Ram07 / bitnet-1b

bitnet-1b-final

Converted BitNet model in SafeTensors format.

Model Details

• Parameters : 195,132
• Architecture : BitNet with BitLinear layers
• Quantization : Ternary weights (2-bit), 8-bit activations
• Format : SafeTensors (bfloat16)

Usage

from safetensors.torch import load_file
import torch

# Load the model
state_dict = load_file("model.safetensors")

# Initialize your BitNet model architecture
model = YourBitNetModel(config)  # Replace with your model initialization

# Load the weights
model.load_state_dict(state_dict)

# Use the model
model.eval()
with torch.no_grad():
    output = model(input_ids)

Files

• model.safetensors - Model weights in SafeTensors format
• config.json - Model configuration (if available)
• conversion_info.json - Details about the conversion

Conversion Details

• Converted on : 2025-10-03T10:51:07.866251
• Original format : PyTorch (.pt)
• Converter version : 1.0.0