phoenix/litellm-mirror
0
0
Fork 1
mirror of https://github.com/BerriAI/litellm.git synced 2025-04-26 03:04:13 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

(test) claude fixes

Browse source
This commit is contained in:
ishaan-jaff 2024-03-04 08:20:26 -08:00
parent d179ae376e
commit ba460e6838
3 changed files with 18 additions and 8 deletions
Download patch file Download diff file Expand all files Collapse all files

14
litellm/llms/anthropic.py
View file

@ -125,12 +125,19 @@ def completion(
format messages for anthropic format messages for anthropic
1. Anthropic supports roles like "user" and "assistant", (here litellm translates system-> assistant) 1. Anthropic supports roles like "user" and "assistant", (here litellm translates system-> assistant)
2. The first message always needs to be of role "user" 2. The first message always needs to be of role "user"
3. Each message must alternate between "user" and "assistant" (this is not addressed as now by litellm)
4. final assistant content cannot end with trailing whitespace (anthropic raises an error otherwise)
""" """
# 1. Anthropic only supports roles like "user" and "assistant" # 1. Anthropic only supports roles like "user" and "assistant"
for message in messages: for idx, message in enumerate(messages):
if message["role"] == "system": if message["role"] == "system":
message["role"] = "assistant" message["role"] = "assistant"
# if this is the final assistant message, remove trailing whitespace
# TODO: only do this if it's the final assistant message
if message["role"] == "assistant":
message["content"] = message["content"].strip()
# 2. The first message always needs to be of role "user" # 2. The first message always needs to be of role "user"
if len(messages) > 0: if len(messages) > 0:
if messages[0]["role"] != "user": if messages[0]["role"] != "user":
@ -211,6 +218,11 @@ def completion(
message=str(completion_response["error"]), message=str(completion_response["error"]),
status_code=response.status_code, status_code=response.status_code,
) )
elif len(completion_response["content"]) == 0:
raise AnthropicError(
message="No content in response",
status_code=response.status_code,
)
else: else:
text_content = completion_response["content"][0].get("text", None) text_content = completion_response["content"][0].get("text", None)
model_response.choices[0].message.content = text_content # type: ignore model_response.choices[0].message.content = text_content # type: ignore

3
litellm/tests/test_add_function_to_prompt.py
View file

@ -47,8 +47,9 @@ test_function_call_non_openai_model()
## case 2: add_function_to_prompt set ## case 2: add_function_to_prompt set
def test_function_call_non_openai_model_litellm_mod_set(): def test_function_call_non_openai_model_litellm_mod_set():
litellm.add_function_to_prompt = True litellm.add_function_to_prompt = True
litellm.set_verbose = True
try: try:
model = "claude-instant-1" model = "claude-instant-1.2"
messages = [{"role": "user", "content": "what's the weather in sf?"}] messages = [{"role": "user", "content": "what's the weather in sf?"}]
functions = [ functions = [
{ {

9
litellm/tests/test_completion.py
View file

@ -92,7 +92,6 @@ def test_completion_claude_3():
response = completion( response = completion(
model="anthropic/claude-3-opus-20240229", model="anthropic/claude-3-opus-20240229",
messages=messages, messages=messages,
max_tokens=10,
) )
# Add any assertions, here to check response args # Add any assertions, here to check response args
print(response) print(response)
@ -198,19 +197,17 @@ def test_completion_mistral_api_modified_input():
def test_completion_claude2_1(): def test_completion_claude2_1():
try: try:
litellm.set_verbose = True
print("claude2.1 test request") print("claude2.1 test request")
messages = [ messages = [
{ {
"role": "system", "role": "system",
"content": "Your goal is generate a joke on the topic user gives", "content": "Your goal is generate a joke on the topic user gives.",
}, },
{"role": "assistant", "content": "Hi, how can i assist you today?"},
{"role": "user", "content": "Generate a 3 liner joke for me"}, {"role": "user", "content": "Generate a 3 liner joke for me"},
] ]
# test without max tokens # test without max tokens
response = completion( response = completion(model="claude-2.1", messages=messages)
model="claude-2.1", messages=messages, request_timeout=10, max_tokens=10
)
# Add any assertions here to check the response # Add any assertions here to check the response
print(response) print(response)
print(response.usage) print(response.usage)