Text Generation
Transformers
Safetensors
English
olmoe
Mixture of Experts
olmo
conversational
Inference Endpoints
Edit model card
OLMoE Logo.

Model Summary

OLMoE-1B-7B-Instruct is a Mixture-of-Experts LLM with 1B active and 7B total parameters released in September 2024 (0924) that has been adapted via SFT and DPO from OLMoE-1B-7B. It yields state-of-the-art performance among models with a similar cost (1B) and is competitive with much larger models like Llama2-13B-Chat. OLMoE is 100% open-source.

This information and more can also be found on the OLMoE GitHub repository.

Use

Install transformers from source until a release after this PR & torch and run:

from transformers import OlmoeForCausalLM, AutoTokenizer
import torch

DEVICE = "cuda" if torch.cuda.is_available() else "cpu"

# Load different ckpts via passing e.g. `revision=kto`
model = OlmoeForCausalLM.from_pretrained("allenai/OLMoE-1B-7B-0924-Instruct").to(DEVICE)
tokenizer = AutoTokenizer.from_pretrained("allenai/OLMoE-1B-7B-0924-Instruct")
messages = [{"role": "user", "content": "Explain to me like I'm five what is Bitcoin."}]
inputs = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt").to(DEVICE)
out = model.generate(inputs, max_length=100)
print(tokenizer.decode(out[0]))
"""
<|endoftext|><|user|>
Explain to me like I'm five what is Bitcoin.
<|assistant|>
Bitcoin is like a special kind of money that you can use to buy things online. But unlike regular money, like dollars or euros, Bitcoin isn't printed by governments or banks. Instead, it's created by a special computer program that helps people keep track of it.

Here's how it works: imagine you have a bunch of toys, and you want to
"""

Branches:

Evaluation Snapshot

Task (→) MMLU GSM8k BBH Human-Eval Alpaca-Eval 1.0 XSTest IFEval Avg
Setup (→) 0-shot 8-shot CoT 3-shot 0-shot 0-shot 0-shot 0-shot
Metric (→) EM EM EM Pass@10 %win F1 Loose Acc
OLMo-1B (0724) 25.0 7.0 22.5 16.0 - 67.6 20.5 -
+SFT 36.0 12.5 27.2 21.2 41.5 81.9 26.1 35.9
+DPO 36.7 12.5 30.6 22.0 50.9 79.8 24.2 37.4
OLMo-7B (0724) 50.8 32.5 36.9 32.3 - 80.8 19.6 -
+SFT 54.2 25.0 35.7 38.5 70.9 86.1 39.7 49.3
+DPO 52.8 9.0 16.6 35.0 83.5 87.5 37.9 49.1
JetMoE-2B-9B 45.6 43.0 37.2 54.6 - 68.2 20.0 -
+SFT 46.1 53.5 35.6 64.8 69.3 55.6 30.5 50.4
DeepSeek-3B-16B 37.7 18.5 39.4 48.3 - 65.9 13.5 -
+Chat 48.5 46.5 40.8 70.1 74.8 85.6 32.3 57.0
Qwen1.5-3B-14B 60.4 13.5 27.2 60.2 - 73.4 20.9 -
+Chat 58.9 55.5 21.3 59.7 83.9 85.6 36.2 57.3
OLMoE (This Model) 49.8 3.0 33.6 22.4 - 59.7 16.6 -
+SFT 51.4 40.5 38.0 51.6 69.2 84.1 43.3 54.0
+DPO 51.9 45.5 37.0 54.8 84.0 82.6 48.1 57.7

Citation

@misc{muennighoff2024olmoeopenmixtureofexpertslanguage,
      title={OLMoE: Open Mixture-of-Experts Language Models}, 
      author={Niklas Muennighoff and Luca Soldaini and Dirk Groeneveld and Kyle Lo and Jacob Morrison and Sewon Min and Weijia Shi and Pete Walsh and Oyvind Tafjord and Nathan Lambert and Yuling Gu and Shane Arora and Akshita Bhagia and Dustin Schwenk and David Wadden and Alexander Wettig and Binyuan Hui and Tim Dettmers and Douwe Kiela and Ali Farhadi and Noah A. Smith and Pang Wei Koh and Amanpreet Singh and Hannaneh Hajishirzi},
      year={2024},
      eprint={2409.02060},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2409.02060}, 
}
Downloads last month
16,337
Safetensors
Model size
6.92B params
Tensor type
BF16
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for allenai/OLMoE-1B-7B-0924-Instruct

Finetuned
(2)
this model
Quantizations
8 models

Dataset used to train allenai/OLMoE-1B-7B-0924-Instruct

Spaces using allenai/OLMoE-1B-7B-0924-Instruct 3

Collection including allenai/OLMoE-1B-7B-0924-Instruct