Skip to content

Commit 0d6972e

Browse files
committed
jpg
Signed-off-by: Chris Abraham <[email protected]>
1 parent 0895fe8 commit 0d6972e

File tree

10 files changed

+9
-9
lines changed

10 files changed

+9
-9
lines changed

_posts/2025-05-02-pt-korea-user-group-recap.md

Lines changed: 9 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@ author: 'Jiho Kim, PyTorch Korea User Group'
77
At the end of March, the PyTorch Korea User Group hosted a special meetup that brought together prominent speakers for deep discussions on the PyTorch core and its broader ecosystem. With the event more than doubling in size compared to past gatherings, we were able to connect with even more developers and share insights. Huge thanks to [goorm](https://goorm.co/) for sponsoring the fantastic venue! 😄
88

99

10-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg1.png){:style="width:100%"}
10+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg1.jpg){:style="width:100%"}
1111

1212

1313

@@ -20,15 +20,15 @@ This recap is for those who couldn’t attend in person, as well as for particip
2020
Representing the PyTorch Foundation, part of the Linux Foundation, Jaeung provided an overview of how PyTorch is driving core open source technologies forward. He shared PyTorch's growth story, the many global projects currently in motion, and the ecosystem’s impressive 20%+ annual growth. The session also covered how the foundation operates, how member organizations are involved, and upcoming plans that are particularly useful for practitioners.
2121

2222

23-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg2.png){:style="width:100%"}
23+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg2.jpg){:style="width:100%"}
2424

2525

2626
## 2️⃣ Alban Desmaison | PyTorch Roadmap
2727

2828
Alban shared the design philosophy behind PyTorch and Meta’s official contribution roadmap ([link](https://dev-discuss.pytorch.org/t/meta-pytorch-team-2025-h1-roadmaps/2794)). He provided a deep technical dive into the differences between Eager and Compiled modes, especially breaking down the backend architecture of device Eager execution. Practical tools and improvements were also introduced—such as memory profilers, enhanced custom operator support, and pinned memory optimizations.
2929

3030

31-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg3.png){:style="width:100%"}
31+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg3.jpg){:style="width:100%"}
3232

3333

3434

@@ -37,15 +37,15 @@ Alban shared the design philosophy behind PyTorch and Meta’s official contribu
3737

3838
Rebellions is building runtime integration for their proprietary NPU architecture, fully aligned with the structural changes in PyTorch 2.0. This talk introduced the performance and scalability of their upcoming chip, their integration strategy with the PyTorch runtime, and challenges in supporting Eager Mode. Hongseok also previewed their roadmap toward releasing these features within the year.
3939

40-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg4.png){:style="width:100%"}
40+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg4.jpg){:style="width:100%"}
4141

4242

4343

4444
## 4️⃣ Kyujin Cho | Backend.AI: A Unified Platform for All AI Accelerators
4545

4646
Backend.AI abstracts and integrates various AI accelerators into a unified workflow. As the diversity of accelerator architectures grows, the need for portability and infrastructure unification becomes even more important. This session showcased features across development and operations—from NPU scheduling and resource allocation to monitoring. Backend.AI currently supports accelerators from NVIDIA, Intel, Tenstorrent, Rebellions, and more.
4747

48-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg5.png){:style="width:100%"}
48+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg5.jpg){:style="width:100%"}
4949

5050

5151

@@ -54,30 +54,30 @@ Backend.AI abstracts and integrates various AI accelerators into a unified workf
5454
This talk focused on the challenges of inference in real-world industrial applications of AI models. As new state-of-the-art models emerge rapidly, there’s a growing need for environments that can quickly validate device compatibility—ideally with one-click ease. NetsPresso is actively working on a static graph representation compatible with PyTorch, offering efficient support for model development, optimization, and testing.
5555

5656

57-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg6.png){:style="width:100%"}
57+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg6.jpg){:style="width:100%"}
5858

5959

6060
## 6️⃣ Jungyeop Lee | The Journey to Reproduce Deepseek-R1
6161

6262
Jungyeop took us through his journey of reproducing Deepseek, a large language model—an effort that involved 201 experiments. He shared real-world lessons from training with Korean data, tokenizer modifications, and fine-tuning strategies. His practical insights and next steps were especially valuable for those building or re-implementing large models from scratch.
6363

6464

65-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg7.png){:style="width:100%"}
65+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg7.jpg){:style="width:100%"}
6666

6767

6868
## 7️⃣ Sol Kim | A journey from TCP architecture to production-level LLMs
6969

7070
Sol presented an integrated optimization approach to deploying large models using the TCP(Tensor Contraction Processor) architecture, which supports tensor contraction at the hardware level. The talk highlighted optimization techniques built on hardware abstraction layers (HALs) and bottom-up integration strategies with PyTorch—offering a hybrid hardware-software perspective.
7171

7272

73-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg8.png){:style="width:100%"}
73+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg8.jpg){:style="width:100%"}
7474

7575
## 💡 Panel Talk & Q&A 💡
7676

7777
The event wrapped up with an engaging panel discussion. Attendees asked sharp questions, and the speakers offered insightful answers. It was a powerful moment that captured the community’s enthusiasm for PyTorch and their hunger for deeper technical understanding.
7878

7979

80-
![people at a conference](/assets/images/pt-korea-user-group-recap/fg9.png){:style="width:100%"}
80+
![people at a conference](/assets/images/pt-korea-user-group-recap/fg9.jpg){:style="width:100%"}
8181

8282

8383
## Final Thoughts
Loading
Loading
Loading
Loading
Loading
Loading
Loading
Loading
Loading

0 commit comments

Comments
 (0)