estel/dify
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

fix: better gard nan value from numpy for issue #11827 (#11864)

Browse Source 
Signed-off-by: yihong0618 <zouzou0208@gmail.com>
This commit is contained in:
yihong
2024-12-20 09:28:32 +08:00
committed by GitHub
parent 95a7e50137
commit 463fbe2680
5 changed files with 18 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
api/core/model_runtime/model_providers/azure_openai/text_embedding/text_embedding.py
View File

@@ -92,7 +92,10 @@ class AzureOpenAITextEmbeddingModel(_CommonAzureOpenAI, TextEmbeddingModel):
average = embeddings_batch[0]
else:
average = np.average(_result, axis=0, weights=num_tokens_in_batch[i])
embeddings[i] = (average / np.linalg.norm(average)).tolist()
embedding = (average / np.linalg.norm(average)).tolist()
if np.isnan(embedding).any():
raise ValueError("Normalized embedding is nan please try again")
embeddings[i] = embedding
# calc usage
usage = self._calc_response_usage(model=model, credentials=credentials, tokens=used_tokens)

5
api/core/model_runtime/model_providers/cohere/text_embedding/text_embedding.py
View File

@@ -88,7 +88,10 @@ class CohereTextEmbeddingModel(TextEmbeddingModel):
average = embeddings_batch[0]
else:
average = np.average(_result, axis=0, weights=num_tokens_in_batch[i])
embeddings[i] = (average / np.linalg.norm(average)).tolist()
embedding = (average / np.linalg.norm(average)).tolist()
if np.isnan(embedding).any():
raise ValueError("Normalized embedding is nan please try again")
embeddings[i] = embedding
# calc usage
usage = self._calc_response_usage(model=model, credentials=credentials, tokens=used_tokens)

5
api/core/model_runtime/model_providers/openai/text_embedding/text_embedding.py
View File

@@ -97,7 +97,10 @@ class OpenAITextEmbeddingModel(_CommonOpenAI, TextEmbeddingModel):
average = embeddings_batch[0]
else:
average = np.average(_result, axis=0, weights=num_tokens_in_batch[i])
embeddings[i] = (average / np.linalg.norm(average)).tolist()
embedding = (average / np.linalg.norm(average)).tolist()
if np.isnan(embedding).any():
raise ValueError("Normalized embedding is nan please try again")
embeddings[i] = embedding
# calc usage
usage = self._calc_response_usage(model=model, credentials=credentials, tokens=used_tokens)

5
api/core/model_runtime/model_providers/upstage/text_embedding/text_embedding.py
View File

@@ -100,7 +100,10 @@ class UpstageTextEmbeddingModel(_CommonUpstage, TextEmbeddingModel):
average = embeddings_batch[0]
else:
average = np.average(_result, axis=0, weights=num_tokens_in_batch[i])
embeddings[i] = (average / np.linalg.norm(average)).tolist()
embedding = (average / np.linalg.norm(average)).tolist()
if np.isnan(embedding).any():
raise ValueError("Normalized embedding is nan please try again")
embeddings[i] = embedding
usage = self._calc_response_usage(model=model, credentials=credentials, tokens=used_tokens)