Phúc
@Phc14097930
Followers
20
Following
3K
Media
1
Statuses
59
I am a Research Intern at Qualcomm AI Research. I'm working on 3D computer vision and diffusion model.
Joined February 2020
RT @janusch_patas: Looking for a great 3DGS dev Discord? (Link below 👇). Look no further and join the MrNeRF & Brush Discord. We have ever….
0
8
0
You should checkout AnyCoder, very impressive.
Kimi k2 + groq in anycoder. Vibe coding 500+ loc Three.js mobile game in seconds
0
0
1
Excited to share minHiWave, my reimplementation of the HiWave paper! Clean code, minimal setup, and ready to run. Check it out:
HiWave. Training-Free High-Resolution Image Generation via Wavelet-Based Diffusion Sampling
2
6
32
RT @ziqi_huang_: 🎬 𝗖𝗩𝗣𝗥 𝟮𝟬𝟮𝟱 𝗧𝘂𝘁𝗼𝗿𝗶𝗮𝗹.𝙁𝙧𝙤𝙢 𝙑𝙞𝙙𝙚𝙤 𝙂𝙚𝙣𝙚𝙧𝙖𝙩𝙞𝙤𝙣 𝙩𝙤 𝙒𝙤𝙧𝙡𝙙 𝙈𝙤𝙙𝙚𝙡. 🚀 Hosted by MMLab@NTU × Kuaishou, etc.📅 June 11 | Nashville.🔗 h….
0
27
0
RT @natolambert: My path into AI.The sort of small wins that accumulate into a real career in AI. When I started grad school AI prof's didn….
interconnects.ai
How I got here. Building a career brick by brick over 8 years.
0
31
0
To easily understand the math in the paper, I suggest you follow this great blog post by @Jianlin_S:
1
1
4
I've implemented a simple version of LoRA-Pro (for educational purposes). You can use it with PEFT to finetune models using LoRA. Currently, my implementation doesn't support multi-gpu training. You can give it a try here:
github.com
A simple implementation of optimizer proposed in LoRA-Pro paper for Low-rank finetuning - laihongphuc/lora_pro
LoRAってちゃんと学習できてるの？.実は「LoRAの学習 = 仮想的な等価なlow rank勾配でのフル学習」→ これを使ってAとBの勾配を調整してフル学習の勾配に近づける→このとき AとBの更新は最適なclosed formな形があることを示した。おもしろい！
1
13
72
RT @gabrielpeyre: Stein's unbiased risk estimate (SURE) is an almost magical formula that enables the computation of the mean squared error….
0
136
0
RT @Michael_J_Black: Build what you need and use what you build. This is a core philosophy of my research. It shifts the focus away from pu….
perceiving-systems.blog
0
260
0
RT @zswitten: I've been dying to shill this harder for six months, and now that Anthropic API is GA, I finally can. The Prompt Engineeri….
docs.google.com
0
67
0
RT @soumithchintala: reading "AI News" (previously Smol Talk) is probably the highest-leverage 45 mins I spend everyday on catching up with….
0
72
0
RT @StasBekman: When finetuning a model has anyone experimented with first running with LR=0 for some 100-1000 iterations to get the optimi….
0
12
0
RT @karpathy: Seeing as I published my Tokenizer video yesterday, I thought it could be fun to take a deepdive into the Gemma tokenizer.….
0
450
0
RT @GoogleAI: Most of an LLM’s memory & compute are consumed by matrix multiplication operations. Today on the blog, learn about techniques….
0
268
0
RT @francoisfleuret: Since these experiments have been popular, here is a recap that will be from now the thread for updates. The motivati….
0
23
0
RT @gblazex: Everybody should pay attention to this guy. 3 models trending + even MistralTrix was trained based on his Colab notebook. He a….
github.com
Course to get into Large Language Models (LLMs) with roadmaps and Colab notebooks. - mlabonne/llm-course
0
93
0
RT @dhh: “This level of purpose picking is available to everyone. Anyone can choose to be needed, and thus choose to be happy. That’s the c….
world.hey.com
Victor Frankl wrote Man’s Search for Meaning after surviving a concentration camp during World War II. He observed the outer extreme of what happens to people who no longer have a WHY to live for....
0
63
0
RT @mervenoyann: Read the MobileSAM paper this weekend 📖 Sharing some insights 🧶 . The idea 💡: SAM model consist of three parts, a heavy im….
0
25
0
RT @fiandola: 🚨 New paper!🚨.A few months ago, Meta released Segment Anything Model (SAM). It's already in photography apps, medicine, and v….
0
267
0
RT @marksaroufim: This is a good question, it gets to the root of the tradeoff between performance and flexibility so how do PyTorch folks….
0
82
0