Tien Dung

tiendung

tiendung

AI & ML interests

None yet

Recent Activity

liked a model 21 days ago

g-group-ai-lab/gipformer-65M-rnnt

liked a model 26 days ago

hynt/Zipformer-30M-RNNT-6000h

liked a model 26 days ago

hataphu/zipformer-k2-rnn-lm-vi

View all activity

Organizations

liked a model 21 days ago

g-group-ai-lab/gipformer-65M-rnnt

Automatic Speech Recognition • Updated 25 days ago • 191 • 23

liked 2 models 26 days ago

hynt/Zipformer-30M-RNNT-6000h

Updated Feb 1 • 1.01k • 45

hataphu/zipformer-k2-rnn-lm-vi

Automatic Speech Recognition • Updated Feb 23 • 4

liked 2 models 2 months ago

mit-oasys/rlm-qwen3-8b-v0.1

8B • Updated Feb 20 • 317 • 57

nvidia/parakeet-ctc-0.6b-Vietnamese

Automatic Speech Recognition • Updated Feb 7 • 778 • 83

liked a model 4 months ago

FunAudioLLM/Fun-ASR-MLT-Nano-2512

Updated Dec 23, 2025 • 182 • 43

liked 2 datasets 4 months ago

NhutP/VietSpeech

Viewer • Updated Apr 25, 2025 • 1.03M • 691 • 31

nampdn-ai/vietspeech

Viewer • Updated Jul 10, 2024 • 272k • 81 • 3

liked a model 9 months ago

SparseLLM/BlockFFN-3B-SFT

Text Generation • Updated Jul 14, 2025 • 14 • 1

liked a model 10 months ago

turboderp/ERNIE-4.5-300B-A47B-PT-exl3

Updated Nov 1, 2025 • 10 • 3

reacted to Jaward's post with 😎👍 10 months ago

Post

2084

I played around with the new RXTX paper (XX^T) and was able to train nanogpt with 4x4 RXTX matmuls in both attention layer and optimizer🤕
It just works (well I had to add some guardrails) but still saves 5% of memory usage:
The Patch:
- Computes attention scores with a 4x4 blockwise RXTX matmuls (no pytorch dot prod)
- Handles arbitrary sequence lengths by padding to the nearest multiple of 4.
- An RXTX variant of shampoo with params reshaped into 4x4 blocks during each optimizer step.
- Uses 5% less ops
Code: https://github.com/Jaykef/ai-algorithms/blob/main/nanogpt-rxtx.ipynb
Paper: https://arxiv.org/pdf/2505.09814