StepLaw/StepLaw-N_536M-D_9.0B-LR3.453e-04-BS2097152

Wandb Model Name: step2v2_0618_h1280_ffnh9048_numh10_numl13_lr3.453e-04_bs1024_ti4768_mlr1.00e-05

This model is part of the StepLaw-N_536M-D_9.0B collection.

Model Specifications

Architecture

• Hidden size (H): 1280
• Feed-forward network size (FFN): 9048
• Attention heads: 10
• Layers: 13
• Parameter count: 536M

Training Parameters

• Learning rate (lr): 3.453e-04
• Batch size (bs): 2097152
• Training iterations: 4768
• Training tokens (D): 10.0B

Model Description

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 3.453e-04 and batch size 2097152 for 4768 iterations, using a total of 10.0B training tokens.

Usage Example

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "StepLaw/StepLaw-N_536M-D_9.0B-LR3.453e-04-BS2097152"
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))