AI/dify
1
0
Fork 0
mirror of https://git.mirrors.martin98.com/https://github.com/langgenius/dify.git synced 2025-08-14 01:15:56 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix/hit test tsne issue (#2581)

Browse Source 
Co-authored-by: jyong <jyong@dify.ai>
This commit is contained in:
Jyong 2024-02-27 17:30:52 +08:00 committed by GitHub
parent ac96d192a6
commit 920b2c2b40
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

BIN
api/celerybeat-schedule.db
View File

Binary file not shown.

2
api/core/features/annotation_reply.py
View File

@ -59,7 +59,7 @@ class AnnotationReplyFeature:
documents = vector.search_by_vector( documents = vector.search_by_vector(
query=query, query=query,
k=1, top_k=1,
score_threshold=score_threshold, score_threshold=score_threshold,
filter={ filter={
'group_id': [dataset.id] 'group_id': [dataset.id]

4
api/core/rag/datasource/retrieval_service.py
View File

@ -101,7 +101,7 @@ class RetrievalService:
documents = keyword.search( documents = keyword.search(
query, query,
k=top_k top_k=top_k
) )
all_documents.extend(documents) all_documents.extend(documents)
@ -121,7 +121,7 @@ class RetrievalService:
documents = vector.search_by_vector( documents = vector.search_by_vector(
query, query,
search_type='similarity_score_threshold', search_type='similarity_score_threshold',
k=top_k, top_k=top_k,
score_threshold=score_threshold, score_threshold=score_threshold,
filter={ filter={
'group_id': [dataset.id] 'group_id': [dataset.id]

5
api/services/hit_testing_service.py
View File

@ -133,8 +133,9 @@ class HitTestingService:
if embedding_length <= 1: if embedding_length <= 1:
return [{'x': 0, 'y': 0}] return [{'x': 0, 'y': 0}]
concatenate_data = np.array(embeddings).reshape(embedding_length, -1) noise = np.random.normal(0, 1e-4, np.array(embeddings).shape)
# concatenate_data = np.concatenate(embeddings) concatenate_data = np.array(embeddings) + noise
concatenate_data = concatenate_data.reshape(embedding_length, -1)
perplexity = embedding_length / 2 + 1 perplexity = embedding_length / 2 + 1
if perplexity >= embedding_length: if perplexity >= embedding_length: