HF transformers-compatible checkpoint + ONNX conversion
#1
by
Xenova
- opened
Hey there! Congrats on the model release. In case it's useful for others, I have uploaded a torch/transformers-compatible fp32 dequantized checkpoint at https://huggingface.co/Xenova/sweep-next-edit-1.5B, along with optimized ONNX conversions for the model. The file onnx/model_quantized.onnx should be the closest to this checkpoint, which uses 8-bits for weights.