AI/ragflow
1
0
Fork 0
mirror of https://git.mirrors.martin98.com/https://github.com/infiniflow/ragflow.git synced 2025-08-12 06:28:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix Bedrock system prompt (#2062)

Browse Source 
### What problem does this PR solve?

Bugfix: usage of Bedrock models require the system prompt (for models
that support it) to be provided in the API in a different way, at least
that was my experience with it just today. This PR fixes it.


https://docs.aws.amazon.com/bedrock/latest/userguide/conversation-inference.html

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
This commit is contained in:
Ran Tavory 2024-08-23 06:44:37 +03:00 committed by GitHub
parent 89b05ad79f
commit 19396998eb
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 4 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
rag/llm/chat_model.py
View File

@ -667,8 +667,6 @@ class BedrockChat(Base):
def chat(self, system, history, gen_conf):
from botocore.exceptions import ClientError
if system:
history.insert(0, {"role": "system", "content": system})
for k in list(gen_conf.keys()):
if k not in ["temperature", "top_p", "max_tokens"]:
del gen_conf[k]
@ -688,7 +686,8 @@ class BedrockChat(Base):
response = self.client.converse(
modelId=self.model_name,
messages=history,
inferenceConfig=gen_conf
inferenceConfig=gen_conf,
system=[{"text": system}] if system else None,
)
# Extract and print the response text.
@ -700,8 +699,6 @@ class BedrockChat(Base):
def chat_streamly(self, system, history, gen_conf):
from botocore.exceptions import ClientError
if system:
history.insert(0, {"role": "system", "content": system})
for k in list(gen_conf.keys()):
if k not in ["temperature", "top_p", "max_tokens"]:
del gen_conf[k]
@ -720,7 +717,8 @@ class BedrockChat(Base):
response = self.client.converse(
modelId=self.model_name,
messages=history,
inferenceConfig=gen_conf
inferenceConfig=gen_conf,
system=[{"text": system}] if system else None,
)
ans = response["output"]["message"]["content"][0]["text"]
return ans, num_tokens_from_string(ans)