allenai/OLMo-2-1124-7B

Model Details

Model Card for OLMo 2 7B

We introduce OLMo 2, a new family of 7B and 13B models featuring a 9-point increase in MMLU, among other evaluation improvements, compared to the original OLMo 7B model. These gains come from training on OLMo-mix-1124 and Dolmino-mix-1124 datasets and staged training approach.

OLMo is a series of Open Language Models designed to enable the science of language models. These models are trained on the Dolma dataset. We are releasing all code, checkpoints, logs (coming soon), and associated training details.

Size	Training Tokens	Layers	Hidden Size	Attention Heads	Context Length
OLMo 2-7B	4 Trillion	32	4096	32	4096
OLMo 2-13B	5 Trillion	40	5120	40	4096

The core models released in this batch include the following:

Stage	OLMo 2 7B	OLMo 2 13B
Base Model	allenai/OLMo-2-1124-7B	allenai/OLMo-2-1124-13B
SFT	allenai/OLMo-2-1124-7B-SFT	allenai/OLMo-2-1124-13B-SFT
DPO	allenai/OLMo-2-1124-7B-DPO	allenai/OLMo-2-1124-13B-DPO
Final Models (RLVR)	allenai/OLMo-2-1124-7B-Instruct	allenai/OLMo-2-1124-13B-Instruct
Reward Model (RM)	allenai/OLMo-2-1124-7B-RM	(Same as 7B)

Installation

OLMo 2 will be supported in the next version of Transformers, and you need to install it from the main branch using:

pip install --upgrade git+https://github.com/huggingface/transformers.git

Inference

You can use OLMo with the standard HuggingFace transformers library:

from transformers import AutoModelForCausalLM, AutoTokenizer
olmo = AutoModelForCausalLM.from_pretrained("allenai/OLMo-2-1124-7B")
tokenizer = AutoTokenizer.from_pretrained("allenai/OLMo-2-1124-7B")
message = ["Language modeling is "]
inputs = tokenizer(message, return_tensors='pt', return_token_type_ids=False)
# optional verifying cuda
# inputs = {k: v.to('cuda') for k,v in inputs.items()}
# olmo = olmo.to('cuda')
response = olmo.generate(**inputs, max_new_tokens=100, do_sample=True, top_k=50, top_p=0.95)
print(tokenizer.batch_decode(response, skip_special_tokens=True)[0])
>> 'Language modeling is  a key component of any text-based application, but its effectiveness...'

For faster performance, you can quantize the model using the following method:

AutoModelForCausalLM.from_pretrained("allenai/OLMo-2-1124-7B", 
    torch_dtype=torch.float16, 
    load_in_8bit=True)  # Requires bitsandbytes

The quantized model is more sensitive to data types and CUDA operations. To avoid potential issues, it's recommended to pass the inputs directly to CUDA using:

inputs.input_ids.to('cuda')

We have released checkpoints for these models. For pretraining, the naming convention is stepXXX-tokensYYYB. For checkpoints with ingredients of the soup, the naming convention is stage2-ingredientN-stepXXX-tokensYYYB

To load a specific model revision with HuggingFace, simply add the argument revision:

olmo = AutoModelForCausalLM.from_pretrained("allenai/OLMo-2-1124-7B", revision="step1000-tokens5B")

Or, you can access all the revisions for the models via the following code snippet:

from huggingface_hub import list_repo_refs
out = list_repo_refs("allenai/OLMo-2-1124-7B")
branches = [b.name for b in out.branches]

Fine-tuning

Model fine-tuning can be done from the final checkpoint (the main revision of this model) or many intermediate checkpoints. Two recipes for tuning are available.

1. Fine-tune with the OLMo repository:

torchrun --nproc_per_node=8 scripts/train.py {path_to_train_config} \
    --data.paths=[{path_to_data}/input_ids.npy] \
    --data.label_mask_paths=[{path_to_data}/label_mask.npy] \
    --load_path={path_to_checkpoint} \
    --reset_trainer_state

For more documentation, see the GitHub readme.

2. Further fine-tuning support is being developing in AI2's Open Instruct repository. Details are here.

Model Description

• Developed by: Allen Institute for AI (Ai2)
• Model type: a Transformer style autoregressive language model.
• Language(s) (NLP): English
• License: The code and model are released under Apache 2.0.
• Contact: Technical inquiries: [email protected]. Press: [email protected]
• Date cutoff: Dec. 2023.

Model Sources

• Project Page: https://allenai.org/olmo
• Repositories:
  • Core repo (training, inference, fine-tuning etc.): https://github.com/allenai/OLMo
  • Evaluation code: https://github.com/allenai/OLMo-Eval
  • Further fine-tuning code: https://github.com/allenai/open-instruct
• Paper: Coming soon

Evaluation

Core model results for OLMo 2 7B and 13B models are found below.

Model	Train FLOPs	Average	ARC/C	HSwag	WinoG	MMLU	DROP	NQ	AGIEval	GSM8k	MMLUPro	TriviaQA
Open weights models:
Llama-2-13B	1.6·10²³	54.1	67.3	83.9	74.9	55.7	45.6	38.4	41.5	28.1	23.9	81.3
Mistral-7B-v0.3	n/a	58.8	78.3	83.1	77.7	63.5	51.8	37.2	47.3	40.1	30	79.3
Llama-3.1-8B	7.2·10²³	61.8	79.5	81.6	76.6	66.9	56.4	33.9	51.3	56.5	34.7	80.3
Mistral-Nemo-12B	n/a	66.9	85.2	85.6	81.5	69.5	69.2	39.7	54.7	62.1	36.7	84.6
Qwen-2.5-7B	8.2·10²³	67.4	89.5	89.7	74.2	74.4	55.8	29.9	63.7	81.5	45.8	69.4
Gemma-2-9B	4.4·10²³	67.8	89.5	87.3	78.8	70.6	63	38	57.3	70.1	42	81.8
Qwen-2.5-14B	16.0·10²³	72.2	94	94	80	79.3	51.5	37.3	71	83.4	52.8	79.1
Partially open models:
StableLM-2-12B	2.9·10²³	62.2	81.9	84.5	77.7	62.4	55.5	37.6	50.9	62	29.3	79.9
Zamba-2-7B	n/c	65.2	92.2	89.4	79.6	68.5	51.7	36.5	55.5	67.2	32.8	78.8
Fully open models:
Amber-7B	0.5·10²³	35.2	44.9	74.5	65.5	24.7	26.1	18.7	21.8	4.8	11.7	59.3
OLMo-7B	1.0·10²³	38.3	46.4	78.1	68.5	28.3	27.3	24.8	23.7	9.2	12.1	64.1
MAP-Neo-7B	2.1·10²³	49.6	78.4	72.8	69.2	58	39.4	28.9	45.8	12.5	25.9	65.1
OLMo-0424-7B	0.9·10²³	50.7	66.9	80.1	73.6	54.3	50	29.6	43.9	27.7	22.1	58.8
DCLM-7B	1.0·10²³	56.9	79.8	82.3	77.3	64.4	39.3	28.8	47.5	46.1	31.3	72.1
OLMo-2-1124-7B	1.8·10²³	62.9	79.8	83.8	77.2	63.7	60.8	36.9	50.4	67.5	31	78
OLMo-2-1124-13B	4.6·10²³	68.3	83.5	86.4	81.5	67.5	70.7	46.7	54.2	75.1	35.1	81.9

Model Details

Pretraining

	OLMo 2 7B	OLMo 2 13B
Pretraining Stage 1 (OLMo-Mix-1124)	4 trillion tokens (1 epoch)	5 trillion tokens (1.2 epochs)
Pretraining Stage 2 (Dolmino-Mix-1124)	50B tokens (3 runs) merged	100B tokens (3 runs) 300B tokens (1 run) merged
Post-training (Tulu 3 SFT OLMo mix)	SFT + DPO + PPO (preference mix)	SFT + DPO + PPO (preference mix)

Stage 1: Initial Pretraining

• Dataset: OLMo-Mix-1124 (3.9T tokens)
• Coverage: 90%+ of total pretraining budget
• 7B Model: ~1 epoch
• 13B Model: 1.2 epochs (5T tokens)

Stage 2: Fine-tuning

• Dataset: Dolmino-Mix-1124 (843B tokens)
• Three training mixes:
  • 50B tokens
  • 100B tokens
  • 300B tokens
• Mix composition: 50% high-quality data + academic/Q&A/instruction/math content

Model Merging

• 7B Model: 3 versions trained on 50B mix, merged via model souping
• 13B Model: 3 versions on 100B mix + 1 version on 300B mix, merged for final checkpoint

Bias, Risks, and Limitations

Like any base language model or fine-tuned model without safety filtering, these models can easily be prompted by users to generate harmful and sensitive content. Such content may also be produced unintentionally, especially in cases involving bias, so we recommend that users consider the risks when applying this technology. Additionally, many statements from OLMo or any LLM are often inaccurate, so facts should be verified.

Citation

A technical manuscript is forthcoming!

Model Card Contact

For errors in this model card, contact [email protected].