phoenix-oss/llama-stack
1
0
Fork 0
forked from phoenix-oss/llama-stack-mirror
Code Pull requests Releases Packages 2 Activity Actions

remove conflicting default for tool prompt format in chat completion (#742)

Browse source 
# What does this PR do?
We are setting a default value of json for tool prompt format, which
conflicts with llama 3.2/3.3 models since they use python list. This PR
changes the defaults to None and in the code, we infer default based on
the model.

Addresses: #695 

Tests:
❯ LLAMA_STACK_BASE_URL=http://localhost:5000 pytest -v
tests/client-sdk/inference/test_inference.py -k
"test_text_chat_completion"

 pytest llama_stack/providers/tests/inference/test_prompt_adapter.py
This commit is contained in:
Dinesh Yeduguru 2025-01-10 10:41:53 -08:00 • committed by GitHub
parent 24fa1adc2f
commit 8af6951106
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
21 changed files with 27 additions and 91 deletions
Download patch file Download diff file Expand all files Collapse all files

5
llama_stack/providers/remote/inference/bedrock/bedrock.py
View file

@ -10,7 +10,6 @@ from typing import AsyncGenerator, AsyncIterator, Dict, List, Optional, Union
from botocore.client import BaseClient
from llama_models.datatypes import CoreModelId
from llama_models.llama3.api.chat_format import ChatFormat
from llama_models.llama3.api.tokenizer import Tokenizer
from llama_stack.apis.common.content_types import InterleavedContent
@ -30,7 +29,6 @@ from llama_stack.apis.inference import (
)
from llama_stack.providers.remote.inference.bedrock.config import BedrockConfig
from llama_stack.providers.utils.bedrock.client import create_bedrock_client
from llama_stack.providers.utils.inference.model_registry import (
build_model_alias,
ModelRegistryHelper,
@ -47,7 +45,6 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
interleaved_content_as_str,
)
MODEL_ALIASES = [
build_model_alias(
"meta.llama3-1-8b-instruct-v1:0",
@ -101,7 +98,7 @@ class BedrockInferenceAdapter(ModelRegistryHelper, Inference):
response_format: Optional[ResponseFormat] = None,
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json,
tool_prompt_format: Optional[ToolPromptFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,
) -> Union[

7
llama_stack/providers/remote/inference/cerebras/cerebras.py
View file

@ -7,11 +7,8 @@
from typing import AsyncGenerator, List, Optional, Union
from cerebras.cloud.sdk import AsyncCerebras
from llama_models.datatypes import CoreModelId
from llama_models.llama3.api.chat_format import ChatFormat
from llama_models.llama3.api.tokenizer import Tokenizer
from llama_stack.apis.common.content_types import InterleavedContent
@ -29,7 +26,6 @@ from llama_stack.apis.inference import (
ToolDefinition,
ToolPromptFormat,
)
from llama_stack.providers.utils.inference.model_registry import (
build_model_alias,
ModelRegistryHelper,
@ -48,7 +44,6 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
from .config import CerebrasImplConfig
model_aliases = [
build_model_alias(
"llama3.1-8b",
@ -130,7 +125,7 @@ class CerebrasInferenceAdapter(ModelRegistryHelper, Inference):
sampling_params: Optional[SamplingParams] = SamplingParams(),
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json,
tool_prompt_format: Optional[ToolPromptFormat] = None,
response_format: Optional[ResponseFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,

7
llama_stack/providers/remote/inference/databricks/databricks.py
View file

@ -7,11 +7,8 @@
from typing import AsyncGenerator, List, Optional
from llama_models.datatypes import CoreModelId
from llama_models.llama3.api.chat_format import ChatFormat
from llama_models.llama3.api.tokenizer import Tokenizer
from openai import OpenAI
from llama_stack.apis.common.content_types import InterleavedContent
@ -28,7 +25,6 @@ from llama_stack.apis.inference import (
ToolDefinition,
ToolPromptFormat,
)
from llama_stack.providers.utils.inference.model_registry import (
build_model_alias,
ModelRegistryHelper,
@ -44,7 +40,6 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
from .config import DatabricksImplConfig
model_aliases = [
build_model_alias(
"databricks-meta-llama-3-1-70b-instruct",
@ -91,7 +86,7 @@ class DatabricksInferenceAdapter(ModelRegistryHelper, Inference):
response_format: Optional[ResponseFormat] = None,
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json,
tool_prompt_format: Optional[ToolPromptFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,
) -> AsyncGenerator:

4
llama_stack/providers/remote/inference/fireworks/fireworks.py
View file

@ -8,7 +8,6 @@ from typing import AsyncGenerator, List, Optional, Union
from fireworks.client import Fireworks
from llama_models.datatypes import CoreModelId
from llama_models.llama3.api.chat_format import ChatFormat
from llama_models.llama3.api.tokenizer import Tokenizer
@ -52,7 +51,6 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
from .config import FireworksImplConfig
MODEL_ALIASES = [
build_model_alias(
"fireworks/llama-v3p1-8b-instruct",
@ -198,7 +196,7 @@ class FireworksInferenceAdapter(
sampling_params: Optional[SamplingParams] = SamplingParams(),
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json,
tool_prompt_format: Optional[ToolPromptFormat] = None,
response_format: Optional[ResponseFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,

5
llama_stack/providers/remote/inference/groq/groq.py
View file

@ -33,6 +33,7 @@ from llama_stack.providers.utils.inference.model_registry import (
build_model_alias_with_just_provider_model_id,
ModelRegistryHelper,
)
from .groq_utils import (
convert_chat_completion_request,
convert_chat_completion_response,
@ -94,9 +95,7 @@ class GroqInferenceAdapter(Inference, ModelRegistryHelper, NeedsRequestProviderD
response_format: Optional[ResponseFormat] = None,
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[
ToolPromptFormat
] = None, # API default is ToolPromptFormat.json, we default to None to detect user input
tool_prompt_format: Optional[ToolPromptFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,
) -> Union[

4
llama_stack/providers/remote/inference/nvidia/nvidia.py
View file

@ -175,9 +175,7 @@ class NVIDIAInferenceAdapter(Inference, ModelRegistryHelper):
response_format: Optional[ResponseFormat] = None,
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[
ToolPromptFormat
] = None, # API default is ToolPromptFormat.json, we default to None to detect user input
tool_prompt_format: Optional[ToolPromptFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,
) -> Union[

4
llama_stack/providers/remote/inference/ollama/ollama.py
View file

@ -9,7 +9,6 @@ from typing import AsyncGenerator, List, Optional, Union
import httpx
from llama_models.datatypes import CoreModelId
from llama_models.llama3.api.chat_format import ChatFormat
from llama_models.llama3.api.tokenizer import Tokenizer
from ollama import AsyncClient
@ -35,7 +34,6 @@ from llama_stack.apis.inference import (
)
from llama_stack.apis.models import Model, ModelType
from llama_stack.providers.datatypes import ModelsProtocolPrivate
from llama_stack.providers.utils.inference.model_registry import (
build_model_alias,
build_model_alias_with_just_provider_model_id,
@ -222,7 +220,7 @@ class OllamaInferenceAdapter(Inference, ModelsProtocolPrivate):
response_format: Optional[ResponseFormat] = None,
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json,
tool_prompt_format: Optional[ToolPromptFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,
) -> AsyncGenerator:

4
llama_stack/providers/remote/inference/tgi/tgi.py
View file

@ -30,13 +30,11 @@ from llama_stack.apis.inference import (
ToolPromptFormat,
)
from llama_stack.apis.models import Model
from llama_stack.providers.datatypes import ModelsProtocolPrivate
from llama_stack.providers.utils.inference.model_registry import (
build_model_alias,
ModelRegistryHelper,
)
from llama_stack.providers.utils.inference.openai_compat import (
get_sampling_options,
OpenAICompatCompletionChoice,
@ -205,7 +203,7 @@ class _HfAdapter(Inference, ModelsProtocolPrivate):
sampling_params: Optional[SamplingParams] = SamplingParams(),
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json,
tool_prompt_format: Optional[ToolPromptFormat] = None,
response_format: Optional[ResponseFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,

2
llama_stack/providers/remote/inference/together/together.py
View file

@ -184,7 +184,7 @@ class TogetherInferenceAdapter(
sampling_params: Optional[SamplingParams] = SamplingParams(),
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json,
tool_prompt_format: Optional[ToolPromptFormat] = None,
response_format: Optional[ResponseFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,

5
llama_stack/providers/remote/inference/vllm/vllm.py
View file

@ -10,7 +10,6 @@ from typing import AsyncGenerator, List, Optional, Union
from llama_models.llama3.api.chat_format import ChatFormat
from llama_models.llama3.api.tokenizer import Tokenizer
from llama_models.sku_list import all_registered_models
from openai import OpenAI
from llama_stack.apis.common.content_types import InterleavedContent
@ -33,7 +32,6 @@ from llama_stack.apis.inference import (
)
from llama_stack.apis.models import Model, ModelType
from llama_stack.providers.datatypes import ModelsProtocolPrivate
from llama_stack.providers.utils.inference.model_registry import (
build_model_alias,
ModelRegistryHelper,
@ -54,7 +52,6 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
from .config import VLLMInferenceAdapterConfig
log = logging.getLogger(__name__)
@ -105,7 +102,7 @@ class VLLMInferenceAdapter(Inference, ModelsProtocolPrivate):
response_format: Optional[ResponseFormat] = None,
tools: Optional[List[ToolDefinition]] = None,
tool_choice: Optional[ToolChoice] = ToolChoice.auto,
tool_prompt_format: Optional[ToolPromptFormat] = ToolPromptFormat.json,
tool_prompt_format: Optional[ToolPromptFormat] = None,
stream: Optional[bool] = False,
logprobs: Optional[LogProbConfig] = None,
) -> AsyncGenerator: