API: Fix a regression

This commit is contained in:
oobabooga 2025-05-16 13:02:27 -07:00
parent 470c822f44
commit e4d3f4449d

View file

@ -146,8 +146,9 @@ class LlamaServer:
pprint.PrettyPrinter(indent=4, sort_dicts=False).pprint(printable_payload) pprint.PrettyPrinter(indent=4, sort_dicts=False).pprint(printable_payload)
print() print()
# Make a direct request with streaming enabled using a context manager # Make the generation request
with self.session.post(url, json=payload, stream=True) as response: response = self.session.post(url, json=payload, stream=True)
try:
response.raise_for_status() # Raise an exception for HTTP errors response.raise_for_status() # Raise an exception for HTTP errors
full_text = "" full_text = ""
@ -184,6 +185,8 @@ class LlamaServer:
print(f"JSON decode error: {e}") print(f"JSON decode error: {e}")
print(f"Problematic line: {line}") print(f"Problematic line: {line}")
continue continue
finally:
response.close()
def generate(self, prompt, state): def generate(self, prompt, state):
output = "" output = ""