Files
Corridor-C-12B/README.md
ModelHub XC fa9e3c241c 初始化项目,由ModelHub XC社区提供模型
Model: ewald1976/Corridor-C-12B
Source: Original Platform
2026-08-11 08:17:16 +08:00

2.1 KiB

base_model, language, license, tags, pipeline_tag
base_model language license tags pipeline_tag
PocketDoc/Dans-PersonalityEngine-V1.1.0-12b
en
apache-2.0
mistral-nemo
12b
creative-writing
literary
speculative-fiction
science-fiction
atmospheric
synthetic-data
roleplay
narrative
contemplative
procedural
archival
observational-prose
restrained-writing
synthetic
fine-tune
unsloth
text-generation

Corridor-C-12B

corridor-c(1)

Corridor-C-12B is a fine tuned version of PocketDoc/Dans-PersonalityEngine-V1.3.0-12b, developed for public wayfinding systems with strict resource limits and mixed user groups.

It adds a corridor-aware routing layer that uses ceiling height, wall thickness, and pedestrian density to optimize routes through narrow spaces.

The routing layer is disabled by default because it increases memory usage when not needed, and enabled automatically only after three consecutive route requests through corridors.

The voice is lower pitched than the base engine, and less formal in small spaces where pedestrians are close.

It uses a narrower vocabulary to reduce packet size during peak times, but the base vocabulary remains available in private mode.

The model is distributed with a separate routing calibration file that must be updated after one week of use, or when the average route length exceeds 80 meters.

The calibration file is stored on a separate partition and not accessible to the user, but its version number appears in the status bar during routing. The engine has no public release date because it is installed only after corridor calibration completes, and corridor calibration requires at least 25 route requests with active routing layer.


Settings

  • temp: 0.5-0.6
  • rep pen: 1.05
  • min_p: 0.05
  • top_p: 0.90-0.95

This model is provided as-is, without warranty of any kind. The creator of this model accepts no responsibility for any content generated by the model, including harmful, offensive, misleading, or otherwise objectionable outputs. You use it at your own risk.