agurung/colar-qwen25-7b-ff-post-rl

CoLAR Qwen2.5-7B Flawed Fictions Post-RL

This repository stores CoLAR exports in a Hugging Face-compatible layout. The repo root works for standard Transformers loading, and extra_state.pt preserves the latent head for latent decoding.

Current Revision

• Current tag: run-sqqdrtop
• Stage: post-RL
• Task: Flawed Fictions continuity error detection
• Compare slug: colar_ff_post_rl

Tagged Checkpoints

Tag	Local reference	Status
run-sqqdrtop	historical post-RL export	current commit

Files

• HF model files at repo root for standard decoding
• extra_state.pt for CoLAR latent decoding
• export_meta.json from the local export
• latent_metadata.json with archival provenance

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('agurung/colar-qwen25-7b-ff-post-rl', revision='run-sqqdrtop', torch_dtype='auto', device_map='auto')
tokenizer = AutoTokenizer.from_pretrained('agurung/colar-qwen25-7b-ff-post-rl', revision='run-sqqdrtop')

For latent decoding, download the same revision and use extra_state.pt together with the repo root model files.

Notes

• Initialized from the Flawed Fictions post-SFT checkpoint owf320j4.
• Original WandB source artifacts were not present on this machine during archival, so this HF upload keeps the exported model, export metadata, and synthesized archival metadata.