prefill
/// has data only if the user asked for it. This will always be filled.
pub(crate) _input_length: u32,
pub(crate) prefill: Vec<PrefillToken>, pub(crate) prefill: Vec<PrefillToken>,
pub(crate) tokens: Vec<Token>, pub(crate) tokens: Vec<Token>,
pub(crate) generated_text: GeneratedText, pub(crate) generated_text: GeneratedText,