When Confidence Misleads: Suffix Anchoring and Anchor-Proximity Confidence Modulation for Diffusion Language Models
Paper β’ 2605.28181 β’ Published β’ 3
Seoul National University
university
Verified
AI & ML interests
None defined yet.
Recent Activity
View all activity
Papers
Can Natural Image Autoencoders Compactly Tokenize fMRI Volumes for Long-Range Dynamics Modeling?
Token Sparse Attention: Efficient Long-Context Inference with Interleaved Token Selection
JwonPΒ
submitted a
paper to Daily Papers about 2 months ago
napalnaΒ
published a
model 2 months ago
concarne2Β
submitted a
paper to Daily Papers 3 months ago
concarne2Β
authored 3
papers 3 months ago
Can Natural Image Autoencoders Compactly Tokenize fMRI Volumes for Long-Range Dynamics Modeling?
Paper β’ 2604.03619 β’ Published β’ 9
SEED: Towards More Accurate Semantic Evaluation for Visual Brain Decoding
Paper β’ 2503.06437 β’ Published
SwiFT: Swin 4D fMRI Transformer
Paper β’ 2307.05916 β’ Published
leey4nΒ
authored a
paper 5 months ago
taeyoun811Β
authored 2
papers 11 months ago
JwonPΒ
authored 3
papers 12 months ago
ReFlex: Text-Guided Editing of Real Images in Rectified Flow via Mid-Step Feature Extraction and Attention Adaptation
Paper β’ 2507.01496 β’ Published β’ 4
Soft Injection of Task Embeddings Outperforms Prompt-Based In-Context Learning
Paper β’ 2507.20906 β’ Published β’ 2
Cross-Attention Head Position Patterns Can Align with Human Visual Concepts in Text-to-Image Generative Models
Paper β’ 2412.02237 β’ Published β’ 1
Post
11435
# PyTorch == 2.5.0 Breaks Transformers' SDPAttention!
When you encounter "RuntimeError: cuDNN Frontend error: [cudnn_frontend] Error: No execution plans support the graph."
We can use workaround like this:
but this slow downs the performance gain from PyTorch 2.5.
Although it is fixed(not "fixed" but default option is turn-off the cuDNN SDPA) at here -- https://github.com/pytorch/pytorch/pull/138587 , but not released yet. (you need to install directly from source)
Fastest way for now : pip install "torch<2.5"
Ref: https://github.com/huggingface/diffusers/issues/9704#issuecomment-2422585273
When you encounter "RuntimeError: cuDNN Frontend error: [cudnn_frontend] Error: No execution plans support the graph."
We can use workaround like this:
torch.backends.cuda.enable_cudnn_sdp(False)but this slow downs the performance gain from PyTorch 2.5.
Although it is fixed(not "fixed" but default option is turn-off the cuDNN SDPA) at here -- https://github.com/pytorch/pytorch/pull/138587 , but not released yet. (you need to install directly from source)
Fastest way for now : pip install "torch<2.5"
Ref: https://github.com/huggingface/diffusers/issues/9704#issuecomment-2422585273
byeonghwikimΒ
authored 3
papers almost 2 years ago
Multi-Level Compositional Reasoning for Interactive Instruction Following
Paper β’ 2308.09387 β’ Published
Online Continual Learning For Interactive Instruction Following Agents
Paper β’ 2403.07548 β’ Published
Factorizing Perception and Policy for Interactive Instruction Following
Paper β’ 2012.03208 β’ Published
Post
19403
#TPU #PyTorch #Jax
When You're trying to use PyTorch or Jax on TPU,
for v2/v3/v4:
use tpu-ubuntu2204-base
for v5p:
use v2-alpha-tpuv5
for v5e:
use v2-alpha-tpuv5-lite
You must use these base images for the system to 'boot'.
Previously used tpu-vm-v4-pt-1.13 images might seem to start the VM, but SSH connections do not work.
I thought it was a firewall issue and spent a lot of time on it before realizing it was a problem with the boot image π₯²
https://cloud.google.com/tpu/docs/runtimes#pytorch_and_jax
When You're trying to use PyTorch or Jax on TPU,
for v2/v3/v4:
use tpu-ubuntu2204-base
for v5p:
use v2-alpha-tpuv5
for v5e:
use v2-alpha-tpuv5-lite
You must use these base images for the system to 'boot'.
Previously used tpu-vm-v4-pt-1.13 images might seem to start the VM, but SSH connections do not work.
I thought it was a firewall issue and spent a lot of time on it before realizing it was a problem with the boot image π₯²
https://cloud.google.com/tpu/docs/runtimes#pytorch_and_jax
Post
12352
π **InfiniTransformer, Gemma/Llama3 based Implementation!** π
> Update @ 2024.04.19: It now supports Llama-3!
> Note: this implementation is unofficial
This implementation is designed to handle virtually infinite context lengths.
Here's the github repo: https://github.com/Beomi/InfiniTransformer
π **Read the original Paper:** https://arxiv.org/abs/2404.07143
## **Focus on Infini-Attention**
- **2 Types of Implementation available:** Attention-layer only implementation / Model & Train-wise implementation
- **Fixed(segment dependent) Memory Usage:** Enables training on larger models and longer sequences without the memory overhead typical of standard Transformer implementations.
- **Infinite Context Capability:** Train with unprecedented sequence lengthsβimagine handling up to 1 million sequence lengths on standard hardware!
- You could train Gemma-2B with 1M sequence length with 2K segmentation size with single H100 GPU.
## **Try InfiniTransformer**
1. **Clone the repository:**
2. **Install necessary tools:**
3. **Dive Deep into Custom Training:**
- Train with extensive sequence lengths using scripts such as
for more detailed info, please visit Repo: https://github.com/Beomi/InfiniTransformer
Look forward to see your feedbacks! π
ps. Training loss plot is here π
> Update @ 2024.04.19: It now supports Llama-3!
> Note: this implementation is unofficial
This implementation is designed to handle virtually infinite context lengths.
Here's the github repo: https://github.com/Beomi/InfiniTransformer
π **Read the original Paper:** https://arxiv.org/abs/2404.07143
## **Focus on Infini-Attention**
- **2 Types of Implementation available:** Attention-layer only implementation / Model & Train-wise implementation
- **Fixed(segment dependent) Memory Usage:** Enables training on larger models and longer sequences without the memory overhead typical of standard Transformer implementations.
- **Infinite Context Capability:** Train with unprecedented sequence lengthsβimagine handling up to 1 million sequence lengths on standard hardware!
- You could train Gemma-2B with 1M sequence length with 2K segmentation size with single H100 GPU.
## **Try InfiniTransformer**
1. **Clone the repository:**
bash
git clone https://github.com/Beomi/InfiniTransformer
2. **Install necessary tools:**
bash
pip install -r requirements.txt
pip install -e git+https://github.com/huggingface/transformers.git@b109257f4f#egg=transformers
3. **Dive Deep into Custom Training:**
- Train with extensive sequence lengths using scripts such as
./train.gemma.infini.noclm.1Mseq.sh.for more detailed info, please visit Repo: https://github.com/Beomi/InfiniTransformer
Look forward to see your feedbacks! π
ps. Training loss plot is here π
- 2 replies
TaeksooΒ
authored a
paper over 2 years ago
byeonghwikimΒ
authored a
paper almost 3 years ago