Gemma 2 9B lineage
According to the model card, it is a fine-tuned version of unsloth/gemma-2-9b-it; tags also record that base with finetune lineage.
Open Source Model Profile · sam-paech
Delirium-v1
Delirium-v1 is a 9.24B-parameter Gemma 2 text-generation fine-tune from sam-paech. According to the model card, it was trained on an experimental Gutenberg plus antislop dataset with SIMPO.
- Publisher
- sam-paech
- Task
- text-generation
- Model type
- gemma2
- License
- gemma
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
Delirium-v1 is published by sam-paech as a text-generation model. The captured configuration identifies Gemma2ForCausalLM with gemma2 type, and Safetensors metadata reports 9,241,705,984 parameters. According to the model card, it is a fine-tune of unsloth/gemma-2-9b-it on sam-paech/gutenbergs_1_2_3_antislop-dpo.
Recorded capabilities
Gutenberg antislop data
According to the model card, training used an experimental Gutenberg plus antislop dataset combining prior Gutenberg sets with the publisher's gutenberg3 subset.
Reported SIMPO process
According to the model card, the model went through SIMPO training with reported validation loss 6.3949 and one epoch using Adam with cosine scheduling.
9.24B Transformers record
Safetensors metadata reports 9,241,705,984 parameters with transformers support and gemma licensing.
Use cases in the source record
- Creative-writing and dialogue experiments aligned with the publisher-listed creative and conversational tags.
- Stylized fiction trials building on the publisher-reported metaphor, voice, and dialogue strengths, with attention to its dark-tone limits.
Limitations and unknowns
- According to the model card, the publisher describes it as brutally dark and grotesque, fixated on repeated phrases, and seemingly unable to write romance or positivity.
- No Ethen-measured evaluation results were extracted; only publisher-reported training loss and reward figures were captured.
- No context-window value was extracted from this record.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: sam-paech/Delirium-v1
Captured: Unknown. Processed: 2026-09-07T19:34:57.317388+00:00.
Delirium v1 GGUFs here: https://huggingface.co/mradermacher/Delirium-v1-GGUF This model was cooked a bit too long during SIMPO training. It writes like Hunter S. Thompson 2 days into an ether binge. It's grotesque, dark, grimy and genius. It's trained on an experimental gutenberg + antislop dataset . This contains the original two gutenberg sets by jondurbin and nbeerbower , as well as a subset of my own set, gutenberg3 . The antislop pairs were generated with gemma-2-9b-it, with one sample generated with the AntiSlop sampler and the rejected sample generated without. Strengths: Interesting, natural dialogue & character voices Surpr…
F001F002F003F004F005F006F007F008F009F010F011F012F013F014F017F018