Instructions to use EleutherAI/pythia-160m-weight-seed2-corrected with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use EleutherAI/pythia-160m-weight-seed2-corrected with Transformers:

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

pipe = pipeline("text-generation", model="EleutherAI/pythia-160m-weight-seed2-corrected")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("EleutherAI/pythia-160m-weight-seed2-corrected")
model = AutoModelForCausalLM.from_pretrained("EleutherAI/pythia-160m-weight-seed2-corrected", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use EleutherAI/pythia-160m-weight-seed2-corrected with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "EleutherAI/pythia-160m-weight-seed2-corrected"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "EleutherAI/pythia-160m-weight-seed2-corrected",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/EleutherAI/pythia-160m-weight-seed2-corrected

SGLang

How to use EleutherAI/pythia-160m-weight-seed2-corrected 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 "EleutherAI/pythia-160m-weight-seed2-corrected" \
    --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": "EleutherAI/pythia-160m-weight-seed2-corrected",
		"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 "EleutherAI/pythia-160m-weight-seed2-corrected" \
        --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": "EleutherAI/pythia-160m-weight-seed2-corrected",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use EleutherAI/pythia-160m-weight-seed2-corrected with Docker Model Runner:
```
docker model run hf.co/EleutherAI/pythia-160m-weight-seed2-corrected
```

Pythia-160M: corrected weight-seed 2

This model was trained from an independent seed-2 weight initialization while keeping the original Pythia training-data order fixed.

The earlier weight-seed releases had a training error: they continued from the original Pythia weights with incorrect Adam optimizer buffers, so they did not actually vary initialization. This release provides the intended independent initialization experiment.

The model was trained for 143,000 steps on the Pile with a global batch size of 1,024 sequences and a context length of 2,048 tokens.

The full set of 154 checkpoints, from step 0 through step 143,000, is available as step revisions.

This revision contains checkpoint step 143000.

Downloads last month: 609

Collection including EleutherAI/pythia-160m-weight-seed2-corrected

PolyPythias

Collection

61 items • Updated 2 days ago • 3