@PyTorch: Now: @Shopify’s continual learning loop with PyTorch and @vLLM_project takes the stage at #PyTorchCon 2026 with Cody Ma…
Summary
Shopify 在 PyTorchCon 2026 主旨演讲中展示基于 PyTorch 与 vLLM 的持续学习闭环：校准评审模型标记低分输出，批评者诊断失败并通过重放修复，使小规模专用模型在特定任务上超越前沿模型，同时通过提示压缩优化推理。
View Cached Full Text
Cached at: 10/21/26, 02:34 PM
Now: @Shopify’s continual learning loop with PyTorch and @vLLM_project takes the stage at #PyTorchCon 2026 with Cody Mazza-Anthony @cmazzaanthony.
Using Buyer Profile as an example, Cody walks through how a calibrated judge flags low-scoring outputs in production. Critics diagnose each failure, and a replay with a critic hint is scored again by the judge. Accepted repairs expand and improve the training data.
Cody shows how a small specialist model can outperform a frontier model on this task, then turns to optimizing inference through prompt compression.
Watch the keynote livestream: https://events.linuxfoundation.org/pytorch-conference-north-america/program/keynote-livestream/…
Keynote Livestream | PyTorch Conference North America: The Deep Learning Conference
Source: https://events.linuxfoundation.org/pytorch-conference-north-america/program/keynote-livestream/ ThePyTorch Conference North America 2026Keynote livestream will take place during the following times:
Tuesday, October 20:
- Morning sessions starting at 9:00 AM
Wednesday, October 21:
- Morning sessions starting at 9:00 AM
All times are in Pacific Daylight Time (PDT), UTC-8. View the fullKeynote Schedulefor further details.
Sign up here to watch the livestream; check your inbox for an email with instructions on how to access the stream. If you don’t receive the email, or need assistance, contact[email protected].
Note: You will need to log in with your LFID to access the livestream. Don’t have an LFID? No problem!Create one now.
Similar Articles
@PyTorch: Simon Mo (@simon_mo_) of @inferact presents the state of @vllm_project at #PyTorchCon 2026, highlighting optimizations …
Simon Mo of Inferact presented the state of the vLLM project at PyTorchCon 2026, highlighting KV cache management and kernel optimizations across TPU, AMD, and NVIDIA hardware to address LLM performance and scalability challenges.
@PyTorch: vLLM (@vllm_project) is a high-throughput, memory-efficient inference and serving engine for LLMs. At #PyTorchCon North…
This post promotes vLLM, a high-throughput inference engine for LLMs, and details its sessions and activities at PyTorchCon North America.
@PyTorch: @marksaroufim Fewer training steps don’t necessarily mean faster training. At #PyTorchCon 2026 (@coreauto) examines the…
在 PyTorch Con 2026 上，Mark Saroufim 探讨了支撑更优优化器的线性代数内核，涵盖 QR、对称特征分解与 Cholesky，并从 GPU MODE 的线性代数竞赛中总结了如何通过暴露并行性、重叠独立阶段和合并更新来减少数据搬运。
@PyTorch: At #PyTorchCon 2026, Sara Hooker @sarahookr (@adaption_ai) examines the limits of scaling larger models and the shift t…
在 PyTorchCon 2026 上，Sara Hooker（Coatue/Adaption AI）发表主旨演讲，探讨扩展更大模型的局限以及向自适应智能的转变，涵盖后训练、测试时扩展、自适应计算和硬件协同设计，并展示 AutoScientist 相关案例。
@PyTorch: Live from Shanghai, @sparkycollier delivers his keynote, “Building Frontier Intelligence in the Open,” at PyTorch Confe…
Live from Shanghai, @sparkycollier delivers his keynote, “Building Frontier Intelligence in the Open,” at PyTorch Conference China 2026. Mark focuses on how PyTorch supports an open ecosystem where researchers, developers, model builders, and hardware companies can build and optimize AI across a growing range of models and compute. From open models to new accelerator architectures, Mark is exploring how open development expands who can build AI, helps breakthroughs move from experimentation to p