Instructions to use deqing/llama-window-2-old with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deqing/llama-window-2-old with Transformers:

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

pipe = pipeline("text-generation", model="deqing/llama-window-2-old")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("deqing/llama-window-2-old")
model = AutoModelForCausalLM.from_pretrained("deqing/llama-window-2-old")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use deqing/llama-window-2-old with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deqing/llama-window-2-old"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "deqing/llama-window-2-old",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/deqing/llama-window-2-old

SGLang

How to use deqing/llama-window-2-old 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 "deqing/llama-window-2-old" \
    --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": "deqing/llama-window-2-old",
		"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 "deqing/llama-window-2-old" \
        --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": "deqing/llama-window-2-old",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use deqing/llama-window-2-old with Docker Model Runner:
```
docker model run hf.co/deqing/llama-window-2-old
```

convergent-llama-300M-muon-window_2

A 300M-parameter language model trained from scratch on FineWeb-Edu 10BT (~9.4B tokens, 1 epoch) as part of the Convergent Evolution project, which investigates how Fourier features emerge in LLM number embeddings.

Model details


Architecture	LLaMA-style Transformer (12 layers, 1024 hidden, 16 heads, GQA)
Parameters	~300M
Optimizer	Muon (for 2D weights) + AdamW (for embeddings/bias/norm)
Data perturbation	window-2 context (bigram-level context only)
Training data	FineWeb-Edu sample-10BT (~9.4B tokens)
Context length	1024
Tokenizer	Llama 3 (128K vocab)
Batch size	512 sequences

Training dynamics

Intermediate checkpoints are saved as branches: tokens-200M, tokens-400M, ..., tokens-9.6B.

from transformers import AutoModelForCausalLM

# Load final checkpoint
model = AutoModelForCausalLM.from_pretrained("deqing/convergent-llama-300M-muon-window_2")

# Load intermediate checkpoint (e.g., at 1B tokens)
model = AutoModelForCausalLM.from_pretrained("deqing/convergent-llama-300M-muon-window_2", revision="tokens-1B")

Citation

Paper forthcoming.

Downloads last month: 33

Safetensors

Model size

0.3B params

Tensor type

F32

deqing
/

llama-window-2-old

convergent-llama-300M-muon-window_2

Model details

Training dynamics

Citation

Dataset used to train deqing/llama-window-2-old