huihui_ai/deepscaler-abliterated

huihui_ai/ deepscaler-abliterated

1,204 Downloads Updated 11 months ago

A fine-tuned version of Deepseek-R1-Distilled-Qwen-1.5B that surpasses the performance of OpenAI’s o1-preview with just 1.5B parameters on popular math evaluations.

1.5b

ollama run huihui_ai/deepscaler-abliterated

curl http://localhost:11434/api/chat \
  -d '{
    "model": "huihui_ai/deepscaler-abliterated",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='huihui_ai/deepscaler-abliterated',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'huihui_ai/deepscaler-abliterated',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

View all →

Name

3 models

Size

Context

Input

deepscaler-abliterated:latest

7.1GB · 128K context window · Text · 11 months ago

deepscaler-abliterated:latest

7.1GB

128K

Text

deepscaler-abliterated:1.5b

latest

7.1GB · 128K context window · Text · 11 months ago

deepscaler-abliterated:1.5b latest

7.1GB

128K

Text

Readme

This is an uncensored version of agentica-org/DeepScaleR-1.5B-Preview created with abliteration (see remove-refusals-with-transformers to know more about it).
This is a crude, proof-of-concept implementation to remove refusals from an LLM model without using TransformerLens.

References

HuggingFace

Donation

Your donation helps us continue our further development and improvement, a cup of coffee can do it.

bitcoin:

  bc1qqnkhuchxw0zqjh2ku3lu4hq45hc6gy84uk70ge