Skip to content
ModelListedExternal
Download

Twscholar Lm Sft 7b Qlora R64

QLoRA SFT adapter for Qwen2.5-7B-Instruct on the twscholar-lm dataset — the SFT stage on its own, before DPO. Most users should prefer twscholar-lm-dpo-7b-final (SFT + DPO); this checkpoint is published for...

Sourcehuggingface.co/Yoda6131027/twscholar-lm-sft-7b-qlora-r64/tree/b6467e5945891916324157a0f013df8fb0eb9461
KindLoRABase modelQwen/Qwen2.5-7B-InstructVersionUnratedgrade

Model source

Kind
LoRA
File size
308 MB
Base model
Qwen/Qwen2.5-7B-Instruct
Version
vBF16
Updated
Nov 3, 2026
Downloads
0
Source
Hugging Face

Source description

twscholar-lm — SFT only (intermediate checkpoint)

QLoRA SFT adapter for Qwen2.5-7B-Instruct on the twscholar-lm dataset — the SFT stage on its own, before DPO. Most users should prefer twscholar-lm-dpo-7b-final (SFT + DPO); this checkpoint is published for reproducibility and for anyone studying the SFT-vs-SFT+DPO comparison directly.

  • r=64, alpha=128, 1 epoch (see the main model card / project repo for why 1 epoch was chosen over 3 — results/m4_ood_glitch_investigation.md)
  • Full project: https://github.com/oscarlin778/twscholar-lm
  • License: cc-by-nc-4.0 (inherited from the training dataset)

Sources

1 source
Hugging FacePrimaryVersion b6467e5945891916324157a0f013df8fb0eb9461
huggingface.co/Yoda6131027/twscholar-lm-sft-7b-qlora-r64/tree/b6467e5945891916324157a0f013df8fb0eb9461

Verified Nov 3

Download files

Repository files

File inventory only; completeness is unverified.

Model artifacts

2 artifacts

adapter_model.safetensors

safetensors · BF16 (unverified) · 308 MB · SHA-256 3b8edb4e95f9…22cc · Hugging Face

Download file

checkpoint-30/adapter_model.safetensors

safetensors · BF16 (unverified) · 308 MB · SHA-256 3b8edb4e95f9…22cc · Hugging Face

Download file

Source excerpts

2 excerpts
ReadmeAug 23huggingface.co/Yoda6131027/twscholar-lm-sft-7b-qlora-r64

--- license: cc-by-nc-4.0 base_model: Qwen/Qwen2.5-7B-Instruct tags: - lora - qlora - sft - traditional-chinese - academic-writing language: - zh --- # twscholar-lm — SFT only (intermediate checkpoint) QLoRA SFT adapter for Qwen2.5-7B-Instruct on the twscholar-lm dataset — the SFT stage on its own, before DPO. Most users should prefer `twscholar-lm-dpo-7b-final` (SFT + DPO); this checkpoint is published for reproducibility and for anyone studying the SFT-vs-SFT+DPO comparison directly. - r=64, alpha=128, 1 epoch (see the main model card / project repo for why 1 epoch was chosen over 3 — `results/m4_ood_glitch_investigation.md`) - Full project: https://github.com/oscarlin778/twscholar-lm - License: cc-by-nc-4.0 (inherited from the training dataset)

NoteAug 23huggingface.co/Yoda6131027/twscholar-lm-sft-7b-qlora-r64

Source context: 0 downloads · 0 likes · Repo Yoda6131027/twscholar-lm-sft-7b-qlora-r64