Instructions to use mechramc/kalavai-phase1-410m-science-specialist-seed42 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use mechramc/kalavai-phase1-410m-science-specialist-seed42 with Transformers:

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

pipe = pipeline("text-generation", model="mechramc/kalavai-phase1-410m-science-specialist-seed42")

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("mechramc/kalavai-phase1-410m-science-specialist-seed42")
model = AutoModelForMultimodalLM.from_pretrained("mechramc/kalavai-phase1-410m-science-specialist-seed42")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use mechramc/kalavai-phase1-410m-science-specialist-seed42 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "mechramc/kalavai-phase1-410m-science-specialist-seed42"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mechramc/kalavai-phase1-410m-science-specialist-seed42",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/mechramc/kalavai-phase1-410m-science-specialist-seed42

SGLang

How to use mechramc/kalavai-phase1-410m-science-specialist-seed42 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 "mechramc/kalavai-phase1-410m-science-specialist-seed42" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mechramc/kalavai-phase1-410m-science-specialist-seed42",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "mechramc/kalavai-phase1-410m-science-specialist-seed42" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "mechramc/kalavai-phase1-410m-science-specialist-seed42",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use mechramc/kalavai-phase1-410m-science-specialist-seed42 with Docker Model Runner:
```
docker model run hf.co/mechramc/kalavai-phase1-410m-science-specialist-seed42
```

KALAVAI — Science Specialist (pythia-410m, seed 42)

Fine-tuned EleutherAI/pythia-410m on Science data as part of the KALAVAI decentralized cooperative training protocol.

Paper: KALAVAI: Predicting When Independent Specialist Fusion Works -- A Quantitative Model for Post-Hoc Cooperative LLM Training
GitHub: mechramc/Kalavai

Paper results

Phase 1 English domains. MoE fusion: +7.72% ±0.02pp over best specialist (3 seeds). Mean divergence 15.65%.

How to use

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("mechramc/kalavai-phase1-410m-science-specialist-seed42")
tokenizer = AutoTokenizer.from_pretrained("EleutherAI/pythia-410m")

This model is one specialist in a KALAVAI cooperative. To reproduce the MoE fusion results from the paper, load multiple domain specialists and combine them with a trained MoE router (see the paper and GitHub for details).

Citation

@article{kumaresan2026kalavai,
  title     = {{KALAVAI}: Predicting When Independent Specialist Fusion Works
               --- A Quantitative Model for Post-Hoc Cooperative {LLM} Training},
  author    = {Kumaresan, Ramchand},
  journal   = {arXiv preprint arXiv:2603.22755},
  year      = {2026},
  url       = {https://arxiv.org/abs/2603.22755}
}