phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-08-05 10:13:05 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

revert some unintentional changes by copying source of truth to llama-models

Browse source
This commit is contained in:
Ashwin Bharambe 2025-04-07 11:00:48 -07:00
parent 53a8086e37
commit cfaf9e0e8b
9 changed files with 133 additions and 113 deletions
Download patch file Download diff file Expand all files Collapse all files

132
docs/_static/llama-stack-spec.html vendored
View file

@ -4163,70 +4163,80 @@
] ]
}, },
"arguments": { "arguments": {
"type": "object", "oneOf": [
"additionalProperties": { {
"oneOf": [ "type": "string"
{ },
"type": "string" {
}, "type": "object",
{ "additionalProperties": {
"type": "integer" "oneOf": [
}, {
{ "type": "string"
"type": "number" },
}, {
{ "type": "integer"
"type": "boolean" },
}, {
{ "type": "number"
"type": "null" },
}, {
{ "type": "boolean"
"type": "array", },
"items": { {
"oneOf": [ "type": "null"
{ },
"type": "string" {
}, "type": "array",
{ "items": {
"type": "integer" "oneOf": [
}, {
{ "type": "string"
"type": "number" },
}, {
{ "type": "integer"
"type": "boolean" },
}, {
{ "type": "number"
"type": "null" },
{
"type": "boolean"
},
{
"type": "null"
}
]
} }
] },
} {
}, "type": "object",
{ "additionalProperties": {
"type": "object", "oneOf": [
"additionalProperties": { {
"oneOf": [ "type": "string"
{ },
"type": "string" {
}, "type": "integer"
{ },
"type": "integer" {
}, "type": "number"
{ },
"type": "number" {
}, "type": "boolean"
{ },
"type": "boolean" {
}, "type": "null"
{ }
"type": "null" ]
} }
] }
} ]
} }
] }
} ]
},
"arguments_json": {
"type": "string"
} }
}, },
"additionalProperties": false, "additionalProperties": false,

52
docs/_static/llama-stack-spec.yaml vendored
View file

@ -2890,30 +2890,34 @@ components:
title: BuiltinTool title: BuiltinTool
- type: string - type: string
arguments: arguments:
type: object oneOf:
additionalProperties: - type: string
oneOf: - type: object
- type: string additionalProperties:
- type: integer oneOf:
- type: number - type: string
- type: boolean - type: integer
- type: 'null' - type: number
- type: array - type: boolean
items: - type: 'null'
oneOf: - type: array
- type: string items:
- type: integer oneOf:
- type: number - type: string
- type: boolean - type: integer
- type: 'null' - type: number
- type: object - type: boolean
additionalProperties: - type: 'null'
oneOf: - type: object
- type: string additionalProperties:
- type: integer oneOf:
- type: number - type: string
- type: boolean - type: integer
- type: 'null' - type: number
- type: boolean
- type: 'null'
arguments_json:
type: string
additionalProperties: false additionalProperties: false
required: required:
- call_id - call_id

9
llama_stack/models/llama/datatypes.py
View file

@ -38,7 +38,14 @@ RecursiveType = Union[Primitive, List[Primitive], Dict[str, Primitive]]
class ToolCall(BaseModel): class ToolCall(BaseModel):
call_id: str call_id: str
tool_name: Union[BuiltinTool, str] tool_name: Union[BuiltinTool, str]
arguments: Dict[str, RecursiveType] # Plan is to deprecate the Dict in favor of a JSON string
# that is parsed on the client side instead of trying to manage
# the recursive type here.
# Making this a union so that client side can start prepping for this change.
# Eventually, we will remove both the Dict and arguments_json field,
# and arguments will just be a str
arguments: Union[str, Dict[str, RecursiveType]]
arguments_json: Optional[str] = None
@field_validator("tool_name", mode="before") @field_validator("tool_name", mode="before")
@classmethod @classmethod

5
llama_stack/models/llama/llama3/quantization/__init__.py Normal file
View file

@ -0,0 +1,5 @@
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.

15
llama_stack/models/llama/llama4/chat_format.py
View file

@ -210,9 +210,12 @@ class ChatFormat:
content = ToolUtils.encode_tool_call(t, tool_prompt_format) content = ToolUtils.encode_tool_call(t, tool_prompt_format)
_process_content(content) _process_content(content)
# Tool calls and Tool Response messages should be eom
eom = False eom = False
if message.role == "assistant": if message.role == "assistant":
eom = message.stop_reason == StopReason.end_of_message eom = message.stop_reason == StopReason.end_of_message or message.tool_calls
elif message.role == "tool":
eom = True
tokens.append(self.tokenizer.special_tokens["<|eom|>" if eom else "<|eot|>"]) tokens.append(self.tokenizer.special_tokens["<|eom|>" if eom else "<|eot|>"])
return tokens, images return tokens, images
@ -247,6 +250,11 @@ class ChatFormat:
if content.startswith(header_str): if content.startswith(header_str):
content = content[len(header_str) :] content = content[len(header_str) :]
ipython = content.startswith("<|python_start|>")
if ipython:
content = content[len("<|python_start|>") :]
content = content.replace("<|python_end|>", "")
if content.endswith("<|eot|>"): if content.endswith("<|eot|>"):
content = content[: -len("<|eot|>")] content = content[: -len("<|eot|>")]
stop_reason = StopReason.end_of_turn stop_reason = StopReason.end_of_turn
@ -277,6 +285,11 @@ class ChatFormat:
} }
if tool_name in BuiltinTool.__members__: if tool_name in BuiltinTool.__members__:
tool_name = BuiltinTool[tool_name] tool_name = BuiltinTool[tool_name]
elif ipython:
tool_name = BuiltinTool.code_interpreter
tool_arguments = {
"code": content,
}
tool_calls = [] tool_calls = []
if tool_name is not None and tool_arguments is not None: if tool_name is not None and tool_arguments is not None:

16
llama_stack/models/llama/llama4/generation.py
View file

@ -4,22 +4,6 @@
# This source code is licensed under the terms described in the LICENSE file in # This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree. # the root directory of this source tree.
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# top-level folder for each specific model found within the models/ directory at
# the top-level of this source tree.
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
# Copyright (c) Meta Platforms, Inc. and affiliates.
# This software may be used and distributed in accordance with the terms of the Llama 3 Community License Agreement.
import codecs import codecs
import io import io
import json import json

5
llama_stack/models/llama/llama4/quantization/__init__.py Normal file
View file

@ -0,0 +1,5 @@
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.

10
llama_stack/models/llama/llama4/tokenizer.py
View file

@ -4,16 +4,6 @@
# This source code is licensed under the terms described in the LICENSE file in # This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree. # the root directory of this source tree.
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the terms described in the LICENSE file in
# top-level folder for each specific model found within the models/ directory at
# the top-level of this source tree.
# Copyright (c) Meta Platforms, Inc. and affiliates.
# This software may be used and distributed in accordance with the terms of the Llama 3 Community License Agreement.
import os import os
from logging import getLogger from logging import getLogger
from pathlib import Path from pathlib import Path

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

@ -96,6 +96,7 @@ def _convert_to_vllm_tool_calls_in_response(
call_id=call.id, call_id=call.id,
tool_name=call.function.name, tool_name=call.function.name,
arguments=json.loads(call.function.arguments), arguments=json.loads(call.function.arguments),
arguments_json=call.function.arguments,
) )
for call in tool_calls for call in tool_calls
] ]
@ -175,6 +176,7 @@ async def _process_vllm_chat_completion_stream_response(
call_id=tool_call_buf.call_id, call_id=tool_call_buf.call_id,
tool_name=tool_call_buf.tool_name, tool_name=tool_call_buf.tool_name,
arguments=args, arguments=args,
arguments_json=args_str,
), ),
parse_status=ToolCallParseStatus.succeeded, parse_status=ToolCallParseStatus.succeeded,
), ),