Fix llava parallelism/fork bug (#315)
This commit is contained in:
@@ -73,9 +73,11 @@ class RuntimeEndpoint(BaseBackend):
|
|||||||
assert res.status_code == 200
|
assert res.status_code == 200
|
||||||
|
|
||||||
def commit_lazy_operations(self, s: StreamExecutor):
|
def commit_lazy_operations(self, s: StreamExecutor):
|
||||||
|
data = {"text": s.text_, "sampling_params": {"max_new_tokens": 0}}
|
||||||
|
self._add_images(s, data)
|
||||||
res = http_request(
|
res = http_request(
|
||||||
self.base_url + "/generate",
|
self.base_url + "/generate",
|
||||||
json={"text": s.text_, "sampling_params": {"max_new_tokens": 0}},
|
json=data,
|
||||||
auth_token=self.auth_token,
|
auth_token=self.auth_token,
|
||||||
api_key=self.api_key,
|
api_key=self.api_key,
|
||||||
verify=self.verify,
|
verify=self.verify,
|
||||||
|
|||||||
@@ -276,6 +276,7 @@ class StreamExecutor:
|
|||||||
exes[i].messages_ = list(self.messages_)
|
exes[i].messages_ = list(self.messages_)
|
||||||
exes[i].cur_role = self.cur_role
|
exes[i].cur_role = self.cur_role
|
||||||
exes[i].fork_start_text_pos = len(self.text_)
|
exes[i].fork_start_text_pos = len(self.text_)
|
||||||
|
exes[i].images_ = list(self.images_)
|
||||||
|
|
||||||
return exes
|
return exes
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user