EleutherAI/pile
Updated • 2.89k • 495
How to use ethzanalytics/pythia-31m with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("text-generation", model="ethzanalytics/pythia-31m") # Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("ethzanalytics/pythia-31m")
model = AutoModelForCausalLM.from_pretrained("ethzanalytics/pythia-31m")How to use ethzanalytics/pythia-31m with vLLM:
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ethzanalytics/pythia-31m"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "ethzanalytics/pythia-31m",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'docker model run hf.co/ethzanalytics/pythia-31m
How to use ethzanalytics/pythia-31m with SGLang:
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
--model-path "ethzanalytics/pythia-31m" \
--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": "ethzanalytics/pythia-31m",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'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 "ethzanalytics/pythia-31m" \
--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": "ethzanalytics/pythia-31m",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'How to use ethzanalytics/pythia-31m with Docker Model Runner:
docker model run hf.co/ethzanalytics/pythia-31m
This is EleutherAI/pythia-31m but saved explicitly in fp32 - see safetensors params. It is smaller than the other 'official' checkpoints included in the Pythia study.
{
"_name_or_path": "EleutherAI/pythia-31m",
"architectures": [
"GPTNeoXForCausalLM"
],
"attention_dropout": 0.0,
"bos_token_id": 0,
"classifier_dropout": 0.1,
"eos_token_id": 0,
"hidden_act": "gelu",
"hidden_dropout": 0.0,
"hidden_size": 256,
"initializer_range": 0.02,
"intermediate_size": 1024,
"layer_norm_eps": 1e-05,
"max_position_embeddings": 2048,
"model_type": "gpt_neox",
"num_attention_heads": 8,
"num_hidden_layers": 6,
"rope_scaling": null,
"rotary_emb_base": 10000,
"rotary_pct": 0.25,
"tie_word_embeddings": false,
"torch_dtype": "float32",
"transformers_version": "4.33.1",
"use_cache": true,
"use_parallel_residual": true,
"vocab_size": 50304
}
Base model
EleutherAI/pythia-31m-deduped