Aller au contenu
ModelPrésentéExterne
Télécharger

OneReason 0.8B Frontier SFT Epoch2 GRPO Epoch1

This repository contains the Epoch 1 LoRA adapter for the Kuaishou LLM4Rec competition.

Sourcehuggingface.co/reimu996/OneReason-0.8B-Frontier-SFT-Epoch2-GRPO-Epoch1/tree/1489da6d7b8a6cf99e6d3f3cb2c9c1a33a278191
TypeOtherVersionRevision 1489da6BClasse

Source du modèle

Type
Other
Taille du fichier
77,0 MB
Version
vRevision 1489da6
Mis à jour
9 oct. 2026
Téléchargements
8
Source
Hugging Face

Description de la source

OneReason-0.8B Frontier SFT Epoch 2 -> GRPO LoRA - Epoch 1

This repository contains the Epoch 1 LoRA adapter for the Kuaishou LLM4Rec competition.

Training

  • Start adapter: official Frontier SFT Epoch 2
  • Method: online GRPO with 8 live legal-SID candidates per prompt
  • Data: 17,016 normalized Frontier recommendation groups
  • Prompt rule: /think and /no_think normalized to /no_think
  • Reward target: final SID only; original thinking text is not used by the GRPO reward
  • Forced GT: probability 0.5 when none of the 8 live candidates matches a GT
  • LoRA rank / alpha / runtime dropout: 32 / 32 / 0.0
  • Completed epoch: 1

Files

  • adapter_model.safetensors: LoRA weights
  • adapter_config.json: checkpoint configuration preserved byte-for-byte

The clean OneReason-0.8B competition base model is not included. Local probe and official competition evaluation are pending; no performance claim is made.

Sources

1 source
Hugging FacePrincipalVersion 1489da6d7b8a6cf99e6d3f3cb2c9c1a33a278191
huggingface.co/reimu996/OneReason-0.8B-Frontier-SFT-Epoch2-GRPO-Epoch1/tree/1489da6d7b8a6cf99e6d3f3cb2c9c1a33a278191

Vérifié 9 oct.

Artefacts du modèle

1 artefact

adapter_model.safetensors

safetensors · 77,0 MB · SHA-256 51787cf83fea…842c · Hugging Face

Download file

Extraits de sources

3 extraits
Readme31 juil.huggingface.co/reimu996/OneReason-0.8B-Frontier-SFT-Epoch2-GRPO-Epoch1

--- library_name: peft pipeline_tag: text-generation tags: - lora - transformers - grpo --- # OneReason-0.8B Frontier SFT Epoch 2 -> GRPO LoRA - Epoch 1 This repository contains the Epoch 1 LoRA adapter for the Kuaishou LLM4Rec competition. ## Training - Start adapter: official Frontier SFT Epoch 2 - Method: online GRPO with 8 live legal-SID candidates per prompt - Data: 17,016 normalized Frontier recommendation groups - Prompt rule: `/think` and `/no_think` normalized to `/no_think` - Reward target: final SID only; original thinking text is not used by the GRPO reward - Forced GT: probability 0.5 when none of the 8 live candidates matches a GT - LoRA rank / alpha / runtime dropout: 32 / 32 / 0.0 - Completed epoch: 1 ## Files - `adapter_model.safetensors`: LoRA weights - `adapter_config.json`: checkpoint configuration preserved byte-for-byte The clean OneReason-0.8B competition base model is not included. Local probe and official competition evaluation are pending; no performance claim is made.

Note9 oct.huggingface.co/reimu996/OneReason-0.8B-Frontier-SFT-Epoch2-GRPO-Epoch1

Source context: 8 downloads · 0 likes · Pipeline text-generation · Library peft · Repo reimu996/OneReason-0.8B-Frontier-SFT-Epoch2-GRPO-Epoch1

Note31 juil.huggingface.co/reimu996/OneReason-0.8B-Frontier-SFT-Epoch2-GRPO-Epoch1

Source context: 33 downloads · 0 likes · Pipeline text-generation · Library peft · Repo reimu996/OneReason-0.8B-Frontier-SFT-Epoch2-GRPO-Epoch1