mirror of
http://112.124.100.131/huang.ze/ebiz-dify-ai.git
synced 2025-12-09 02:46:52 +08:00
Fix: Wrong cutoff length lead to empty input in openai compatible embedding model. (#7133)
This commit is contained in:
@@ -72,7 +72,7 @@ class OllamaEmbeddingModel(TextEmbeddingModel):
|
||||
num_tokens = self._get_num_tokens_by_gpt2(text)
|
||||
|
||||
if num_tokens >= context_size:
|
||||
cutoff = int(len(text) * (np.floor(context_size / num_tokens)))
|
||||
cutoff = int(np.floor(len(text) * (context_size / num_tokens)))
|
||||
# if num tokens is larger than context length, only use the start
|
||||
inputs.append(text[0: cutoff])
|
||||
else:
|
||||
|
||||
@@ -76,7 +76,7 @@ class OAICompatEmbeddingModel(_CommonOAI_API_Compat, TextEmbeddingModel):
|
||||
num_tokens = self._get_num_tokens_by_gpt2(text)
|
||||
|
||||
if num_tokens >= context_size:
|
||||
cutoff = int(len(text) * (np.floor(context_size / num_tokens)))
|
||||
cutoff = int(np.floor(len(text) * (context_size / num_tokens)))
|
||||
# if num tokens is larger than context length, only use the start
|
||||
inputs.append(text[0: cutoff])
|
||||
else:
|
||||
|
||||
@@ -79,7 +79,7 @@ class OAICompatEmbeddingModel(_CommonOAI_API_Compat, TextEmbeddingModel):
|
||||
num_tokens = self._get_num_tokens_by_gpt2(text)
|
||||
|
||||
if num_tokens >= context_size:
|
||||
cutoff = int(len(text) * (np.floor(context_size / num_tokens)))
|
||||
cutoff = int(np.floor(len(text) * (context_size / num_tokens)))
|
||||
# if num tokens is larger than context length, only use the start
|
||||
inputs.append(text[0: cutoff])
|
||||
else:
|
||||
|
||||
Reference in New Issue
Block a user