Orthogonalized Cot Model (Layer 17)

This model is an orthogonalized version of deepseek-ai/DeepSeek-R1-Distill-Llama-8B.

Model Details

Base Model: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
Model Type: Cot
Orthogonalization Layer: 17

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("kureha295/deepseek-ai-DeepSeek-R1-Distill-Llama-8B-ortho-cot-layer-17-harmless")
tokenizer = AutoTokenizer.from_pretrained("kureha295/deepseek-ai-DeepSeek-R1-Distill-Llama-8B-ortho-cot-layer-17-harmless")

Citation

If you use this model, please cite the original model and the orthogonalization method used.

Downloads last month: 1

Safetensors

Model size

8B params

Tensor type

F16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for kureha295/deepseek-ai-DeepSeek-R1-Distill-Llama-8B-ortho-cot-layer-17-harmless

Base model

deepseek-ai/DeepSeek-R1-Distill-Llama-8B

Finetuned

(156)

this model