Predicting Memorization: Fine-tuned Models
Collection
33 items • Updated
How to use from
SGLangUse 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 "jaydeepb/pythia-1.4b-code-run2" \
--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": "jaydeepb/pythia-1.4b-code-run2",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'Quick Links
pythia-1.4b-code-run2
Pythia-1.4B fine-tuned for one epoch on a 1M-sequence (128-token) sample of The Stack.
This is a run2 model, fine-tuned on a 1M sample drawn disjoint from run1. It provides the fresh held-out data used to evaluate the memorization-prediction classifier (predict-before-training validation). The matching run1 model supplies the labels used to train the classifier.
- Base model, EleutherAI/pythia-1.4b
- Dataset, The Stack (bigcode/the-stack)
- Training, 1 epoch, seq len 128, learning rate 5e-5, cosine schedule, seed 42
- Downloads last month
- 308
Model tree for jaydeepb/pythia-1.4b-code-run2
Base model
EleutherAI/pythia-1.4b
Install from pip and serve model
# Install SGLang from pip: pip install sglang# Start the SGLang server: python3 -m sglang.launch_server \ --model-path "jaydeepb/pythia-1.4b-code-run2" \ --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": "jaydeepb/pythia-1.4b-code-run2", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'