Compact Math & Language Models
Collection
Lightweight Transformer checkpoints for symbolic 1D linear algebra (T5) and compact causal language modeling (Pythia-70M). โข 2 items โข Updated
How to use entity2260/pythia-70m with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("text-generation", model="entity2260/pythia-70m") # pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("entity2260/pythia-70m")
model = AutoModelForCausalLM.from_pretrained("entity2260/pythia-70m", device_map="auto")How to use entity2260/pythia-70m with vLLM:
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "entity2260/pythia-70m"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "entity2260/pythia-70m",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'docker model run hf.co/entity2260/pythia-70m
How to use entity2260/pythia-70m with SGLang:
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
--model-path "entity2260/pythia-70m" \
--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": "entity2260/pythia-70m",
"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 "entity2260/pythia-70m" \
--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": "entity2260/pythia-70m",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'How to use entity2260/pythia-70m with Docker Model Runner:
docker model run hf.co/entity2260/pythia-70m
This is a t5-small fine-tuned version on the math_dataset/algebra_linear_1d for solving algebra 1d equations mission.
To load the model: (necessary packages: !pip install transformers sentencepiece)
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("entity2260/pythia-70m")
model = AutoModelForCausalLM.from_pretrained("entity2260/pythia-70m")
You can then use this model to solve algebra 1d equations into numbers.
query = "Solve 0 = 1026*x - 2474 + 46592 for x"
input_text = f"{query} </s>"
features = tokenizer([input_text], return_tensors='pt')
model.to('cuda')
output = model.generate(input_ids=features['input_ids'].cuda(),
attention_mask=features['attention_mask'].cuda())
tokenizer.decode(output[0])
# <pad> -41</s>
Another examples: