klbr/khoj
1
0
Fork 0
mirror of https://github.com/khoaliber/khoj.git synced 2026-03-09 21:29:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

Use n_batch parameter to increase resource consumption on host machine (and implicitly engage GPU)

Browse Source
This commit is contained in:
sabaimran
2023-07-31 23:25:08 -07:00
parent e55e9a7b67
commit 8054bdc896
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/khoj/processor/conversation/gpt4all/chat_model.py
View File

@@ -58,7 +58,7 @@ def extract_questions_offline(
next_christmas_date=next_christmas_date,
)
message = system_prompt + example_questions
response = gpt4all_model.generate(message, max_tokens=200, top_k=2, temp=0)
response = gpt4all_model.generate(message, max_tokens=200, top_k=2, temp=0, n_batch=128)
# Extract, Clean Message from GPT's Response
try:
@@ -161,7 +161,7 @@ def llm_thread(g, messages: List[ChatMessage], model: GPT4All):
templated_system_message = prompts.system_prompt_llamav2.format(message=system_message.content)
templated_user_message = prompts.general_conversation_llamav2.format(query=user_message.content)
prompted_message = templated_system_message + chat_history + templated_user_message
response_iterator = model.generate(prompted_message, streaming=True, max_tokens=1000)
response_iterator = model.generate(prompted_message, streaming=True, max_tokens=1000, n_batch=256)
for response in response_iterator:
g.send(response)
g.close()