Welcome to mLLMCelltype Discussions! #7
Replies: 2 comments 11 replies
-
|
你好，我是一名中国的用户，我在=== Starting check_consensus function ===环节反复遇到错误。 1、我使用OpenRouter 作为统一接入点； 2、这是我的原始代码： 3、我的问题在于每次都存在共识检查失败： （2）然而我明明设置了 consensus_check_model = "anthropic/claude-3-7-sonnet-20250219"，为什么无法按照我设置的模型进行共识检查呢？已经更换多种模型依然无法有效解决。 似乎，程序总是按照GPT 4o作为共识检查模型。你明白的，这在中国大陆不是很容易。而且OpenRouter用一条api集成多种模型，对中国大陆的用户非常方便。 所以您能解决这一问题吗？或者您还需要我提供更多详细信息吗？谢谢您！ Hello, I am a user from China, and I repeatedly encounter errors at the === Starting check_consensus function === stage.
consensus_results <- interactive_consensus_annotation(
input = scRNA_harmony_marker2,
tissue_name = "human osteosarcoma tissues", # Directly provide the natural language description of the tissue
models = c(
"anthropic/claude-3-7-sonnet-20250219",
"meta-llama/llama-3-70b-instruct",
"google/gemini-2.5-pro-preview-03-25"
),
api_keys = list(
openrouter = "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"
),
top_gene_count = 10,
controversy_threshold = 0.7,
entropy_threshold = 1,
max_discussion_rounds = 1,
consensus_check_model = "anthropic/claude-3-7-sonnet-20250219",
cache_dir = cache_dir
)
(1) The error message is as follows: (2) However, I have clearly set It seems that the program always uses GPT-4o as the consensus check model. You understand that this is not very easy in mainland China. Moreover, OpenRouter integrates multiple models with a single API, which is very convenient for users in mainland China. So can you solve this issue? Or do you need me to provide more detailed information? Thank you! |
Beta Was this translation helpful? Give feedback.
-
|
我在熬大夜修这个问题，预计 1 小时之后完成，再给我一些时间((((( |
Beta Was this translation helpful? Give feedback.
-
|
因为我自己都是 openrouter 和其他 model 混用的，所以没有碰到过这个问题，不好意思，在解决了😭 |
Beta Was this translation helpful? Give feedback.
-
|
问题解决了，现在纯 openrouter 加 consensus_check_model = "anthropic/claude-3-7-sonnet-20250219" 的方案可以 work 了 |
Beta Was this translation helpful? Give feedback.
-
|
Hey, I am getting this error in my macbook m3 pro ImportError: cannot import name 'genai' from 'google' (unknown location) I tried pip install google-generativeai, and import google.generativeai as genai But still the same. I was actually trying to use deepseek r1 and I am not sure why it was trying to force import genai. |
Beta Was this translation helpful? Give feedback.
-
|
Hi @mhbsiam, I've investigated the OpenRouter API errors you're encountering. The issue is with the model IDs you're trying to use. Here's what's happening and how to fix it: The Problem
Your attempts:
The SolutionHere's how to use free OpenRouter models correctly: consensus_results = interactive_consensus_annotation(
marker_genes=marker_genes,
species="human",
models=[
{"provider": "openrouter", "model": "microsoft/mai-ds-r1:free"},
{"provider": "openrouter", "model": "qwen/qwq-32b:free"}
],
api_keys={
"openrouter": "your-openrouter-api-key"
},
max_discussion_rounds=3
)Other Free Models You Can TryHere are some other free models that are currently available on OpenRouter:
Checking Available ModelsYou can check the current list of available models with this code: import requests
import os
# Get your OpenRouter API key
api_key = "your-openrouter-api-key"
# Get available models
response = requests.get(
"https://openrouter.ai/api/v1/models",
headers={"Authorization": f"Bearer {api_key}"}
)
# Print all free models
models = response.json()["data"]
print("Available free OpenRouter models:")
for model in models:
model_id = model["id"]
is_free = model.get("pricing", {}).get("prompt") == 0 and model.get("pricing", {}).get("completion") == 0
if is_free:
print(f" - {model_id}")I've updated our documentation to make this clearer for future users. Let me know if you have any other questions! Best, |
Beta Was this translation helpful? Give feedback.
-
|
Hi Chen, I actually opted to use just one model: mai-ds-r1:free I think, if you add an example code (i.e. instruction) in your README on how to use just one free model from OpenRouter using Python, many people will benefit from it. Thanks again. |
Beta Was this translation helpful? Give feedback.
-
|
Hi @mhbsiam, I'm pleased to let you know that I've added an example of using a single free OpenRouter model to our documentation at https://github.com/cafferychen777/mLLMCelltype#using-a-single-free-openrouter-model Thank you for your helpful suggestion! I agree this will benefit many users who want a simpler and more cost-effective solution for cell type annotation. I'm glad to hear that the mai-ds-r1:free model worked well for you - it's indeed both fast and accurate for this purpose. Your feedback helps us continuously improve mLLMCelltype. Please don't hesitate to share any other suggestions or questions you might have. Best regards, |
Beta Was this translation helpful? Give feedback.
-
|
Hi Chen, Just a quick thought. In the documentation, instead of manually finding out and assigning the markers for each cluster, one can also rank the top genes for each cluster and use that to identify cell types. This is what I used: This could be useful for someone trying to see what's potentially in the dataset. |
Beta Was this translation helpful? Give feedback.
-
|
Hi @mhbsiam, Thank you for your excellent suggestion about extracting marker genes from AnnData objects! I've tested your method and found it works perfectly - it produces identical results to the standard approach we were using. I've added your suggested code as an example in our README under the "Using a Single Free OpenRouter Model" section. The new example shows how to:
This makes it much easier for Scanpy users to integrate mLLMCelltype into their workflows without having to manually extract marker genes. I appreciate your contribution to making our package more user-friendly! The code is now available in our repository, and I agree that the MAI-DS-R1 free model is both fast and accurate for cell type annotation tasks. Thanks again for sharing this helpful approach with our community! Best regards, |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
👋 Welcome to the mLLMCelltype Discussion Forum!
We've created this discussion space as a platform to connect our community members. We hope you'll use this forum to:
We believe in building an inclusive, supportive community where everyone can contribute regardless of their background or expertise level. Whether you're a bioinformatician, computational biologist, AI researcher, or just curious about the intersection of LLMs and single-cell analysis, you're welcome here!
To get started, feel free to introduce yourself in the comments and share what brought you to mLLMCelltype.
Happy discussing! 🧬🤖
Beta Was this translation helpful? Give feedback.
All reactions