Article
From Zero to Reasoning Hero: How DeepSeek-R1 Leverages Reinforcement Learning to Master Complex Reasoning
By
•
•
11Bui Van Hop
hllj
AI & ML interests
Computer Vision, Deep Learning, NLP
Recent Activity
View all activity
Organizations
hllj's activity
upvoted
an
article
11 days ago
updated
2
datasets
12 days ago
published
a
dataset
12 days ago
liked
a
dataset
13 days ago
published
a
dataset
14 days ago
liked
2
datasets
5 months ago
updated
a
collection
6 months ago
updated
a
dataset
6 months ago
liked
2
datasets
7 months ago
upvoted
a
collection
7 months ago
upvoted
an
article
7 months ago
Article
Fine-tune Llama 3.1 Ultra-Efficiently with Unsloth
By
•
•
282
upvoted
a
collection
7 months ago
liked
2
datasets
7 months ago
liked
a
model
7 months ago
reacted to
kenshinn's
post with ❤️
7 months ago
Post
2031
Sparse MoE (SMoE) has an unavoidable drawback: the performance of SMoE heavily relies on the choice of hyper-parameters, such as the number of activated experts per token (top-k) and the number of experts.
Also, identifying the optimal hyper-parameter without a sufficient number of ablation studies is challenging. As the size of the models continues to grow, this limitation could result in a significant waste of computational resources, and in turn, could hinder the efficiency of training MoE-based models in practice.
(READ MORE ↓↓↓) Now, our DynMoE addresses these challenges! 🙌 DynMoE incorporates:
(1) a novel gating method that enables each token to automatically determine the number of experts to activate.
(2) An adaptive process automatically adjusts the number of experts during training. Extensive numerical results across Vision, Language, and Vision-Language tasks demonstrate the effectiveness of our approach to achieve competitive performance compared to GMoE for vision and language tasks, and MoE-LLaVA for vision-language tasks, while maintaining efficiency by activating fewer parameters.
Our code is available at https://github.com/LINs-lab/DynMoE, also see the checkpoints at LINs-lab/dynmoe-family-665ed5a331a7e84463cab01a
Also, identifying the optimal hyper-parameter without a sufficient number of ablation studies is challenging. As the size of the models continues to grow, this limitation could result in a significant waste of computational resources, and in turn, could hinder the efficiency of training MoE-based models in practice.
(READ MORE ↓↓↓) Now, our DynMoE addresses these challenges! 🙌 DynMoE incorporates:
(1) a novel gating method that enables each token to automatically determine the number of experts to activate.
(2) An adaptive process automatically adjusts the number of experts during training. Extensive numerical results across Vision, Language, and Vision-Language tasks demonstrate the effectiveness of our approach to achieve competitive performance compared to GMoE for vision and language tasks, and MoE-LLaVA for vision-language tasks, while maintaining efficiency by activating fewer parameters.
Our code is available at https://github.com/LINs-lab/DynMoE, also see the checkpoints at LINs-lab/dynmoe-family-665ed5a331a7e84463cab01a