nebulette/clip-l-sized-lfm-350m

LFM 350M for anime diffusion models

An encoder replacement for CLIP L. The tokenizer has remained unchanged.

The model has been verified for both the LFM (1024) and the CLIP (768) dimensions.

The English stop words have been removed from the captions.

The comma-seperated keywords were scrambled and adjusted for the model.

How it's done

The model was initialised using a low learning rate and natural-language captions from CC (Moondream) and Danbooru (Qwen3.5).

The masked language model then continued to learn from a combination of comma-separated and natural language texts.

The artists' given names, nicknames and fantasy names have been deliberately excluded from the dataset, as it would be difficult to predict these based on the text alone.

Changes

Compared to the 230M model, this has a cleaner, more recent dataset with more stable loss values. It's advisable to use this model.

Source data:

• anime_captions
• anime_faces_256px_v2
• artbench_captions
• cc12m_2mp_realistic
• danbooru_multitier_captions_202606
• dtg_character_tags
• gelbooru_characters_enriched
• gelbooru_tags_full
• live gelbooru feed, top 2k colors and verbs
• samples for a certain locked tag