Skip to content

NPU needs to be initialized when starting a new process #2843

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merrymercy merged 2 commits into from
Dec 28, 2023
+4 −1
Merged

NPU needs to be initialized when starting a new process #2843

Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ed8069e
NPU needs to be initialized when starting a new process
jq460494839 Dec 21, 2023
beab1da
format
jq460494839 Dec 21, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Next Next commit
NPU needs to be initialized when starting a new process
  • Loading branch information
@jq460494839
jq460494839 committed Dec 21, 2023
commit ed8069edade958e68e02ce9aece06bef6567d5a4
8 changes: 5 additions & 3 deletions fastchat/serve/model_worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,7 +31,6 @@
str_to_torch_dtype,
)


worker_id = str(uuid.uuid4())[:8]
logger = build_logger("model_worker", f"model_worker_{worker_id}.log")

Expand Down Expand Up @@ -101,6 +100,9 @@ def __init__(
self.init_heart_beat()

def generate_stream_gate(self, params):
if self.device == "npu":
import torch_npu
torch_npu.npu.set_device("npu:0")
self.call_ct += 1

try:
Expand Down Expand Up @@ -216,8 +218,8 @@ def get_embeddings(self, params):
all_embeddings = []
all_token_num = 0
for i in range(0, input_ids.size(1), self.context_len):
chunk_input_ids = input_ids[:, i : i + self.context_len]
chunk_attention_mask = attention_mask[:, i : i + self.context_len]
chunk_input_ids = input_ids[:, i: i + self.context_len]
chunk_attention_mask = attention_mask[:, i: i + self.context_len]

chunk_embeddings, token_num = self.__process_embed_chunk(
chunk_input_ids, chunk_attention_mask, **model_type_dict
Expand Down