github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #2049] Embedding API could return empty embedding while using completion API from LiteLLM #26945

New Issue
Open
opened 2026-04-22 03:44:24 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-22 03:44:24 -05:00
Owner
Copy Link

Originally created by @James4Ever0 on GitHub (Jan 18, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/2049

To reproduce:

Launch a LiteLLM service:

litellm --model ollama/openhermes2.5-mistral --drop_params

Call the service /completion API continuously first, meanwhile you call embedding API via Langchain, and hopefully during the very gap (very short) between each /completion call you get empty embedding from Langchain.

To call the /completion API:

import os
os.environ['OPENAI_API_KEY'] = 'any'
os.environ['OPENAI_API_BASE'] = 'http://0.0.0.0:8000'

from contextlib import contextmanager
from langchain.llms import OpenAI
import tiktoken

def print_center(banner: str):
    print(banner.center(50, "="))


class LLM:
    """
    A class for running a Language Model Chain.
    """

    def __init__(self, prompt: str, temperature=0, gpt_4=False):
        """
        Initializes the LLM class.
        Args:
            prompt (PromptTemplate): The prompt template to use.
            temperature (int): The temperature to use for the model.
            gpt_4 (bool): Whether to use GPT-4 or Text-Davinci-003.
        Side Effects:
            Sets the class attributes.
        """
        self.prompt = prompt
        self.prompt_size = self.number_of_tokens(prompt)
        self.temperature = temperature
        self.gpt_4 = gpt_4
        self.model_name = "gpt-4" if self.gpt_4 else "text-davinci-003"
        self.max_tokens = 4097 * 2 if self.gpt_4 else 4097
        self.show_init_config()

    def show_init_config(self):
        print_center("init params")
        print(f"Model: {self.model_name}")
        print(f"Max Tokens: {self.max_tokens}")
        print(f"Prompt Size: {self.prompt_size}")
        print(f"Temperature: {self.temperature}")
        print_center("init config")
        print(self.prompt)

    def run(self, query):
        """
        Runs the Language Model Chain.
        Args:
            code (str): The code to use for the chain.
            **kwargs (dict): Additional keyword arguments.
        Returns:
            str: The generated text.
        """
        llm = OpenAI(
            temperature=self.temperature,
            max_tokens=-1,
            model_name=self.model_name,
            disallowed_special=(),  # to suppress error when special tokens within the input text (encode special tokens as normal text)
        )
        # chain = LLMChain(llm=llm, prompt=self.prompt)
        chunk_list = []
        print_center("query")
        print(query)
        print_center("response")
        _input = "\n".join([self.prompt, query])
        for chunk in llm.stream(input=_input):
            print(chunk, end="", flush=True)
            chunk_list.append(chunk)
        print()

        result = "".join(chunk_list)
        return result

    def number_of_tokens(self, text):
        """
        Counts the number of tokens in a given text.
        Args:
            text (str): The text to count tokens for.
        Returns:
            int: The number of tokens in the text.
        """
        encoding = tiktoken.encoding_for_model("gpt-4")
        return len(encoding.encode(text, disallowed_special=()))


@contextmanager
def llm_context(prompt: str, temperature=0, gpt_4=False):
    model = LLM(prompt, temperature=temperature, gpt_4=gpt_4)
    try:
        yield model
    finally:
        del model

if __name__ == "__main__":
    while True:
        with llm_context("You are a helpful assistant.") as model:
            model.run("Write me a 100 words news.")

To call the embedding API:

from langchain.embeddings import OllamaEmbeddings

ollama_emb = OllamaEmbeddings(
    model="openhermes2.5-mistral:latest",
)

while True:
    embed = ollama_emb.embed_query("Hello world")
    print("empty embedding?", embed is None)
Originally created by @James4Ever0 on GitHub (Jan 18, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/2049 To reproduce: Launch a LiteLLM service: ```bash litellm --model ollama/openhermes2.5-mistral --drop_params ``` Call the service `/completion` API continuously first, meanwhile you call embedding API via Langchain, and hopefully during the very gap (very short) between each `/completion` call you get empty embedding from Langchain. To call the `/completion` API: ```python import os os.environ['OPENAI_API_KEY'] = 'any' os.environ['OPENAI_API_BASE'] = 'http://0.0.0.0:8000' from contextlib import contextmanager from langchain.llms import OpenAI import tiktoken def print_center(banner: str): print(banner.center(50, "=")) class LLM: """ A class for running a Language Model Chain. """ def __init__(self, prompt: str, temperature=0, gpt_4=False): """ Initializes the LLM class. Args: prompt (PromptTemplate): The prompt template to use. temperature (int): The temperature to use for the model. gpt_4 (bool): Whether to use GPT-4 or Text-Davinci-003. Side Effects: Sets the class attributes. """ self.prompt = prompt self.prompt_size = self.number_of_tokens(prompt) self.temperature = temperature self.gpt_4 = gpt_4 self.model_name = "gpt-4" if self.gpt_4 else "text-davinci-003" self.max_tokens = 4097 * 2 if self.gpt_4 else 4097 self.show_init_config() def show_init_config(self): print_center("init params") print(f"Model: {self.model_name}") print(f"Max Tokens: {self.max_tokens}") print(f"Prompt Size: {self.prompt_size}") print(f"Temperature: {self.temperature}") print_center("init config") print(self.prompt) def run(self, query): """ Runs the Language Model Chain. Args: code (str): The code to use for the chain. **kwargs (dict): Additional keyword arguments. Returns: str: The generated text. """ llm = OpenAI( temperature=self.temperature, max_tokens=-1, model_name=self.model_name, disallowed_special=(), # to suppress error when special tokens within the input text (encode special tokens as normal text) ) # chain = LLMChain(llm=llm, prompt=self.prompt) chunk_list = [] print_center("query") print(query) print_center("response") _input = "\n".join([self.prompt, query]) for chunk in llm.stream(input=_input): print(chunk, end="", flush=True) chunk_list.append(chunk) print() result = "".join(chunk_list) return result def number_of_tokens(self, text): """ Counts the number of tokens in a given text. Args: text (str): The text to count tokens for. Returns: int: The number of tokens in the text. """ encoding = tiktoken.encoding_for_model("gpt-4") return len(encoding.encode(text, disallowed_special=())) @contextmanager def llm_context(prompt: str, temperature=0, gpt_4=False): model = LLM(prompt, temperature=temperature, gpt_4=gpt_4) try: yield model finally: del model if __name__ == "__main__": while True: with llm_context("You are a helpful assistant.") as model: model.run("Write me a 100 words news.") ``` To call the embedding API: ```python from langchain.embeddings import OllamaEmbeddings ollama_emb = OllamaEmbeddings( model="openhermes2.5-mistral:latest", ) while True: embed = ollama_emb.embed_query("Hello world") print("empty embedding?", embed is None) ```
GiteaMirror added the embeddingsbugapi labels 2026-04-22 03:44:24 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label api bug embeddings
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#26945
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?