metadata
license: apache-2.0
datasets:
- NewEden/Roleplay-Logs-Sharegpt-Ngram-cleaned
language:
- en
base_model:
- toasterai/Qwen2.5-7B-SambaV1DialogTesting-001
pipeline_tag: text-generation
tags:
- unsloth
- dialogue
ChatML as always. Full-precision this time. Quants will come later.
An experiment to make causal models more conversational. Many of them can already chat, but suffer from problems like occasional dullness, incoherence and verbatim repetition. This model DOES NOT follow assistant-style instructions and IS NOT INTENDED TO.