Commit @34507832c38b82cadc804db50d3bf199e865c079

Y
List All
Feedback
- This Project
- All Projects

윤영준 2024-04-03

fixed embed function desc

@34507832c38b82cadc804db50d3bf199e865c079

ebafef9

3450783

document_load_embedding.py

--- document_load_embedding.py

+++ document_load_embedding.py


 )
 # print(vllm_instance.complete("hello?"))
 
-# max_length must be set to 512, or it will crash for high dimension multilingual models...... WHY DO YOU NOT MENTION THIS ON DOCUMENT!?
+# max_length must be set to embed weights output dense dimension, and there are problems with executing sparse one, or it will crash for high dimension multilingual models...... WHY DO YOU NOT MENTION THIS ON DOCUMENT!?
 # https://github.com/run-llama/llama_index/issues/9272
 # Why am I setting dimension manually instead of llama_index deciding it?
 # it's because their default parameters are fit to openai and that causes problems EVERYWHERE

Add a comment

Open 0
Closed 0

List

...	...	@@ -12,7 +12,7 @@
12	12	)
13	13	# print(vllm_instance.complete("hello?"))
14	14
15		-# max_length must be set to 512, or it will crash for high dimension multilingual models...... WHY DO YOU NOT MENTION THIS ON DOCUMENT!?
	15	+# max_length must be set to embed weights output dense dimension, and there are problems with executing sparse one, or it will crash for high dimension multilingual models...... WHY DO YOU NOT MENTION THIS ON DOCUMENT!?
16	16	# https://github.com/run-llama/llama_index/issues/9272
17	17	# Why am I setting dimension manually instead of llama_index deciding it?
18	18	# it's because their default parameters are fit to openai and that causes problems EVERYWHERE

Delete comment