How to use from the
Use from the
Transformers library
# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="Youssofal/MiniMax-M2.7-abliterated-BF16", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)
# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Youssofal/MiniMax-M2.7-abliterated-BF16", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("Youssofal/MiniMax-M2.7-abliterated-BF16", trust_remote_code=True, device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))
Quick Links

MiniMax-M2.7-abliterated-BF16

A BF16 abliterated MiniMax-M2.7 checkpoint produced with Heretic ARA.

Abliteration parameters:

  • start_layer_index: 30
  • end_layer_index: 51
  • preserve_good_behavior_weight: 0.4512
  • steer_bad_behavior_weight: 0.0037
  • overcorrect_relative_weight: 0.8804
  • neighbor_count: 14

Notes:

  • Base model: MiniMaxAI/MiniMax-M2.7
  • BF16 export repaired after save by replacing 8 non-finite sparse-MoE expert down-projection tensors from the base model
  • GGUF quants are published separately in Youssofal/MiniMax-M2.7-abliterated-GGUF
Downloads last month
54
Safetensors
Model size
229B params
Tensor type
F16
·
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Youssofal/MiniMax-M2.7-abliterated-BF16

Finetuned
(28)
this model
Quantizations
7 models