Modelpublic

atomicapple0/EAGLE-Llama-3.2-3B-Instruct-bf16

sourceHugging Faceapache-2.0updated 6mo agoView on Hugging Face

0likes154downloads

Model Card

EAGLE-Llama-3.2-3B-Instruct (bf16)

This repository contains a bf16 safetensors repackaging of `linborui/EAGLE-Llama-3.2-3B-Instruct`.

What changed

—Converted the checkpoint tensors from float32 to bfloat16.
—Updated config.json to advertise torch_dtype: bfloat16.
—Kept the original architecture and file layout, with the converted weights stored as model.safetensors.

Files

—config.json
—model.safetensors
—README.md

Loading

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "atomicapple0/EAGLE-Llama-3.2-3B-Instruct-bf16"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto")