XMemTransfer Qwen3.5-0.8B source memory (50M)

This repository contains the released Qwen3.5-0.8B source-memory artifact used in the XMemTransfer scaling experiments.

Project links

Files

  • memory_best.pt: best checkpoint selected during source training
  • memory_config.json: hash and memory configuration
  • source_adaptor_best.pt: source-side adaptor checkpoint used with the source memory
  • config.json: training configuration for the source-memory run

Intended use

This artifact is meant to be paired with Qwen target-side adaptors from the XMemTransfer project.

Provenance

  • source model family: Qwen3.5-0.8B
  • source corpus: WikiText
  • source token budget: 50M
  • released checkpoint: best selected source-memory checkpoint only

Usage and limitations

These files are research artifacts for the XMemTransfer implementation, not standalone Transformers pipeline checkpoints. Follow the project code to construct the memory reader and load the checkpoint. Results depend on the paired backbone, source memory, data preprocessing, and configuration released with the model.

Citation

Please cite the XMemTransfer paper when using these artifacts.

Downloads last month
5
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for OLAResearchX/xmemtransfer-qwen35-08b-source-50m

Finetuned
(100)
this model

Dataset used to train OLAResearchX/xmemtransfer-qwen35-08b-source-50m

Collection including OLAResearchX/xmemtransfer-qwen35-08b-source-50m

Paper for OLAResearchX/xmemtransfer-qwen35-08b-source-50m