mirror of
https://github.com/meta-llama/llama-stack.git
synced 2025-10-03 19:57:35 +00:00
2e544ecd8a
This is a sweeping change to clean up some gunk around our "Tool" definitions. First, we had two types `Tool` and `ToolDef`. The first of these was a "Resource" type for the registry but we had stopped registering tools inside the Registry long back (and only registered ToolGroups.) The latter was for specifying tools for the Agents API. This PR removes the former and adds an optional `toolgroup_id` field to the latter. Secondly, as pointed out by @bbrowning in https://github.com/llamastack/llama-stack/pull/3003#issuecomment-3245270132, we were doing a lossy conversion from a full JSON schema from the MCP tool specification into our ToolDefinition to send it to the model. There is no necessity to do this -- we ourselves aren't doing any execution at all but merely passing it to the chat completions API which supports this. By doing this (and by doing it poorly), we encountered limitations like not supporting array items, or not resolving $refs, etc. To fix this, we replaced the `parameters` field by `{ input_schema, output_schema }` which can be full blown JSON schemas. Finally, there were some types in our llama-related chat format conversion which needed some cleanup. We are taking this opportunity to clean those up. This PR is a substantial breaking change to the API. However, given our window for introducing breaking changes, this suits us just fine. I will be landing a concurrent `llama-stack-client` change as well since API shapes are changing.
317 lines
12 KiB
Python
317 lines
12 KiB
Python
# 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.
|
|
|
|
import io
|
|
import json
|
|
import uuid
|
|
from dataclasses import dataclass
|
|
|
|
import torch
|
|
from PIL import Image as PIL_Image
|
|
|
|
# TODO: either fork these or move them to the common package
|
|
from ..datatypes import (
|
|
BuiltinTool,
|
|
RawContent,
|
|
RawMediaItem,
|
|
RawMessage,
|
|
RawTextItem,
|
|
Role,
|
|
StopReason,
|
|
ToolCall,
|
|
ToolPromptFormat,
|
|
)
|
|
from ..llama3.tool_utils import ToolUtils
|
|
from .args import VisionArgs
|
|
from .datatypes import LLMInput
|
|
from .preprocess import ResizeNormalizeImageTransform, VariableSizeImageTransform
|
|
from .tokenizer import Tokenizer
|
|
|
|
|
|
def role_str(role: Role) -> str:
|
|
role_strs = {
|
|
Role.user: "user",
|
|
Role.system: "system",
|
|
Role.tool: "ipython", # special
|
|
Role.assistant: "assistant",
|
|
}
|
|
return role_strs[role]
|
|
|
|
|
|
@dataclass
|
|
class TransformedImage:
|
|
image_tiles: torch.Tensor
|
|
# is the aspect ratio needed anywhere?
|
|
aspect_ratio: tuple[int, int]
|
|
|
|
|
|
def convert_image_to_rgb(image: PIL_Image.Image, bg: tuple[int, int, int] = (255, 255, 255)) -> PIL_Image.Image:
|
|
if image.mode == "RGBA":
|
|
image.load() # for png.split()
|
|
new_img = PIL_Image.new("RGB", image.size, bg)
|
|
new_img.paste(image, mask=image.split()[3]) # 3 is the alpha channel
|
|
return new_img
|
|
return image.convert("RGB")
|
|
|
|
|
|
class ChatFormat:
|
|
possible_headers: dict[Role, str]
|
|
|
|
def __init__(
|
|
self,
|
|
tokenizer: Tokenizer,
|
|
vision_args: VisionArgs | None = None,
|
|
max_num_chunks: int = 16,
|
|
):
|
|
self.tokenizer = tokenizer
|
|
self.vision_args = vision_args
|
|
self.max_num_chunks = max_num_chunks
|
|
|
|
self.possible_headers = {role: f"<|header_start|>{role_str(role)}<|header_end|>\n\n" for role in Role}
|
|
|
|
self.image_transform = None
|
|
self.dynamic_image_transform = None
|
|
if vision_args:
|
|
self.dynamic_image_transform = VariableSizeImageTransform(vision_args.image_size.width)
|
|
self.image_transform = ResizeNormalizeImageTransform(
|
|
vision_args.image_size.width, vision_args.image_size.height
|
|
)
|
|
|
|
def _encode_header(self, role: str) -> list[int]:
|
|
tokens = []
|
|
tokens.append(self.tokenizer.special_tokens["<|header_start|>"])
|
|
|
|
# TODO: need to check if this is correct
|
|
tokens.extend(self.tokenizer.encode("ipython" if role == "tool" else role, bos=False, eos=False))
|
|
tokens.append(self.tokenizer.special_tokens["<|header_end|>"])
|
|
tokens.extend(self.tokenizer.encode("\n\n", bos=False, eos=False))
|
|
return tokens
|
|
|
|
def encode_content(self, content: RawContent) -> LLMInput:
|
|
tokens, images = self._encode_content(content, bos=True)
|
|
return self._model_input_from_tokens_images(tokens, images)
|
|
|
|
def _encode_image(
|
|
self,
|
|
transformed_image: TransformedImage,
|
|
) -> list[int]:
|
|
assert self.vision_args is not None, "The model is not vision-enabled"
|
|
|
|
image_tensor = transformed_image.image_tiles
|
|
image_channels = image_tensor.shape[-3]
|
|
image_height = image_tensor.shape[-2]
|
|
image_width = image_tensor.shape[-1]
|
|
image_chunks = image_tensor.view(-1, image_channels, image_height, image_width).shape[0]
|
|
|
|
patch_height = self.vision_args.patch_size.height
|
|
patch_width = self.vision_args.patch_size.width
|
|
|
|
if image_height % patch_height != 0:
|
|
raise ValueError(f"{image_height=} not divisible by {patch_height=}")
|
|
if image_width % patch_width != 0:
|
|
raise ValueError(f"{image_width=} not divisible by {patch_width=}")
|
|
|
|
ds_ratio = int(round(1.0 / (self.vision_args.pixel_shuffle_ratio**2)))
|
|
n_patches_per_chunk = int((image_height // patch_height) * (image_width // patch_width) // ds_ratio)
|
|
|
|
image_ar = transformed_image.aspect_ratio
|
|
tokens = [self.tokenizer.special_tokens["<|image_start|>"]]
|
|
if image_chunks == 1:
|
|
tokens += [self.tokenizer.special_tokens["<|image|>"]]
|
|
tokens += [self.tokenizer.special_tokens["<|patch|>"]] * n_patches_per_chunk
|
|
tokens += [self.tokenizer.special_tokens["<|image_end|>"]]
|
|
else:
|
|
ratio_h, ratio_w = image_ar
|
|
for _ in range(ratio_h):
|
|
for xx in range(ratio_w):
|
|
tokens += [self.tokenizer.special_tokens["<|patch|>"]] * n_patches_per_chunk
|
|
if xx < ratio_w - 1:
|
|
tokens.append(self.tokenizer.special_tokens["<|tile_x_separator|>"])
|
|
|
|
tokens.append(self.tokenizer.special_tokens["<|tile_y_separator|>"])
|
|
|
|
tokens += [self.tokenizer.special_tokens["<|image|>"]]
|
|
tokens += [self.tokenizer.special_tokens["<|patch|>"]] * n_patches_per_chunk
|
|
tokens += [self.tokenizer.special_tokens["<|image_end|>"]]
|
|
|
|
return tokens
|
|
|
|
def _encode_content(self, content: RawContent, bos: bool = False) -> tuple[list[int], list[TransformedImage]]:
|
|
tokens = []
|
|
tranformed_images = []
|
|
|
|
added_bos = False
|
|
|
|
def _process(c):
|
|
nonlocal added_bos, bos
|
|
|
|
if isinstance(c, str) or isinstance(c, RawTextItem):
|
|
if isinstance(c, RawTextItem):
|
|
c = c.text
|
|
tokens.extend(self.tokenizer.encode(c, bos=False if added_bos else bos, eos=False))
|
|
added_bos = True
|
|
|
|
elif isinstance(c, RawMediaItem):
|
|
if not self.vision_args:
|
|
raise ValueError("The model is not vision-enabled, but a media item was found")
|
|
|
|
bos = False if added_bos else bos
|
|
if bos:
|
|
tokens.append(self.tokenizer.special_tokens["<|begin_of_text|>"])
|
|
added_bos = True
|
|
|
|
bytes_io = io.BytesIO(c.data) if isinstance(c.data, bytes) else c.data
|
|
image = PIL_Image.open(bytes_io)
|
|
image = convert_image_to_rgb(image)
|
|
image_tiles, ar = self.dynamic_image_transform(image, max_num_chunks=self.max_num_chunks)
|
|
|
|
if image_tiles.shape[0] > 1:
|
|
image_global = self.image_transform(image)
|
|
image_global = image_global.unsqueeze(0)
|
|
image_combine = torch.cat((image_tiles, image_global), dim=0)
|
|
image_tiles = image_combine
|
|
|
|
transformed_image = TransformedImage(image_tiles=image_tiles, aspect_ratio=ar)
|
|
tokens.extend(self._encode_image(transformed_image))
|
|
tranformed_images.append(transformed_image)
|
|
|
|
if isinstance(content, list):
|
|
for c in content:
|
|
_process(c)
|
|
else:
|
|
_process(content)
|
|
|
|
return tokens, tranformed_images
|
|
|
|
def encode_message(
|
|
self, message: RawMessage, tool_prompt_format: ToolPromptFormat
|
|
) -> tuple[list[int], list[TransformedImage]]:
|
|
tokens = self._encode_header(message.role)
|
|
images = []
|
|
|
|
def _process_content(c):
|
|
toks, imgs = self._encode_content(c)
|
|
tokens.extend(toks)
|
|
images.extend(imgs)
|
|
|
|
_process_content(message.content)
|
|
|
|
if message.role == "user" and message.context is not None:
|
|
# This is RAG context; why is it here in the chat format? I don't think
|
|
# this is needed and can be moved upwards
|
|
_process_content("\n\n")
|
|
_process_content(message.context)
|
|
|
|
if message.role == "assistant":
|
|
for t in message.tool_calls:
|
|
content = ToolUtils.encode_tool_call(t, tool_prompt_format)
|
|
_process_content(content)
|
|
|
|
# Tool calls and Tool Response messages should be eom
|
|
eom = False
|
|
if message.role == "assistant":
|
|
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|>"])
|
|
return tokens, images
|
|
|
|
def encode_dialog_prompt(
|
|
self,
|
|
messages: list[RawMessage],
|
|
tool_prompt_format: ToolPromptFormat = ToolPromptFormat.json,
|
|
) -> LLMInput:
|
|
tokens = []
|
|
images = []
|
|
tokens.append(self.tokenizer.special_tokens["<|begin_of_text|>"])
|
|
for message in messages:
|
|
toks, imgs = self.encode_message(message, tool_prompt_format)
|
|
tokens.extend(toks)
|
|
images.extend(imgs)
|
|
|
|
# Add the start of an assistant message for the model to complete.
|
|
tokens.extend(self._encode_header("assistant"))
|
|
|
|
return self._model_input_from_tokens_images(tokens, images)
|
|
|
|
# TODO(this should be generic, not only for assistant messages)
|
|
def decode_assistant_message(self, tokens: list[int], stop_reason: StopReason) -> RawMessage:
|
|
content = self.tokenizer.decode(tokens)
|
|
|
|
return self.decode_assistant_message_from_content(content, stop_reason)
|
|
|
|
def decode_assistant_message_from_content(self, content: str, stop_reason: StopReason) -> RawMessage:
|
|
content = content.strip(" ")
|
|
header_str = self.possible_headers[Role.assistant]
|
|
if content.startswith(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|>"):
|
|
content = content[: -len("<|eot|>")]
|
|
stop_reason = StopReason.end_of_turn
|
|
elif content.endswith("<|eom|>"):
|
|
content = content[: -len("<|eom|>")]
|
|
stop_reason = StopReason.end_of_message
|
|
|
|
tool_name = None
|
|
tool_arguments = {}
|
|
|
|
custom_tool_info = ToolUtils.maybe_extract_custom_tool_call(content)
|
|
if custom_tool_info is not None:
|
|
tool_name, tool_arguments = custom_tool_info
|
|
# Sometimes when agent has custom tools alongside builin tools
|
|
# Agent responds for builtin tool calls in the format of the custom tools
|
|
# This code tries to handle that case
|
|
if tool_name in BuiltinTool.__members__:
|
|
tool_name = BuiltinTool[tool_name]
|
|
tool_arguments = {
|
|
"query": list(tool_arguments.values())[0],
|
|
}
|
|
else:
|
|
builtin_tool_info = ToolUtils.maybe_extract_builtin_tool_call(content)
|
|
if builtin_tool_info is not None:
|
|
tool_name, query = builtin_tool_info
|
|
tool_arguments = {
|
|
"query": query,
|
|
}
|
|
if tool_name in BuiltinTool.__members__:
|
|
tool_name = BuiltinTool[tool_name]
|
|
elif ipython:
|
|
tool_name = BuiltinTool.code_interpreter
|
|
tool_arguments = {
|
|
"code": content,
|
|
}
|
|
|
|
tool_calls = []
|
|
if tool_name is not None and tool_arguments is not None:
|
|
call_id = str(uuid.uuid4())
|
|
tool_calls.append(
|
|
ToolCall(
|
|
call_id=call_id,
|
|
tool_name=tool_name,
|
|
arguments=json.dumps(tool_arguments),
|
|
)
|
|
)
|
|
content = ""
|
|
|
|
return RawMessage(
|
|
role="assistant",
|
|
content=content,
|
|
stop_reason=stop_reason,
|
|
tool_calls=tool_calls,
|
|
)
|
|
|
|
def _model_input_from_tokens_images(self, tokens: list[int], images: list[TransformedImage]) -> LLMInput:
|
|
return LLMInput(
|
|
tokens=tokens,
|
|
images=[x.image_tiles for x in images] if len(images) > 0 else None,
|
|
)
|