gustavecortal/Oneirogen-7B

Oneirogen-7B: A Specialized Model for Dream Generation and Analysis

Oneirogen-7B, developed by gustavecortal, is a 7.6 billion parameter language model built upon the Qwen2 architecture. Its core distinction lies in its specialized training on the DreamBank corpus, a collection of over 27,000 real dream narratives. This focused training enables Oneirogen-7B to understand and generate content within the unique domain of dreams.

Key Capabilities

• Novel Dream Narrative Generation: The primary function of Oneirogen-7B is to create new, coherent dream narratives, offering a unique tool for creative applications or psychological studies.
• Dream Analysis Foundation: Beyond generation, the model can serve as a base for advanced dream analysis tasks. For instance, it can be further fine-tuned on annotated datasets, such as Hall and Van de Castle annotations, to predict elements like characters and emotions within dream reports.
• Contextual Understanding of Dream Language: Its training on a vast corpus of dreams provides it with a deep understanding of the often surreal and symbolic language used in dream descriptions.

When to Use Oneirogen-7B

• Creative Writing: Ideal for generating unique dream sequences for stories, games, or artistic projects.
• Psychological Research: Useful for researchers studying dream content, potentially aiding in the automated analysis or synthesis of dream data.
• Exploration of Consciousness: Offers a computational tool to explore the nature of dream states and their linguistic representation.

Oneirogen-7B was also instrumental in creating The Android and The Machine, a dataset comprising 10,000 real and 10,000 generated dreams, showcasing its generation quality.