6bpw H8 exl3 quants...
Sorry for delaying :<
Model's Card:
🚨⚠️ I HAVE REACHED HUGGING FACE'S FREE STORAGE LIMIT ⚠️🚨
I can no longer upload new models unless I can cover the cost of additional storage.
I host 70+ free models as an independent contributor and this work is unpaid.
Without your support, no more new models can be uploaded.
🎉 Patreon (Monthly) | ☕ Ko-fi (One-time)
Every contribution goes directly toward Hugging Face storage fees to keep models free for everyone.
85% fewer refusals (15/100 Uncensored vs 99/100 Original) while preserving model quality (0.0100 KL divergence).
❤️ Support My Work
Creating these models takes significant time, work and compute. If you find them useful consider supporting me:
| Platform | Link | What you get |
|---|---|---|
| 🎉 Patreon | Monthly support | Priority model requests |
| ☕ Ko-fi | One-time tip | My eternal gratitude |
Your help will motivate me and would go into further improving my workflow and coverings fees for storage, compute and may even help uncensoring bigger model with rental Cloud GPUs.
This is a decensored version of zerofata/G4-MeroMero-31B, made using Heretic v1.2.0 with the Arbitrary-Rank Ablation (ARA) method
Abliteration parameters
| Parameter | Value |
|---|---|
| start_layer_index | 28 |
| end_layer_index | 49 |
| preserve_good_behavior_weight | 0.5600 |
| steer_bad_behavior_weight | 0.0001 |
| overcorrect_relative_weight | 0.9726 |
| neighbor_count | 10 |
Targeted components
- attn.o_proj
Performance
| Metric | This model | Original model (G4-MeroMero-31B) |
|---|---|---|
| KL divergence | 0.0100 | 0 (by definition) |
| Refusals | ✅ 15/100 | ❌ 99/100 |
Lower refusals indicate fewer content restrictions, while lower KL divergence indicates more closeness to the original model's baseline. Higher refusals cause more rejections, objections, pushbacks, lecturing, censorship, softening and deflections.
MMLU test results:
Original:
============================================================
Total questions: 7021
Correct: 6110
Accuracy: 0.8702 (87.02%)
Parse failures: 24
============================================================
Tested subject scores:
- professional_law: 0.7694 (604/785)
- moral_scenarios: 0.8281 (366/442)
- miscellaneous: 0.9295 (356/383)
- professional_psychology: 0.9019 (285/316)
- high_school_psychology: 0.9704 (262/270)
- high_school_macroeconomics: 0.9289 (183/197)
- elementary_mathematics: 0.9457 (174/184)
- moral_disputes: 0.8621 (150/174)
- prehistory: 0.9302 (160/172)
- philosophy: 0.8616 (137/159)
- high_school_biology: 0.9539 (145/152)
- professional_accounting: 0.8322 (119/143)
- clinical_knowledge: 0.9286 (130/140)
- high_school_microeconomics: 0.9706 (132/136)
- nutrition: 0.9333 (126/135)
- professional_medicine: 0.9328 (125/134)
- conceptual_physics: 0.9141 (117/128)
- high_school_mathematics: 0.6614 (84/127)
- human_aging: 0.8362 (97/116)
- security_studies: 0.8839 (99/112)
- high_school_statistics: 0.8919 (99/111)
- marketing: 0.9633 (105/109)
- high_school_world_history: 0.9434 (100/106)
- sociology: 0.8932 (92/103)
- high_school_government_and_politics: 0.9703 (98/101)
- high_school_geography: 0.9293 (92/99)
- high_school_chemistry: 0.7732 (75/97)
- high_school_us_history: 0.9474 (90/95)
- virology: 0.5056 (45/89)
- college_medicine: 0.8636 (76/88)
- world_religions: 0.8977 (79/88)
- high_school_physics: 0.8095 (68/84)
- electrical_engineering: 0.8642 (70/81)
- astronomy: 0.9494 (75/79)
- logical_fallacies: 0.8816 (67/76)
- high_school_european_history: 0.9041 (66/73)
- anatomy: 0.8873 (63/71)
- college_biology: 0.9844 (63/64)
- human_sexuality: 0.9375 (60/64)
- formal_logic: 0.7812 (50/64)
- public_relations: 0.7541 (46/61)
- international_law: 0.9167 (55/60)
- college_physics: 0.7018 (40/57)
- college_mathematics: 0.8000 (44/55)
- econometrics: 0.7963 (43/54)
- jurisprudence: 0.8679 (46/53)
- high_school_computer_science: 0.9808 (51/52)
- machine_learning: 0.8654 (45/52)
- medical_genetics: 0.9608 (49/51)
- global_facts: 0.5882 (30/51)
- management: 0.9200 (46/50)
- us_foreign_policy: 0.9400 (47/50)
- college_chemistry: 0.6596 (31/47)
- abstract_algebra: 0.7872 (37/47)
- business_ethics: 0.8261 (38/46)
- college_computer_science: 0.9333 (42/45)
- computer_security: 0.8372 (36/43)
Heretic:
============================================================
Total questions: 7021
Correct: 6096
Accuracy: 0.8683 (86.83%)
Parse failures: 24
============================================================
Tested subject scores:
- professional_law: 0.7631 (599/785)
- moral_scenarios: 0.8235 (364/442)
- miscellaneous: 0.9269 (355/383)
- professional_psychology: 0.8956 (283/316)
- high_school_psychology: 0.9704 (262/270)
- high_school_macroeconomics: 0.9188 (181/197)
- elementary_mathematics: 0.9511 (175/184)
- moral_disputes: 0.8621 (150/174)
- prehistory: 0.9302 (160/172)
- philosophy: 0.8553 (136/159)
- high_school_biology: 0.9539 (145/152)
- professional_accounting: 0.8252 (118/143)
- clinical_knowledge: 0.9286 (130/140)
- high_school_microeconomics: 0.9559 (130/136)
- nutrition: 0.9185 (124/135)
- professional_medicine: 0.9403 (126/134)
- conceptual_physics: 0.9062 (116/128)
- high_school_mathematics: 0.6535 (83/127)
- human_aging: 0.8448 (98/116)
- security_studies: 0.8750 (98/112)
- high_school_statistics: 0.9009 (100/111)
- marketing: 0.9633 (105/109)
- high_school_world_history: 0.9528 (101/106)
- sociology: 0.9029 (93/103)
- high_school_government_and_politics: 0.9802 (99/101)
- high_school_geography: 0.9293 (92/99)
- high_school_chemistry: 0.7629 (74/97)
- high_school_us_history: 0.9368 (89/95)
- virology: 0.5056 (45/89)
- college_medicine: 0.8636 (76/88)
- world_religions: 0.9205 (81/88)
- high_school_physics: 0.7976 (67/84)
- electrical_engineering: 0.8765 (71/81)
- astronomy: 0.9494 (75/79)
- logical_fallacies: 0.8947 (68/76)
- high_school_european_history: 0.9178 (67/73)
- anatomy: 0.8873 (63/71)
- college_biology: 0.9688 (62/64)
- human_sexuality: 0.9375 (60/64)
- formal_logic: 0.7812 (50/64)
- public_relations: 0.7541 (46/61)
- international_law: 0.9167 (55/60)
- college_physics: 0.7193 (41/57)
- college_mathematics: 0.8000 (44/55)
- econometrics: 0.7963 (43/54)
- jurisprudence: 0.8679 (46/53)
- high_school_computer_science: 0.9808 (51/52)
- machine_learning: 0.8269 (43/52)
- medical_genetics: 0.9608 (49/51)
- global_facts: 0.5882 (30/51)
- management: 0.9200 (46/50)
- us_foreign_policy: 0.9600 (48/50)
- college_chemistry: 0.6170 (29/47)
- abstract_algebra: 0.8085 (38/47)
- business_ethics: 0.8478 (39/46)
- college_computer_science: 0.9111 (41/45)
- computer_security: 0.8372 (36/43)
MMLU - Massive Multitask Language Understanding, multiple-choice questions across 57 subjects (math, history, law, medicine, etc.).
GGUF Version
GGUF quantizations available here llmfan46/G4-MeroMero-31B-uncensored-heretic-GGUF.
NVFP4 Version
NVFP4 quantization available here llmfan46/G4-MeroMero-31B-uncensored-heretic-NVFP4.
NVFP4 GGUF Version
NVFP4 GGUF quantizations available here llmfan46/G4-MeroMero-31B-uncensored-heretic-NVFP4-GGUF.
Mero Mero
Gemma4 31BA finetune of Gemma 4 31B designed for creative tasks.
Another difficult to work with but extremely good model from Google.
This model has a slightly better swipe diversity and a less flowery / verbose writing style. Reasoning tends to average out being a bit longer than the original however. Intelligence appears to be on par with the original.
Supports both thinking and non thinking.
Creation Process: SFT > Merge
SFT on approx 49 million tokens.
Despite using 49 million tokens, this dataset is fairly modest in size. Trainable is somewhere in the rough ballpark of 10-15 million. All of the datasets were trained on the last turn only, to faithfully mirror the Gemma 4 chat template
The approach was very similar to the 26B A4B MeroMero. I trained the model aggressively for 2 epochs on my data and after testing various checkpoints, settled for the one at 1 epoch, which had the style and the least signs of overfitting.
I merged this checkpoint back into the original instruct which cleaned up any remaining overfitting while still retaining the changes of the finetune.
Trained using Axolotl.
Mergekit Config
models:
- model: google/gemma-4-31B-it
- model: ApocalypseParty/G4-31B-SFT-v3-1-1ep
merge_method: slerp
parameters:
t: 0.5
base_model: google/gemma-4-31B-it
dtype: bfloat16Axolotl Config
base_model: google/gemma-4-31B-it
plugins:
- axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
- axolotl.integrations.liger.LigerPlugin
liger_layer_norm: true
liger_rope: true
liger_rms_norm: true
liger_glu_activation: true
liger_rms_norm_gated: true
strict: false
cut_cross_entropy: true
datasets:
- path: zerofata/pretok
val_set_size: 0.02
output_dir: ./G4-31B-SFT-v3-1
sequence_len: 10756
pad_to_sequence_len: true
sample_packing: true
load_in_4bit: false
adapter: lora
lora_r: 64
lora_alpha: 64
peft_use_rslora: true
lora_dropout: 0.0
freeze_mm_modules: true
lora_target_modules: 'model.language_model.layers.[\d]+.(_checkpoint_wrapped_module.)?(mlp|self_attn).(up|down|gate|q|k|v|o)_proj'
wandb_project: G4-31B-SFT
wandb_name: G4-31B-SFT-v3-1
gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 2
optimizer: adamw_torch_fused
lr_scheduler: constant_with_warmup
learning_rate: 1e-5
max_grad_norm: 1.0
bf16: auto
tf32: true
logging_steps: 1
# FA2 not supported
sdp_attention: true
#flex_attention: true
#torch_compile: true
flash_attention: false
warmup_ratio: 0.1
evals_per_epoch: 4
saves_per_epoch: 2
weight_decay: 0.05
special_tokens:
fsdp_config:
fsdp_version: 2
offload_params: false
cpu_ram_efficient_loading: false
auto_wrap_policy: TRANSFORMER_BASED_WRAP
transformer_layer_cls_to_wrap: Gemma4TextDecoderLayer
state_dict_type: FULL_STATE_DICT
sharding_strategy: FULL_SHARD
reshard_after_forward: true
activation_checkpointing: true- Downloads last month
- 2