huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-06-10 03:12:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix: run pre-commit

Browse Source
This commit is contained in:
drbh 2024-05-21 21:31:26 +00:00
parent 9b08e4ab32
commit baf2adfb69
3 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
clients/python/text_generation/client.py
View File

@ -982,4 +982,4 @@ class AsyncClient:
except ValidationError: except ValidationError:
# If we failed to parse the payload, then it is an error payload # If we failed to parse the payload, then it is an error payload
raise parse_error(resp.status, json_payload) raise parse_error(resp.status, json_payload)
yield response yield response

2
clients/python/text_generation/types.py
View File

@ -453,4 +453,4 @@ class StreamResponse(BaseModel):
# Inference API currently deployed model # Inference API currently deployed model
class DeployedModel(BaseModel): class DeployedModel(BaseModel):
model_id: str model_id: str
sha: str sha: str

6
router/src/server.rs
View File

@ -597,7 +597,7 @@ async fn completions(
let span = tracing::Span::current(); let span = tracing::Span::current();
metrics::increment_counter!("tgi_request_count"); metrics::increment_counter!("tgi_request_count");
let CompletionRequest { let CompletionRequest {
max_tokens, max_tokens,
seed, seed,
stop, stop,
@ -642,13 +642,13 @@ async fn completions(
} }
let generate_requests: Vec<GenerateRequest> = req let generate_requests: Vec<GenerateRequest> = req
.prompt .prompt
.iter() .iter()
.map(|prompt| GenerateRequest { .map(|prompt| GenerateRequest {
inputs: prompt.to_string(), inputs: prompt.to_string(),
parameters: GenerateParameters { parameters: GenerateParameters {
best_of: None, best_of: None,
temperature: temperature, temperature,
repetition_penalty: req.repetition_penalty, repetition_penalty: req.repetition_penalty,
frequency_penalty: req.frequency_penalty, frequency_penalty: req.frequency_penalty,
top_k: None, top_k: None,