Add PytorchConf 2024

pytorchconf-2024/category.json

@@ -0,0 +1,3 @@
+{
+  "title": "PyTorch Conference 2024"
+}

pytorchconf-2024/videos/a-distributed-stateful-dataloader-for-large-scale-pretraining-davis-wertheimer-linsong-chu.json

@@ -0,0 +1,25 @@
+{
+  "description": "A Distributed Stateful Dataloader for Large-Scale Pretraining - Davis Wertheimer, IBM & Linsong Chu, IBM Research\n\nLarge-scale model pretraining crucially relies on specialized and dedicated dataloaders that can, for example, partition and stream data asynchronously across multiple processes and physical nodes. In this talk we discuss one of the torch-native dataloaders we built and use at IBM Research for addressing these needs. Intended for use in large-scale model pretraining, particularly in research settings where rapid iteration between datasets may be required, our dataloader is distributed, stateful, checkpointable, composable and rescalable \u2013 while remaining a simple extension of the existing PyTorch dataloading framework. It automatically and invisibly handles data sharding, shuffling, subdataset weighting, checkpoint saving and loading, and custom user-defined preprocessing functions, with minimal overhead and high throughput. We discuss these properties and how we achieved them, such as reducing overhead by implementing a custom LCG random number generator, and demonstrate proof of concept on production-scale training of a 7B parameter Llama model over 4 trillion tokens.",
+  "duration": 1433,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Davis Wertheimer",
+    "Linsong Chu"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/VtT4rdph4Qs/maxresdefault.webp",
+  "title": "A Distributed Stateful Dataloader for Large-Scale Pretraining",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=VtT4rdph4Qs"
+    }
+  ]
+}

pytorchconf-2024/videos/blobs-to-clips-efficient-end-to-end-video-data-loading-andrew-ho-ahmad-sharif-meta.json

@@ -0,0 +1,25 @@
+{
+  "description": "Blobs to Clips: Efficient End-to-End Video Data Loading - Andrew Ho & Ahmad Sharif, Meta\n\nThe PyTorch team has improved training speed by an order of magnitude for teams at Meta working on Small-to-Large-Scale MultiModal Video models. In this talk we\u2019ll share our learnings on reducing GPU starvation by overcoming data loading challenges such as dealing with large distributed datasets, worker imbalance, compute-bottlenecks due to parallel video decoding and sampling, checkpointing, and debuggability. As part of our commitment to open-source, we are releasing a new decoding library and updating existing PyTorch libraries on GitHub, and invite feedback and contributions from the community.",
+  "duration": 1544,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Andrew Ho",
+    "Ahmad Sharif"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/-cBplHNM4RA/maxresdefault.webp",
+  "title": "Blobs to Clips: Efficient End-to-End Video Data Loading",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=-cBplHNM4RA"
+    }
+  ]
+}

pytorchconf-2024/videos/building-pytorch-computer-vision-algorithms-for-100-skin-shades-emmanuel-acheampong-robomua.json

@@ -0,0 +1,24 @@
+{
+  "description": "Building PyTorch Computer Vision Algorithms for 100 Skin Shades - Emmanuel Acheampong, roboMUA\n\nAt roboMUA we're leading the charge in building predictive AI models for diverse skin shades with the use of Convolutional Neural Networks (CNNs), and harnessing the power of Generative Adversarial Networks (GANs) specifically for generating realistic images of black hairstyles. Our session showcases PyTorch's versatility in both predictive and generative tasks, offering a comprehensive approach to inclusive AI. For predictive AI models, we leverage PyTorch's flexible framework to develop CNNs. Through innovative techniques in feature engineering and model architecture design, we demonstrate how PyTorch enables accurate prediction across 100 skin shades. Simultaneously, we showcase the transformative potential of GANs in the realm of black hairstyles. By training GANs on a curated dataset of diverse hair textures and styles, we illustrate how PyTorch facilitates the generation of lifelike images that celebrate the beauty and diversity of black hair. Attendees will gain insights into the data preprocessing, model training, and evaluation processes and and learn how PyTorch empowers developers to build inclusive solutions.",
+  "duration": 882,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Emmanuel Acheampong"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/eGZ3-8ZbFj0/maxresdefault.webp",
+  "title": "Building PyTorch Computer Vision Algorithms for 100 Skin Shades",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=eGZ3-8ZbFj0"
+    }
+  ]
+}

pytorchconf-2024/videos/building-scientific-computing-infrastructure-software-with-the-pytorch-ecosystem-bharath-ramsundar.json

@@ -0,0 +1,24 @@
+{
+  "description": "Building Scientific Computing Infrastructure Software with the PyTorch Ecosystem - Bharath Ramsundar, Deep Forest Sciences\n\nThe DeepChem library is a scientific computing library that implements deep learning infrastructure for drug discovery, materials discovery, and biology. The DeepChem community is one of the largest scientific open source projects built in PyTorch, with over 5K stars on Github and thousands of citations. The DeepChem community has learned a number of useful lessons for building and maintaining high quality scientific code built on top of PyTorch. In this talk, I will share our learnings with the PyTorch community and also highlight opportunities for improving scientific support in the ecosystem.",
+  "duration": 1685,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Bharath Ramsundar"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/Ru-CDK_79NQ/maxresdefault.webp",
+  "title": "Building Scientific Computing Infrastructure Software with the PyTorch Ecosystem",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=Ru-CDK_79NQ"
+    }
+  ]
+}

pytorchconf-2024/videos/data-dependent-shapes-in-pt2-edward-yang-meta.json

@@ -0,0 +1,24 @@
+{
+  "description": "Data-Dependent Shapes in PT2 - Edward Yang, Meta\n\nData-dependent shapes are ubiquitous whenever you want to take advantage of sparsity in your data representation, whether it is in recommendation systems, mixture of experts or other use cases. We have made a lot of improvements to torch.compile's support for capturing and compiling data dependent shapes, but they also require some user knowledge to work with effectively. This talk will give an overview of PT2's facilities for data dependent compute and how to use them effectively.",
+  "duration": 1524,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Edward Yang"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/VYuFJUfyWSU/maxresdefault.webp",
+  "title": "Data-Dependent Shapes in PT2",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=VYuFJUfyWSU"
+    }
+  ]
+}

pytorchconf-2024/videos/dl-compiler-panel-discussion-p-tillet-j-ansel-j-pienaar-t-chen-m-zolotukhin-p-wu.json

@@ -0,0 +1,29 @@
+{
+  "description": "DL Compiler Panel Discussion - Philip Tillet, OpenAI; Jason Ansel, Meta; Jacques Pienaar, Google; Tianqi Chen, CMU & OctoAI; Mikhail Zolotukhin, Modular; Peng Wu, Meta\n\nSince the release of PyTorch 2 in 2023, torch.compile() has spurred significant new thinking around DL compiler designs at the framework level. In this session, we invite leaders in this space to share their insights based on real experiences of building DL compilers \u2013 Triton, TorchInductor, Halide, TVM, OpenXLA, and Mojo \u2013 and growing their ecosystems. We also invite a \u2018compiler user representative,\u2019 together.ai, to share their recent journey of redesigning the LLM inference stack around torch.compile(). Each leader will give a 10-minute lightning talk and an engaging panel discussion.",
+  "duration": 2131,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Philip Tillet",
+    "Jason Ansel",
+    "Jacques Pienaar",
+    "Tianqi Chen",
+    "Mikhail Zolotukhin",
+    "Peng Wu"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/YWDzHGx8PrY/maxresdefault.webp",
+  "title": "DL Compiler Panel Discussion",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=YWDzHGx8PrY"
+    }
+  ]
+}

pytorchconf-2024/videos/executorch-beta-and-on-device-generative-ai-support-mergen-nachin-mengtao-martin-yuan-meta.json

@@ -0,0 +1,25 @@
+{
+  "description": "ExecuTorch Beta and on-Device Generative AI Support - Mergen Nachin & Mengtao (Martin) Yuan, Meta\n\nDuring this session, we will discuss real-life case studies focusing on the productionization of PyTorch models onto edge devices and welcome the community to begin adopting ExecuTorch. Since announcing the ExecuTorch MVP at the previous PTC, we have made significant progress in terms of stability, model coverage, accelerator performance, and developer experience, reaching a milestone that marks the transition to beta status. In addition to the above improvements, we continue to support generative AI models. Since the alpha launch that initially enabled support for LLama2/3 models, we have now expanded our capabilities to include multimodal use cases and developed mobile demo apps showcasing these new features.",
+  "duration": 1213,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Mergen Nachin",
+    "Mengtao Yuan"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/45yNTi7c1Q0/maxresdefault.webp",
+  "title": "ExecuTorch Beta and on-Device Generative AI Support",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=45yNTi7c1Q0"
+    }
+  ]
+}

pytorchconf-2024/videos/hacks-to-make-llm-training-faster-daniel-han-unsloth-ai.json

@@ -0,0 +1,24 @@
+{
+  "description": "Hacks to Make LLM Training Faster - Daniel Han, Unsloth AI\n\nAs open-source LLMs have become more capable, a substantial ecosystem has developed around the fine-tuning of these models. A thriving community of researchers, developers, practitioners and hobbyists has emerged which focuses on topics ranging from memory efficiency, parameter-efficient fine-tuning and quantization to performance at scale and reproducible evaluations. The goal of this mini-summit is to bring this community together to discuss ideas, share knowledge and build connections.\n\nThe agenda features a keynote from Joe Spisak on the state of the Llama ecosystem followed by invited talks from the founders of Axolotl, Unsloth and torchtune. We conclude the summit with a riveting discussion on what\u2019s next for LLMs, fine-tuning and the PyTorch ecosystem with a fabulous panel of experts - Tim Dettmers (author of bitsandbytes and QLoRA), Hailey Schoelkopf (maintainer of LM Eval Harness at EleutherAI), Aakanksha Chowdhery (Lead author on PaLM and Gemini) and Alexis Conneau (Research Lead at OpenAI)",
+  "duration": 477,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Daniel Han"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/PdtKkc5jB4g/maxresdefault.webp",
+  "title": "Hacks to Make LLM Training Faster",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=PdtKkc5jB4g"
+    }
+  ]
+}

pytorchconf-2024/videos/halide-a-halide-backend-for-torchinductor-jason-ansel-meta.json

@@ -0,0 +1,24 @@
+{
+  "description": "[HALIDE] A Halide Backend for TorchInductor - Jason Ansel, Meta\n\nThis talk will focus on a new Halide backend for TorchInductor, which is in addition to the existing Triton and C++ backends.  The Halide backend is meant to serve as a reference backend to make it easier to extend TorchInductor to support new backend compilers and hardware devices.  Halide has been the inspiration (either in ideas or through forking) of numerous other compiler projects, so it is a good starting point for adding new backends that follow a Halide-like model.",
+  "duration": 489,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Jason Ansel"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/hCvlWZEXRrg/maxresdefault.webp",
+  "title": "[HALIDE] A Halide Backend for TorchInductor",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=hCvlWZEXRrg"
+    }
+  ]
+}

pytorchconf-2024/videos/implementing-a-custom-torch-compile-backend-a-case-study-maanav-dalal-yulong-wang-microsoft.json

@@ -0,0 +1,25 @@
+{
+  "description": "Implementing a Custom Torch.Compile Backend - A Case Study - Maanav Dalal & Yulong Wang, Microsoft\n\nThis presentation will dive into the development of the ONNXRuntime (ORT) backend for torch.compile. We'll cover the implementation process, starting with a PyTorch 2.0 generated FX graph, highlighting the unique challenges encountered when serving ORT-specific scenarios and how we solved them. Attendees will gain insights into optimizing performance, overcoming integration hurdles, and achieving efficient execution. Whether you're a developer looking to extend PyTorch's capabilities for your own use cases, keen to learn about ONNX Runtime, or interested in backend performance optimization, and the many steps we've taken to get to where we are now, this session promises valuable takeaways and practical knowledge.",
+  "duration": 666,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Maanav Dalal",
+    "Yulong Wang"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/MnNO_13fLtU/maxresdefault.webp",
+  "title": "Implementing a Custom Torch.Compile Backend - A Case Study",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=MnNO_13fLtU"
+    }
+  ]
+}

pytorchconf-2024/videos/intel-gpu-in-upstream-pytorch-expanding-gpu-choices-and-enhancing-back-eikan-wang-min-jean-cho.json

@@ -0,0 +1,25 @@
+{
+  "description": "Intel GPU in Upstream PyTorch: Expanding GPU Choices and Enhancing Backend Flexibility - Eikan Wang & Min Jean Cho, Intel\n\nThe integration of Intel GPU support into PyTorch marks a pivotal enhancement for PyTorch device and runtime. We generalized the PyTorch device and runtime to accommodate streaming devices. The generalization not only facilitates the deployment of PyTorch on ubiquitous hardware but also makes the integration of different HW backends easier. In addition, PyTorch with Intel GPU supports various Intel GPUs from the data center to the client. It enriches and democratizes PyTorch HW ecosystem. Particularly in AIPC scenarios where Intel's integrated and discrete GPUs are prevalent, Pytorch with Intel GPU can deliver promising performance and improved OOB experience in the AIPC domain that can extend PyTorch's applicability significantly.",
+  "duration": 1398,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Eikan Wang",
+    "Min Jean Cho"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/HRdda_kVEh4/maxresdefault.webp",
+  "title": "Intel GPU in Upstream PyTorch: Expanding GPU Choices and Enhancing Backend Flexibility",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=HRdda_kVEh4"
+    }
+  ]
+}

pytorchconf-2024/videos/keynote-building-an-advanced-knowledge-assistant-jerry-liu-co-founder-ceo-llamaindex.json

@@ -0,0 +1,24 @@
+{
+  "description": "Keynote: Building an Advanced Knowledge Assistant - Jerry Liu, Co-Founder & CEO, LlamaIndex\n\nA huge promise for LLMs is being able to answer questions and solve tasks of arbitrary complexity over an arbitrary number of data sources. The world has started to shift from simple RAG stacks, which are mostly good for answering pointed questions, to agents that can more autonomously reason over a diverse set of inputs, and interleave retrieval and tool use to produce sophisticated outputs.\n\nBuilding a reliable multi-agent system is challenging. There's a core question of developer ergonomics and production deployment - what makes sense outside a notebook setting. In this talk we outline some core building blocks for building advanced research assistants, including advanced RAG modules, event-driven workflow orchestration, and more.",
+  "duration": 896,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Jerry Liu"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/EjGdum0UdaI/maxresdefault.webp",
+  "title": "Keynote: Building an Advanced Knowledge Assistant",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=EjGdum0UdaI"
+    }
+  ]
+}

pytorchconf-2024/videos/keynote-community-awards.json

@@ -0,0 +1,22 @@
+{
+  "description": "Keynote: Community Awards",
+  "duration": 359,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/1dSVYKPk3N0/maxresdefault.webp",
+  "title": "Keynote: Community Awards",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=1dSVYKPk3N0"
+    }
+  ]
+}

pytorchconf-2024/videos/keynote-enabling-generative-ai-on-the-edge-cormac-brick-principal-engineer-google.json

@@ -0,0 +1,24 @@
+{
+  "description": "Keynote: Enabling Generative AI on the Edge - Cormac Brick, Principal Engineer, Google\n\nGenerative AI is no longer just in the cloud - recently it's also getting deployed on edge devices. A disruptive goal of this work is AI-powered applications that respond instantly, work offline, and protect user privacy by processing data locally. In this talk, we'll explore the cutting edge of edge-based generative AI, showcasing open models that are pushing the boundaries of what's possible today on the edge. We'll dive deep into the PyTorch ecosystem, looking at projects that are making it easier than ever to author, optimize, and deploy these models across a wide range of devices.",
+  "duration": 848,
+  "language": "eng",
+  "recorded": "2024-09-18",
+  "related_urls": [
+    {
+      "label": "Conference Website",
+      "url": "https://pytorch.org/event/pytorch-conference-2024/"
+    }
+  ],
+  "speakers": [
+    "Cormac Brick"
+  ],
+  "tags": [],
+  "thumbnail_url": "https://i.ytimg.com/vi_webp/n2DsEyLbuJg/maxresdefault.webp",
+  "title": "Keynote: Enabling Generative AI on the Edge",
+  "videos": [
+    {
+      "type": "youtube",
+      "url": "https://www.youtube.com/watch?v=n2DsEyLbuJg"
+    }
+  ]
+}