mirror of
https://github.com/BerriAI/litellm.git
synced 2025-04-25 18:54:30 +00:00
Litellm dev 11 21 2024 (#6837)
* Fix Vertex AI function calling invoke: use JSON format instead of protobuf text format. (#6702) * test: test tool_call conversion when arguments is empty dict Fixes https://github.com/BerriAI/litellm/issues/6833 * fix(openai_like/handler.py): return more descriptive error message Fixes https://github.com/BerriAI/litellm/issues/6812 * test: skip overloaded model * docs(anthropic.md): update anthropic docs to show how to route to any new model * feat(groq/): fake stream when 'response_format' param is passed Groq doesn't support streaming when response_format is set * feat(groq/): add response_format support for groq Closes https://github.com/BerriAI/litellm/issues/6845 * fix(o1_handler.py): remove fake streaming for o1 Closes https://github.com/BerriAI/litellm/issues/6801 * build(model_prices_and_context_window.json): add groq llama3.2b model pricing Closes https://github.com/BerriAI/litellm/issues/6807 * fix(utils.py): fix handling ollama response format param Fixes https://github.com/BerriAI/litellm/issues/6848#issuecomment-2491215485 * docs(sidebars.js): refactor chat endpoint placement * fix: fix linting errors * test: fix test * test: fix test * fix(openai_like/handler): handle max retries * fix(streaming_handler.py): fix streaming check for openai-compatible providers * test: update test * test: correctly handle model is overloaded error * test: update test * test: fix test * test: mark flaky test --------- Co-authored-by: Guowang Li <Guowang@users.noreply.github.com>
This commit is contained in:
Krish Dholakia
GitHub
parent
9ef254ff35
commit
4eca6ede4e
31 changed files with 747 additions and 403 deletions
|
|
@ -1739,15 +1739,15 @@ def supports_response_schema(model: str, custom_llm_provider: Optional[str]) ->
|
|||
|
||||
Does not raise error. Defaults to 'False'. Outputs logging.error.
|
||||
"""
|
||||
## GET LLM PROVIDER ##
|
||||
model, custom_llm_provider, _, _ = get_llm_provider(
|
||||
model=model, custom_llm_provider=custom_llm_provider
|
||||
)
|
||||
|
||||
if custom_llm_provider == "predibase": # predibase supports this globally
|
||||
return True
|
||||
|
||||
try:
|
||||
## GET LLM PROVIDER ##
|
||||
model, custom_llm_provider, _, _ = get_llm_provider(
|
||||
model=model, custom_llm_provider=custom_llm_provider
|
||||
)
|
||||
|
||||
if custom_llm_provider == "predibase": # predibase supports this globally
|
||||
return True
|
||||
|
||||
## GET MODEL INFO
|
||||
model_info = litellm.get_model_info(
|
||||
model=model, custom_llm_provider=custom_llm_provider
|
||||
|
|
@ -1755,12 +1755,17 @@ def supports_response_schema(model: str, custom_llm_provider: Optional[str]) ->
|
|||
|
||||
if model_info.get("supports_response_schema", False) is True:
|
||||
return True
|
||||
return False
|
||||
except Exception:
|
||||
verbose_logger.error(
|
||||
f"Model not supports response_schema. You passed model={model}, custom_llm_provider={custom_llm_provider}."
|
||||
## check if provider supports response schema globally
|
||||
supported_params = get_supported_openai_params(
|
||||
model=model,
|
||||
custom_llm_provider=custom_llm_provider,
|
||||
request_type="chat_completion",
|
||||
)
|
||||
return False
|
||||
if supported_params is not None and "response_schema" in supported_params:
|
||||
return True
|
||||
|
||||
return False
|
||||
|
||||
|
||||
def supports_function_calling(
|
||||
|
|
@ -2710,6 +2715,7 @@ def get_optional_params( # noqa: PLR0915
|
|||
non_default_params["response_format"] = type_to_response_format_param(
|
||||
response_format=non_default_params["response_format"]
|
||||
)
|
||||
|
||||
if "tools" in non_default_params and isinstance(
|
||||
non_default_params, list
|
||||
): # fixes https://github.com/BerriAI/litellm/issues/4933
|
||||
|
|
@ -3259,24 +3265,14 @@ def get_optional_params( # noqa: PLR0915
|
|||
)
|
||||
_check_valid_arg(supported_params=supported_params)
|
||||
|
||||
if max_tokens is not None:
|
||||
optional_params["num_predict"] = max_tokens
|
||||
if stream:
|
||||
optional_params["stream"] = stream
|
||||
if temperature is not None:
|
||||
optional_params["temperature"] = temperature
|
||||
if seed is not None:
|
||||
optional_params["seed"] = seed
|
||||
if top_p is not None:
|
||||
optional_params["top_p"] = top_p
|
||||
if frequency_penalty is not None:
|
||||
optional_params["repeat_penalty"] = frequency_penalty
|
||||
if stop is not None:
|
||||
optional_params["stop"] = stop
|
||||
if response_format is not None and response_format["type"] == "json_object":
|
||||
optional_params["format"] = "json"
|
||||
optional_params = litellm.OllamaConfig().map_openai_params(
|
||||
non_default_params=non_default_params,
|
||||
optional_params=optional_params,
|
||||
)
|
||||
elif custom_llm_provider == "ollama_chat":
|
||||
supported_params = litellm.OllamaChatConfig().get_supported_openai_params()
|
||||
supported_params = get_supported_openai_params(
|
||||
model=model, custom_llm_provider=custom_llm_provider
|
||||
)
|
||||
|
||||
_check_valid_arg(supported_params=supported_params)
|
||||
|
||||
|
|
@ -3494,24 +3490,16 @@ def get_optional_params( # noqa: PLR0915
|
|||
)
|
||||
_check_valid_arg(supported_params=supported_params)
|
||||
|
||||
if temperature is not None:
|
||||
optional_params["temperature"] = temperature
|
||||
if max_tokens is not None:
|
||||
optional_params["max_tokens"] = max_tokens
|
||||
if top_p is not None:
|
||||
optional_params["top_p"] = top_p
|
||||
if stream is not None:
|
||||
optional_params["stream"] = stream
|
||||
if stop is not None:
|
||||
optional_params["stop"] = stop
|
||||
if tools is not None:
|
||||
optional_params["tools"] = tools
|
||||
if tool_choice is not None:
|
||||
optional_params["tool_choice"] = tool_choice
|
||||
if response_format is not None:
|
||||
optional_params["response_format"] = response_format
|
||||
if seed is not None:
|
||||
optional_params["seed"] = seed
|
||||
optional_params = litellm.GroqChatConfig().map_openai_params(
|
||||
non_default_params=non_default_params,
|
||||
optional_params=optional_params,
|
||||
model=model,
|
||||
drop_params=(
|
||||
drop_params
|
||||
if drop_params is not None and isinstance(drop_params, bool)
|
||||
else False
|
||||
),
|
||||
)
|
||||
elif custom_llm_provider == "deepseek":
|
||||
supported_params = get_supported_openai_params(
|
||||
model=model, custom_llm_provider=custom_llm_provider
|
||||
|
|
@ -6178,5 +6166,7 @@ class ProviderConfigManager:
|
|||
return litellm.OpenAIO1Config()
|
||||
elif litellm.LlmProviders.DEEPSEEK == provider:
|
||||
return litellm.DeepSeekChatConfig()
|
||||
elif litellm.LlmProviders.GROQ == provider:
|
||||
return litellm.GroqChatConfig()
|
||||
|
||||
return OpenAIGPTConfig()
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue