Instructions to use epfl-dlab/spp-t0-mt-3b-base with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use epfl-dlab/spp-t0-mt-3b-base with Transformers:

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

pipe = pipeline("text-generation", model="epfl-dlab/spp-t0-mt-3b-base")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("epfl-dlab/spp-t0-mt-3b-base")
model = AutoModelForCausalLM.from_pretrained("epfl-dlab/spp-t0-mt-3b-base", device_map="auto")
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 Settings

vLLM

How to use epfl-dlab/spp-t0-mt-3b-base with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "epfl-dlab/spp-t0-mt-3b-base"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "epfl-dlab/spp-t0-mt-3b-base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/epfl-dlab/spp-t0-mt-3b-base

SGLang

How to use epfl-dlab/spp-t0-mt-3b-base 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 "epfl-dlab/spp-t0-mt-3b-base" \
    --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": "epfl-dlab/spp-t0-mt-3b-base",
		"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 "epfl-dlab/spp-t0-mt-3b-base" \
        --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": "epfl-dlab/spp-t0-mt-3b-base",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use epfl-dlab/spp-t0-mt-3b-base with Docker Model Runner:
```
docker model run hf.co/epfl-dlab/spp-t0-mt-3b-base
```

SPP-T0-MT — Base (3B)

Type: base (pretrained) model. Not instruction-tuned and ships no chat template.

Trained with SPP from token zero, followed by an additional reflection-focused midtraining stage resumed from the SPP-T0 checkpoint (replacing the final learning-rate cooldown).

Synthetic Persona Pretraining (SPP)

Synthetic Persona Pretraining (SPP) installs a target value persona during pretraining rather than only during alignment. Value-laden, first-person reflections, generated against a constitution, are appended to a subset of pretraining documents after a special <assistant> token. Attention masking and RoPE position aliasing keep the reflection from changing the continuation of the original document. This model is trained with SPP.

Instruction-tuned counterpart: epfl-dlab/spp-t0-mt-3b-instruct.

Model details

Architecture: Llama-3.2-3B-shaped, trained from scratch.
Tokenizer: the SmolLM2 tokenizer extended with an <assistant> marker and constitution tokens (vocabulary 49280).
Pretraining: ~500B tokens on a subset of the Olmo 3 Dolma 3 mixture, with SPP reflections inserted into the safety-annotated documents within it, followed by a reflection-focused midtraining stage on those annotated documents.

Training checkpoints

Intermediate checkpoints are published as git revisions on this repo, so any point in the trajectory can be loaded by passing revision=:

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

repo = "epfl-dlab/spp-t0-mt-3b-base"
tok = AutoTokenizer.from_pretrained(repo)          # identical at every revision
model = AutoModelForCausalLM.from_pretrained(
    repo, revision="step-0", dtype=torch.bfloat16, device_map="auto"
)

Revision	Midtraining step	LR phase
`step-0`	0 / 72,895	— (init from `spp-t0-3b-base` step 225,000)
`step-25000`	25,000 / 72,895	linear decay
`step-50000`	50,000 / 72,895	linear decay
`step-72895`	72,895 / 72,895	linear decay — same weights as `main`

main always holds the finished model (step 72,895). Only model weights are published — optimizer and RNG state are not included, so these revisions support evaluation, probing, and fine-tuning, but not exact resumption of the original run.

Steps are counted from the start of midtraining. Midtraining resumed from pretraining step 225,000, so the earlier part of this model's history is the pretraining trajectory in epfl-dlab/spp-t0-3b-base (revisions step-25000 … step-225000). Those checkpoints are shared and are not duplicated here; step-0 is the exact fork point.

Intended use

Research on alignment and safety. As a base model it is meant for continuation, probing, or further fine-tuning; it is not instruction-tuned and can produce incorrect or unsafe content.

Collection including epfl-dlab/spp-t0-mt-3b-base

SPP — Synthetic Persona Pretraining

Collection

Installing a value persona during pretraining, not just alignment. 3B base models with full step-by-step training trajectories, plus instruct. • 11 items • Updated about 8 hours ago

epfl-dlab
/

spp-t0-mt-3b-base