Skip to content

update Doubao-1.5-Embedding #2575

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Samoed merged 9 commits into from
Apr 30, 2025
Merged

update Doubao-1.5-Embedding #2575

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
cfd2ba9
update seed-embedding
namespace-Pt Apr 24, 2025
666a6ee
update seed models
Apr 28, 2025
ee2abe0
fix linting and tiktoken problem
Apr 29, 2025
5e46c5b
fix tiktoken bug
Apr 29, 2025
6def35c
fix lint
namespace-Pt Apr 29, 2025
5a78089
update name
namespace-Pt Apr 29, 2025
9b5a685
Update mteb/models/seed_models.py
namespace-Pt Apr 30, 2025
0af2d83
update logging
namespace-Pt Apr 30, 2025
81cb4da
update lint
namespace-Pt Apr 30, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions mteb/models/overview.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,6 +72,7 @@
ru_sentence_models,
salesforce_models,
searchmap_models,
seed_models,
sentence_transformers_models,
shuu_model,
siglip_models,
Expand Down Expand Up @@ -163,6 +164,7 @@
ara_models,
b1ade_models,
nb_sbert,
seed_models,
]
MODEL_REGISTRY = {}

Expand Down
167 changes: 167 additions & 0 deletions mteb/models/seed_models.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,167 @@
from __future__ import annotations

import logging
import os
import time
from functools import partial
from typing import Any

import numpy as np
import tqdm

from mteb.encoder_interface import PromptType
from mteb.model_meta import ModelMeta
from mteb.models.wrapper import Wrapper
from mteb.requires_package import requires_package

logger = logging.getLogger(__name__)


class SeedTextEmbeddingModel(Wrapper):
def __init__(
self,
model_name: str,
rate_limit_per_minute: int = 300,
**kwargs,
) -> None:
requires_package(
self,
"openai",
model_name,
install_instruction="pip install 'mteb[openai]'",
)
from openai import OpenAI

requires_package(
self,
"tiktoken",
model_name,
install_instruction="pip install 'mteb[openai]'",
)
import tiktoken

self.model_name = model_name
self.rate_limit_per_minute = rate_limit_per_minute
self.last_request_time = 0
self.tokenizer = tiktoken.get_encoding("cl100k_base")
self.client = OpenAI(
api_key=os.environ["ARK_API_KEY"],
base_url="https://ark.cn-beijing.volces.com/api/v3",
)

def _enforce_rate_limit(self):
"""Enforce rate limiting"""
current_time = time.time()
time_since_last_request = current_time - self.last_request_time
min_interval = 60.0 / self.rate_limit_per_minute

if time_since_last_request < min_interval:
time.sleep(min_interval - time_since_last_request)

self.last_request_time = time.time()

def _truncate_text(self, text: str, max_tokens: int = 32000) -> str:
"""Truncate text to fit within token limit"""
tokens = self.tokenizer.encode(text)
if len(tokens) > max_tokens:
tokens = tokens[:max_tokens]
text = self.tokenizer.decode(tokens)
return text

def _format_instruction(self, instruction: str, input_: str) -> str:
if isinstance(instruction, dict):
return input_
elif isinstance(instruction, str) and len(instruction):
return instruction + "\n" + input_
else:
return input_

def _embed(
self,
sentences: list[str],
instruction: str,
show_progress_bar: bool = False,
retries: int = 5,
) -> np.ndarray:
max_batch_size = 20
batches = [
sentences[i : i + max_batch_size]
for i in range(0, len(sentences), max_batch_size)
]

all_embeddings = []

for batch in tqdm.tqdm(batches, leave=False, disable=not show_progress_bar):
# Truncate texts
batch = [self._truncate_text(text) for text in batch]

# Add instruction to each text
batch = [self._format_instruction(instruction, text) for text in batch]

while retries > 0:
try:
self._enforce_rate_limit()
response = self.client.embeddings.create(
model=self.model_name, input=batch, encoding_format="float"
)
embeddings = [x.embedding for x in response.data]
break
except Exception as e:
logger.warning(
f"Retrying... {retries} retries left. Error: {str(e)}"
)
retries -= 1
if retries == 0:
raise e

all_embeddings.extend(embeddings)

return np.array(all_embeddings)

def encode(
self,
sentences: list[str],
*,
task_name: str,
prompt_type: PromptType | None = None,
**kwargs: Any,
) -> np.ndarray:
logger.warning("The API will be publicly available soon. Stay tuned!")

instruction = self.get_instruction(task_name, prompt_type)
show_progress_bar = kwargs.pop("show_progress_bar", False)

return self._embed(
sentences,
instruction=instruction,
show_progress_bar=show_progress_bar,
)


seed_embedding = ModelMeta(
name="ByteDance-Seed/Doubao-1.5-Embedding",
revision="1",
release_date="2025-04-25",
languages=[
"eng-Latn",
"zho-Hans",
],
loader=partial(
SeedTextEmbeddingModel,
model_name="doubao-1-5-embedding",
rate_limit_per_minute=300,
),
max_tokens=32768,
embed_dim=2048,
open_weights=False,
n_parameters=None,
memory_usage_mb=None,
license=None,
reference="https://console.volcengine.com/ark/region:ark+cn-beijing/model/detail?Id=doubao-1-5-embedding",
similarity_fn_name="cosine",
framework=["API"],
use_instructions=True,
training_datasets=None,
public_training_code=None,
public_training_data=None,
)