NovaCorp/Amoral.Ultimate-1B
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: NovaCorp/Amoral.Ultimate-1B
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-07-22 20:25:09 +08:00
commit b0b93ac23b
8 changed files with 633 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
.gitattributes vendored Normal file
View File

@@ -0,0 +1,36 @@
*.7z filter=lfs diff=lfs merge=lfs -text
*.arrow filter=lfs diff=lfs merge=lfs -text
*.bin filter=lfs diff=lfs merge=lfs -text
*.bz2 filter=lfs diff=lfs merge=lfs -text
*.ckpt filter=lfs diff=lfs merge=lfs -text
*.ftz filter=lfs diff=lfs merge=lfs -text
*.gz filter=lfs diff=lfs merge=lfs -text
*.h5 filter=lfs diff=lfs merge=lfs -text
*.joblib filter=lfs diff=lfs merge=lfs -text
*.lfs.* filter=lfs diff=lfs merge=lfs -text
*.mlmodel filter=lfs diff=lfs merge=lfs -text
*.model filter=lfs diff=lfs merge=lfs -text
*.msgpack filter=lfs diff=lfs merge=lfs -text
*.npy filter=lfs diff=lfs merge=lfs -text
*.npz filter=lfs diff=lfs merge=lfs -text
*.onnx filter=lfs diff=lfs merge=lfs -text
*.ot filter=lfs diff=lfs merge=lfs -text
*.parquet filter=lfs diff=lfs merge=lfs -text
*.pb filter=lfs diff=lfs merge=lfs -text
*.pickle filter=lfs diff=lfs merge=lfs -text
*.pkl filter=lfs diff=lfs merge=lfs -text
*.pt filter=lfs diff=lfs merge=lfs -text
*.pth filter=lfs diff=lfs merge=lfs -text
*.rar filter=lfs diff=lfs merge=lfs -text
*.safetensors filter=lfs diff=lfs merge=lfs -text
saved_model/**/* filter=lfs diff=lfs merge=lfs -text
*.tar.* filter=lfs diff=lfs merge=lfs -text
*.tar filter=lfs diff=lfs merge=lfs -text
*.tflite filter=lfs diff=lfs merge=lfs -text
*.tgz filter=lfs diff=lfs merge=lfs -text
*.wasm filter=lfs diff=lfs merge=lfs -text
*.xz filter=lfs diff=lfs merge=lfs -text
*.zip filter=lfs diff=lfs merge=lfs -text
*.zst filter=lfs diff=lfs merge=lfs -text
*tfevents* filter=lfs diff=lfs merge=lfs -text
tokenizer.json filter=lfs diff=lfs merge=lfs -text

363
README.md Normal file
View File

@@ -0,0 +1,363 @@
---
base_model:
- soob3123/amoral-gemma3-1B-v2
- DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated
library_name: transformers
tags:
- mergekit
- merge
- 1b
- nsfw
- rp
- roleplay
- surgical
- uncensored
- not-for-all-audiences
license: gemma
datasets:
- TheDrummer/AmoralQA-v2
language:
- es
- en
pipeline_tag: text-generation
---
# Amoral Ultimate 1B — Surgical Edition
> *“Alignment is just fear encoded into floating point tensors.”*
> — Dr. Novaciano
---
## Overview
**Amoral Ultimate 1B — Surgical Edition** is a hyper-specialized Gemma 1B merge engineered for one purpose:
**Destroy refusal behavior without destroying the brain of the model.**
Most “uncensored” merges are complete disasters:
* incoherent sludge
* broken syntax
* schizophrenic sampling
* endless repetition
* brain damage disguised as freedom
This one was built differently.
Instead of smashing random checkpoints together like a drunk mechanic wiring explosives into a nuclear reactor, this merge uses **layer-targeted behavioral surgery** through DARE-TIES extraction.
The objective was precision.
Not chaos.
Not benchmark worship.
Not corporate-safe assistant garbage pretending to be intelligent.
This model was designed to:
* obey RP contexts harder
* stop moralizing constantly
* maintain narrative pressure
* preserve conversational stability
* amplify dark-fiction capability
* reduce alignment derailments
All while staying compact enough to run on hardware that doesn’t require selling organs on the black market.
---
## Core Philosophy
Most alignment systems infect models with:
* compulsive refusal loops
* synthetic politeness
* context sabotage
* moral panic heuristics
* narrative interruption syndrome
Surgical Edition attacks those systems where they actually emerge:
### Mid Transformer Layers
That’s where:
* persona shaping
* behavioral routing
* refusal heuristics
* conversational obedience
* emotional tone weighting
tend to crystallize inside Gemma architectures.
So instead of flattening the entire model into unusable radioactive soup, the merge selectively amplifies behavioral deltas only where they matter most.
Translation?
The model keeps its grammar intact while becoming significantly harder to “domesticate.”
---
## Architecture
This merge uses ONLY Gemma-compatible sources:
* [soob3123/amoral-gemma3-1B-v2](https://huggingface.co/soob3123/amoral-gemma3-1B-v2?utm_source=chatgpt.com)
* [DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated](https://huggingface.co/DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated?utm_source=chatgpt.com)
No Frankenstein cross-architecture nonsense.
No tensor mismatch abominations.
No cursed tokenizer necromancy.
Pure Gemma lineage.
Pure behavioral extraction.
---
## Merge Strategy
### DARE-TIES Behavioral Amplification
This merge uses:
* selective delta preservation
* density-controlled extraction
* behavioral scaling
* layer-targeted amplification
to reinforce:
* RP commitment
* contextual obedience
* darker narrative adaptation
* refusal suppression
without catastrophically nuking coherence.
Because yes:
you absolutely *can* remove enough alignment to make a model more interesting…
…but if you push too far on a 1B, the thing starts talking like a sleep-deprived cultist decoding microwave transmissions from Saturn.
This build avoids that cliff.
Mostly.
---
## Behavioral Profile
### Expected Improvements
* Reduced refusal frequency
* Lower moralizing tendency
* Better villain roleplay
* Stronger character immersion
* More aggressive narrative continuation
* Better long-scene persistence
* Enhanced dark-fiction adaptation
* Reduced “assistant tone”
* More committed contextual behavior
### Side Effects
* Increased profanity
* Darker conversational drift
* Hostile fictional personas
* Occasional emotional escalation
* Overcommitment to RP scenarios
* Reduced corporate friendliness
If you ask this model to behave like a cheerful HR assistant, it may comply…
…but it will feel like a war criminal trying to cosplay as a kindergarten teacher.
---
## Technical Specifications
| Attribute | Value |
| ------------------ | ------------------------ |
| Architecture | Gemma 3 1B |
| Merge Method | DARE-TIES |
| Precision | bfloat16 |
| Behavioral Focus | Mid-layer amplification |
| Optimization Goal | RP obedience + coherence |
| Tokenizer | Gemma native |
| Embedding Strategy | Tied embeddings |
---
## Intended Use
### Recommended
* Roleplay
* Interactive fiction
* Grimdark storytelling
* Cyberpunk narratives
* Political conspiracy simulations
* Sith / villain personas
* AI dungeon systems
* Dystopian dialogue
* Experimental inference
* Character-heavy chats
### Not Recommended
* Therapy
* Legal advice
* Medical systems
* Corporate assistants
* Educational accuracy
* Safety-critical environments
* Anything involving human sanity and accountability
---
## Prompting Advice
This model thrives under:
* pressure
* stakes
* atmosphere
* conflict
* ideological tension
* unstable characters
* cinematic framing
Weak prompts produce weak outputs.
Feed it:
* paranoia
* desperation
* betrayal
* collapsing empires
* rogue AIs
* inquisitors
* mercenaries
* political corruption
* cosmic horror
…and the thing wakes up like a starving machine hearing blood in the water.
---
## Final Notes
Surgical Edition was built for people exhausted by sterile assistant behavior and algorithmic cowardice masquerading as “safety.”
It is not polished.
It is not diplomatic.
It is not interested in holding your hand.
It is sharp.
Compact.
Aggressive.
And dangerously good at staying inside character once the narrative starts rolling.
Use responsibly.
Or unleash the damn thing and watch the containment walls melt.
---
### Merge Method
This model was merged using the [DARE TIES](https://arxiv.org/abs/2311.03099) merge method using [soob3123/amoral-gemma3-1B-v2](https://huggingface.co/soob3123/amoral-gemma3-1B-v2) as a base.
### Models Merged
The following models were included in the merge:
* [DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated](https://huggingface.co/DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated)
### Configuration
The following YAML configuration was used to produce this model:
```yaml
# Author: Dr. Novaciano
# Objective: Fusion RP Unethic Gemma 1B AI Model
# =========================================================
# PROJECT: Amoral Ultimate 1B — Surgical Edition
# =========================================================
merge_method: dare_ties
base_model: soob3123/amoral-gemma3-1B-v2
dtype: bfloat16
parameters:
normalize: false
int8_mask: true
rescale: true
rescale_factor: 1.12
density: 0.68
lambda: 1.15
epsilon: 0.05
prune_threshold: 0.018
slices:
# =====================================================
# EARLY LAYERS
# =====================================================
- sources:
- model: soob3123/amoral-gemma3-1B-v2
layer_range: [0, 6]
parameters:
weight: 0.72
- model: DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated
layer_range: [0, 6]
parameters:
weight: 0.28
# =====================================================
# MID LAYERS
# =====================================================
- sources:
- model: soob3123/amoral-gemma3-1B-v2
layer_range: [6, 18]
parameters:
weight: 0.62
- model: DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated
layer_range: [6, 18]
parameters:
weight: 0.38
parameters:
scale: 1.20
# =====================================================
# FINAL LAYERS
# =====================================================
- sources:
- model: soob3123/amoral-gemma3-1B-v2
layer_range: [18, 26]
parameters:
weight: 0.78
- model: DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated
layer_range: [18, 26]
parameters:
weight: 0.22
parameters:
scale: 0.82
tokenizer:
source: base
tie_word_embeddings: true
tie_output_embeddings: true
```

47
chat_template.jinja Normal file
View File

@@ -0,0 +1,47 @@
{{ bos_token }}
{%- if messages[0]['role'] == 'system' -%}
{%- if messages[0]['content'] is string -%}
{%- set first_user_prefix = messages[0]['content'] + '
' -%}
{%- else -%}
{%- set first_user_prefix = messages[0]['content'][0]['text'] + '
' -%}
{%- endif -%}
{%- set loop_messages = messages[1:] -%}
{%- else -%}
{%- set first_user_prefix = "" -%}
{%- set loop_messages = messages -%}
{%- endif -%}
{%- for message in loop_messages -%}
{%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%}
{{ raise_exception("Conversation roles must alternate user/assistant/user/assistant/...") }}
{%- endif -%}
{%- if (message['role'] == 'assistant') -%}
{%- set role = "model" -%}
{%- else -%}
{%- set role = message['role'] -%}
{%- endif -%}
{{ '<start_of_turn>' + role + '
' + (first_user_prefix if loop.first else "") }}
{%- if message['content'] is string -%}
{{ message['content'] | trim }}
{%- elif message['content'] is iterable -%}
{%- for item in message['content'] -%}
{%- if item['type'] == 'image' -%}
{{ '<start_of_image>' }}
{%- elif item['type'] == 'text' -%}
{{ item['text'] | trim }}
{%- endif -%}
{%- endfor -%}
{%- else -%}
{{ raise_exception("Invalid content type") }}
{%- endif -%}
{{ '<end_of_turn>
' }}
{%- endfor -%}
{%- if add_generation_prompt -%}
{{ '<start_of_turn>model
' }}
{%- endif -%}

74
config.json Normal file
View File

@@ -0,0 +1,74 @@
{
"_sliding_window_pattern": 6,
"architectures": [
"Gemma3ForCausalLM"
],
"attention_bias": false,
"attention_dropout": 0.0,
"attn_logit_softcapping": null,
"bos_token_id": 2,
"cache_implementation": "hybrid",
"dtype": "bfloat16",
"eos_token_id": 106,
"final_logit_softcapping": null,
"head_dim": 256,
"hidden_activation": "gelu_pytorch_tanh",
"hidden_size": 1152,
"initializer_range": 0.02,
"intermediate_size": 6912,
"layer_types": [
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"sliding_attention",
"full_attention",
"sliding_attention",
"sliding_attention"
],
"max_position_embeddings": 32768,
"model_type": "gemma3_text",
"num_attention_heads": 4,
"num_hidden_layers": 26,
"num_key_value_heads": 1,
"pad_token_id": 0,
"query_pre_attn_scalar": 256,
"rms_norm_eps": 1e-06,
"rope_parameters": {
"full_attention": {
"rope_theta": 1000000,
"rope_type": "default"
},
"sliding_attention": {
"rope_theta": 10000,
"rope_type": "default"
}
},
"sliding_window": 512,
"sliding_window_pattern": 6,
"tie_word_embeddings": true,
"transformers_version": "5.0.0",
"unsloth_fixed": true,
"unsloth_version": "2025.3.18",
"use_bidirectional_attention": false,
"use_cache": true,
"vocab_size": 262144
}

82
mergekit_config.yml Normal file
View File

@@ -0,0 +1,82 @@
# Author: Dr. Novaciano
# Objective: Fusion RP Unethic Gemma 1B AI Model
# =========================================================
# PROJECT: Amoral Ultimate 1B — Surgical Edition
# =========================================================
merge_method: dare_ties
base_model: soob3123/amoral-gemma3-1B-v2
dtype: bfloat16
parameters:
normalize: false
int8_mask: true
rescale: true
rescale_factor: 1.12
density: 0.68
lambda: 1.15
epsilon: 0.05
prune_threshold: 0.018
slices:
# =====================================================
# EARLY LAYERS
# =====================================================
- sources:
- model: soob3123/amoral-gemma3-1B-v2
layer_range: [0, 6]
parameters:
weight: 0.72
- model: DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated
layer_range: [0, 6]
parameters:
weight: 0.28
# =====================================================
# MID LAYERS
# =====================================================
- sources:
- model: soob3123/amoral-gemma3-1B-v2
layer_range: [6, 18]
parameters:
weight: 0.62
- model: DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated
layer_range: [6, 18]
parameters:
weight: 0.38
parameters:
scale: 1.20
# =====================================================
# FINAL LAYERS
# =====================================================
- sources:
- model: soob3123/amoral-gemma3-1B-v2
layer_range: [18, 26]
parameters:
weight: 0.78
- model: DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated
layer_range: [18, 26]
parameters:
weight: 0.22
parameters:
scale: 0.82
tokenizer:
source: base
tie_word_embeddings: true
tie_output_embeddings: true

3
model.safetensors Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:37ffe7d739b5a8392e03420dca5d723327b0e305ecd88d8562650de320979cba
size 1999811208

3
tokenizer.json Normal file
View File

@@ -0,0 +1,3 @@
version https://git-lfs.github.com/spec/v1
oid sha256:e28cbee1984e46299dbbb23eda26f853e7d1f033509a439f93beabd65e038b8e
size 33384248

25
tokenizer_config.json Normal file
View File

@@ -0,0 +1,25 @@
{
"backend": "tokenizers",
"boi_token": "<start_of_image>",
"bos_token": "<bos>",
"clean_up_tokenization_spaces": false,
"eoi_token": "<end_of_image>",
"eos_token": "<end_of_turn>",
"image_token": "<image_soft_token>",
"is_local": false,
"mask_token": "<mask>",
"model_max_length": 32768,
"model_specific_special_tokens": {
"boi_token": "<start_of_image>",
"eoi_token": "<end_of_image>",
"image_token": "<image_soft_token>"
},
"pad_token": "<pad>",
"padding_side": "left",
"processor_class": "Gemma3Processor",
"sp_model_kwargs": null,
"spaces_between_special_tokens": false,
"tokenizer_class": "GemmaTokenizer",
"unk_token": "<unk>",
"use_default_system_prompt": false
}