Instructions to use formalmathatepfl/OLMo-3-7B-finetuned with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use formalmathatepfl/OLMo-3-7B-finetuned with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="formalmathatepfl/OLMo-3-7B-finetuned")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("formalmathatepfl/OLMo-3-7B-finetuned")
model = AutoModelForCausalLM.from_pretrained("formalmathatepfl/OLMo-3-7B-finetuned")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use formalmathatepfl/OLMo-3-7B-finetuned with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "formalmathatepfl/OLMo-3-7B-finetuned"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "formalmathatepfl/OLMo-3-7B-finetuned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/formalmathatepfl/OLMo-3-7B-finetuned

SGLang

How to use formalmathatepfl/OLMo-3-7B-finetuned with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "formalmathatepfl/OLMo-3-7B-finetuned" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "formalmathatepfl/OLMo-3-7B-finetuned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "formalmathatepfl/OLMo-3-7B-finetuned" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "formalmathatepfl/OLMo-3-7B-finetuned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use formalmathatepfl/OLMo-3-7B-finetuned with Docker Model Runner:
```
docker model run hf.co/formalmathatepfl/OLMo-3-7B-finetuned
```

olmo3-lean-sft

This model is a fine-tuned version of allenai/OLMo-3-1B on the lean_sft dataset. It achieves the following results on the evaluation set:

Loss: 0.0655

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 2
total_train_batch_size: 32
total_eval_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.05
num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss
0.2558	0.0194	100	0.2367
0.1315	0.0388	200	0.1327
0.11	0.0582	300	0.1104
0.1012	0.0775	400	0.1001
0.0926	0.0969	500	0.0948
0.092	0.1163	600	0.0913
0.0866	0.1357	700	0.0884
0.0846	0.1551	800	0.0864
0.0843	0.1745	900	0.0848
0.08	0.1939	1000	0.0834
0.0813	0.2132	1100	0.0823
0.0801	0.2326	1200	0.0812
0.0814	0.2520	1300	0.0801
0.0781	0.2714	1400	0.0792
0.0774	0.2908	1500	0.0785
0.0754	0.3102	1600	0.0777
0.078	0.3296	1700	0.0771
0.0758	0.3489	1800	0.0763
0.074	0.3683	1900	0.0758
0.0731	0.3877	2000	0.0751
0.0739	0.4071	2100	0.0746
0.0736	0.4265	2200	0.0740
0.0739	0.4459	2300	0.0737
0.0692	0.4653	2400	0.0731
0.0712	0.4846	2500	0.0726
0.0714	0.5040	2600	0.0721
0.0698	0.5234	2700	0.0718
0.0718	0.5428	2800	0.0712
0.0691	0.5622	2900	0.0709
0.0693	0.5816	3000	0.0705
0.0691	0.6009	3100	0.0700
0.0707	0.6203	3200	0.0696
0.0693	0.6397	3300	0.0693
0.0694	0.6591	3400	0.0689
0.0687	0.6785	3500	0.0686
0.0659	0.6979	3600	0.0682
0.0658	0.7173	3700	0.0678
0.0677	0.7366	3800	0.0676
0.0671	0.7560	3900	0.0672
0.0662	0.7754	4000	0.0670
0.0675	0.7948	4100	0.0667
0.0665	0.8142	4200	0.0665
0.0644	0.8336	4300	0.0662
0.0643	0.8530	4400	0.0661
0.0636	0.8723	4500	0.0659
0.0667	0.8917	4600	0.0658
0.0644	0.9111	4700	0.0657
0.0654	0.9305	4800	0.0656
0.0662	0.9499	4900	0.0655
0.0651	0.9693	5000	0.0655
0.0642	0.9887	5100	0.0655

Framework versions

Transformers 4.57.3
Pytorch 2.9.0+cu129
Datasets 4.0.0
Tokenizers 0.22.1

Downloads last month: 26

Safetensors

Model size

528k params

Tensor type

BF16