You can deploy this model and start earning money today!

Harvard-DCML/
boomerang-qwen3-4.9B

Name: boomerang-qwen3-4.9B
Availability: InStock

safetensors

qwen3

EleutherAI/the_pile_deduplicated

2510.05064

apache-2.0

Views

No views yet

Model Description

Boomerang distillation is a phenomenon in LLMs where we can distill a teacher model into a student and reincorporate teacher layers to create intermediate-sized models with no additional training. This is the student model distilled from Qwen3-8B-Base from our paper.

Training Procedure

This model was initialized from Qwen3-8B-Base by copying every other layer and the last 2 layers. It was distilled on 2.1B tokens of The Pile deduplicated with cross entropy, KL, and cosine loss to match the activations of Qwen3-8B-Base. We used the following hyperparameters:

Learning rate: 3e-4
Learning rate scheduler: cosine
Warmup ratio: 0.01
Optimizer: AdamW
Adam betas: (0.9, 0.95)
Adam epsilon: 1e-8
Weight decay: 0.1
Max. gradient norm: 1.0
Number of training steps: 500
Max. sequence length: 2048
Effective batch size: 2048
Mixed precision: bf16
KLDiv weight: 0.1
Cosine distance weight per layer: 0.1

Use

To interpolate between this model and Qwen3-8B-Base, please use the build_intermediate_model function from our github repository:

python3

import torch
from patching.patch import build_intermediate_model

intermediate_model = build_intermediate_model(
  teacher_name_or_path = "Qwen/Qwen3-8B-Base",
  student_name_or_path = "Harvard-DCML/boomerang-qwen3-4.9B",
  num_layers_to_patch = 2,
  patch_first_k_layers = False,
  dtype = torch.bfloat16,
)

Notes:

Changing num_layers_to_patch changes the size of the intermediate model by patching different numbers of student layers.
patch_first_k_layers should be set to False for this model for optimal interpolation performance.

Citation

@article{kangaslahti2025boomerang,
  title={Boomerang Distillation Enables Zero-Shot Model Size Interpolation},
  author={Kangaslahti, Sara and Nayak, Nihal V and Geuter, Jonathan and Fumero, Marco and Locatello, Francesco and Alvarez-Melis, David},
  journal={arXiv preprint arXiv:2510.05064},
  year={2025},
  url={https://arxiv.org/abs/2510.05064}
}