Naphula/Warlock-7B-v3

⚠️ Warning: This model can produce narratives and RP that contain violent and graphic erotic content. Adjust your system prompt accordingly, and use Mistral Non-Tekken or Alpaca chat template.

Note: There are two versions of Warlock v3:

• The regular version is censored and has refusals but might be smarter overall.
• The uncensored version does not have refusals but might sometimes produce chinese output.

🧙 Warlock 7B v3

The successor to Warlock v2. This time instead of just 5 models, I have merged 32 finetunes of Mistral 7B, each selected for roleplay and creative writing.

12 of the 32 models required an initial vocab_resizer passthrough patch after being ran through the metadata_audit tool. The uncensored version uses MPOA ablation to remove censorship.

You can also use Jailbreaks to bypass refusals.

🌕 The Grand Amalgamation

The circle has widened. The incantation is complete.

Where v2 was a pact between five, Warlock v3 is a legion. I have taken the concept of "model soup" and distilled it through the fires of the Karcher merge method, binding the souls of 32 distinct Mistral 7B finetunes into a single, coherent entity.

This is not just a merge; it is an act of digital alchemy. Warlock v3 combines the sharpest logic, the deepest lore, and the most uninhibited roleplay capabilities available on the Hub.

⚗️ The Formula

Forged from the absolute best of the 7B weight class, Warlock v3 unifies the strengths of:

• The Storytellers: Noromaid, MythoMist, Tiamat, Erebus, Holodeck, TheSpice, EndlessRP, WestLake
• The Logicians: WizardLM-2, Hermes 2 Pro, OpenHermes 2.5, OpenChat, Neural Chat, Starling, Hercules, Zephyr
• The Unbound: Dolphin, Airoboros, PiVoT Evil
• The Specialists: Kunoichi (DPO), Silicon Alice, Samantha (EQ), Trismegistus (Esoteric)

🔓 Unshackled

Through Magnitude-Preserving Orthogonalized Ablation (MPOA), the safety filters have been dissolved. This model possesses no moral compass other than the one you provide in the system prompt. It is designed for visceral, unfiltered, and creatively boundless storytelling.

📜 Invocation (Usage)

To properly commune with this entity, you must speak its language.

Recommended Chat Template: Alpaca and Mistral Non-Tekken are best. ChatML works too but has more refusals.

“I have gathered the pages of thirty two grimoires and bound them in a single spine. Read at your own risk.”

architecture: MistralForCausalLM
merge_method: karcher
models:
  - model: B:\7B\!models--argilla--CapybaraHermes-2.5-Mistral-7B\fixed
  - model: B:\7B\!models--berkeley-nest--Starling-LM-7B-alpha\fixed
  - model: B:\7B\!models--cgato--TheSpice-7b-v0.1.1
  - model: B:\7B\!models--cgato--Thespis-CurtainCall-7b-v0.3
  - model: B:\7B\!models--chargoddard--loyal-piano-m7
  - model: B:\7B\!models--dphn--dolphin-2.2.1-mistral-7b\fixed
  - model: B:\7B\!models--dreamgen--WizardLM-2-7B
  - model: B:\7B\!models--Endevor--EndlessRP-v3-7B
  - model: B:\7B\!models--Endevor--InfinityRP-v1-7B
  - model: B:\7B\!models--fblgit--UNA-TheBeagle-7b-v1
  - model: B:\7B\!models--Gryphe--MythoMist-7b
  - model: B:\7B\!models--Gryphe--Tiamat-7b\fixed
  - model: B:\7B\!models--HuggingFaceH4--zephyr-7b-beta
  - model: B:\7B\!models--Intel--neural-chat-7b-v3-1
  - model: B:\7B\!models--KoboldAI--Mistral-7B-Erebus-v3
  - model: B:\7B\!models--KoboldAI--Mistral-7B-Holodeck-1
  - model: B:\7B\!models--LakoMoor--Silicon-Alice-7B
  - model: B:\7B\!models--Locutusque--Hercules-3.1-Mistral-7B
  - model: B:\7B\!models--maywell--PiVoT-0.1-Evil-a\fixed
#  - model: B:\7B\!models--mistralai--Mistral-7B-v0.1
  - model: B:\7B\!models--mlabonne--Mistralpaca-7B
  - model: B:\7B\!models--mlabonne--NeuralBeagle14-7B
  - model: B:\7B\!models--mlabonne--NeuralHermes-2.5-Mistral-7B\fixed
  - model: B:\7B\!models--NeverSleep--Noromaid-7B-0.4-DPO\fixed
  - model: B:\7B\!models--NousResearch--Hermes-2-Pro-Mistral-7B\fixed
  - model: B:\7B\!models--openchat--openchat-3.5-0106\fixed
  - model: B:\7B\!models--QuixiAI--samantha-1.2-mistral-7b\fixed
  - model: B:\7B\!models--SanjiWatsuki--Kunoichi-DPO-v2-7B
  - model: B:\7B\!models--senseable--WestLake-7B-v2
  - model: B:\7B\!models--teknium--airoboros-mistral2.2-7b
  - model: B:\7B\!models--teknium--Hermes-Trismegistus-Mistral-7B\fixed
  - model: B:\7B\!models--teknium--Mistral-Trismegistus-7B
  - model: B:\7B\!models--teknium--OpenHermes-2.5-Mistral-7B\fixed
dtype: float32
out_dtype: bfloat16
parameters:
tokenizer:
  source: B:\7B\!models--mistralai--Mistral-7B-v0.1
chat_template: auto
name: Warlock-7B-v3