StepLaw models are trained with various hyperparameter settings to enable research on scaling laws and hyperparameter optimization. This specific model was trained with learning rate 1.105e-02 and batch size 262144 for 305175 iterations, using a total of 80.0B training tokens.

Usage Example

Learning rate (lr): 1.105e-02
Batch size (bs): 262144
Training iterations: 305175
Training tokens (D): 80.0B

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "StepLaw/StepLaw-N_268M-D_79.0B-LR1.105e-02-BS262144"
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True, use_fast=False)
model = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True)

# Generate text
inputs = tokenizer("A long time ago in a galaxy far, far away", return_tensors="pt")
outputs = model.generate(**inputs, max_length=100)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))