diff --git a/README.md b/README.md index 8a3361f9f41..ce3ccac8606 100644 --- a/README.md +++ b/README.md @@ -26,6 +26,7 @@ ---- :fire: *News* :fire: +- [Jan 2025] Prepare and Serve Large-Scale Image Search with **Vector Database**: [**blog post**](https://blog.skypilot.co/large-scale-vector-database/) [**example**](./examples/vector_database/) - [Jan 2025] Launch and Serve **[DeepSeek-R1](https://github.com/deepseek-ai/DeepSeek-R1)** and **[Janus](https://github.com/deepseek-ai/DeepSeek-Janus)** on Kubernetes or Any Cloud: [**R1 example**](./llm/deepseek-r1/) and [**Janus example**](./llm/deepseek-janus/) - [Oct 2024] :tada: **SkyPilot crossed 1M+ downloads** :tada:: Thank you to our community! [**Twitter/X**](https://x.com/skypilot_org/status/1844770841718067638) - [Sep 2024] Point, Launch and Serve **Llama 3.2** on Kubernetes or Any Cloud: [**example**](./llm/llama-3_2/) @@ -187,7 +188,7 @@ Runnable examples: - [LocalGPT](./llm/localgpt) - [Falcon](./llm/falcon) - Add yours here & see more in [`llm/`](./llm)! -- Framework examples: [PyTorch DDP](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_distributed_torch.yaml), [DeepSpeed](./examples/deepspeed-multinode/sky.yaml), [JAX/Flax on TPU](https://github.com/skypilot-org/skypilot/blob/master/examples/tpu/tpuvm_mnist.yaml), [Stable Diffusion](https://github.com/skypilot-org/skypilot/tree/master/examples/stable_diffusion), [Detectron2](https://github.com/skypilot-org/skypilot/blob/master/examples/detectron2_docker.yaml), [Distributed](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_distributed_tf_app.py) [TensorFlow](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_app_storage.yaml), [Ray Train](examples/distributed_ray_train/ray_train.yaml), [NeMo](https://github.com/skypilot-org/skypilot/blob/master/examples/nemo/), [programmatic grid search](https://github.com/skypilot-org/skypilot/blob/master/examples/huggingface_glue_imdb_grid_search_app.py), [Docker](https://github.com/skypilot-org/skypilot/blob/master/examples/docker/echo_app.yaml), [Cog](https://github.com/skypilot-org/skypilot/blob/master/examples/cog/), [Unsloth](https://github.com/skypilot-org/skypilot/blob/master/examples/unsloth/unsloth.yaml), [Ollama](https://github.com/skypilot-org/skypilot/blob/master/llm/ollama), [llm.c](https://github.com/skypilot-org/skypilot/tree/master/llm/gpt-2), [Airflow](./examples/airflow/training_workflow) and [many more (`examples/`)](./examples). +- Framework examples: [Vector Database](./examples/vector_database/), [PyTorch DDP](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_distributed_torch.yaml), [DeepSpeed](./examples/deepspeed-multinode/sky.yaml), [JAX/Flax on TPU](https://github.com/skypilot-org/skypilot/blob/master/examples/tpu/tpuvm_mnist.yaml), [Stable Diffusion](https://github.com/skypilot-org/skypilot/tree/master/examples/stable_diffusion), [Detectron2](https://github.com/skypilot-org/skypilot/blob/master/examples/detectron2_docker.yaml), [Distributed](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_distributed_tf_app.py) [TensorFlow](https://github.com/skypilot-org/skypilot/blob/master/examples/resnet_app_storage.yaml), [Ray Train](examples/distributed_ray_train/ray_train.yaml), [NeMo](https://github.com/skypilot-org/skypilot/blob/master/examples/nemo/), [programmatic grid search](https://github.com/skypilot-org/skypilot/blob/master/examples/huggingface_glue_imdb_grid_search_app.py), [Docker](https://github.com/skypilot-org/skypilot/blob/master/examples/docker/echo_app.yaml), [Cog](https://github.com/skypilot-org/skypilot/blob/master/examples/cog/), [Unsloth](https://github.com/skypilot-org/skypilot/blob/master/examples/unsloth/unsloth.yaml), [Ollama](https://github.com/skypilot-org/skypilot/blob/master/llm/ollama), [llm.c](https://github.com/skypilot-org/skypilot/tree/master/llm/gpt-2), [Airflow](./examples/airflow/training_workflow) and [many more (`examples/`)](./examples). Case Studies and Integrations: [Community Spotlights](https://blog.skypilot.co/community/) diff --git a/docs/source/_gallery_original/applications/vector_database.md b/docs/source/_gallery_original/applications/vector_database.md new file mode 120000 index 00000000000..ebcd50df736 --- /dev/null +++ b/docs/source/_gallery_original/applications/vector_database.md @@ -0,0 +1 @@ +../../../../examples/vector_database/README.md \ No newline at end of file diff --git a/docs/source/_gallery_original/index.rst b/docs/source/_gallery_original/index.rst index e049a4ad322..8e0d0b16c35 100644 --- a/docs/source/_gallery_original/index.rst +++ b/docs/source/_gallery_original/index.rst @@ -50,6 +50,7 @@ Contents :maxdepth: 1 :caption: Applications + Image Vector Database Tabby: Coding Assistant LocalGPT: Chat with PDF diff --git a/docs/source/docs/index.rst b/docs/source/docs/index.rst index 2e9ca6859c6..ea5d6c6c18e 100644 --- a/docs/source/docs/index.rst +++ b/docs/source/docs/index.rst @@ -108,7 +108,7 @@ Runnable examples: * `LocalGPT `_ * Add yours here & see more in `llm/ `_! -* Framework examples: `PyTorch DDP `_, `DeepSpeed `_, `JAX/Flax on TPU `_, `Stable Diffusion `_, `Detectron2 `_, `Distributed `_ `TensorFlow `_, `NeMo `_, `programmatic grid search `_, `Docker `_, `Cog `_, `Unsloth `_, `Ollama `_, `llm.c `__, `Airflow `_ and `many more `_. +* Framework examples: `Vector Database `_, `PyTorch DDP `_, `DeepSpeed `_, `JAX/Flax on TPU `_, `Stable Diffusion `_, `Detectron2 `_, `Distributed `_ `TensorFlow `_, `NeMo `_, `programmatic grid search `_, `Docker `_, `Cog `_, `Unsloth `_, `Ollama `_, `llm.c `__, `Airflow `_ and `many more `_. Case Studies and Integrations: `Community Spotlights `_ diff --git a/examples/vector_database/README.md b/examples/vector_database/README.md index f127d2c176e..20581cc421c 100644 --- a/examples/vector_database/README.md +++ b/examples/vector_database/README.md @@ -4,7 +4,7 @@ VectorDB with SkyPilot

-### Large-Scale Image Search +## Large-Scale Image Search As the volume of image data grows, the need for efficient and powerful search methods becomes critical. Traditional keyword-based or metadata-based search often fails to capture the full semantic meaning in images. A vector database enables semantic search: you can find images that conceptually match a query (e.g., "a photo of a cloud") rather than relying on textual tags. In particular: @@ -17,7 +17,7 @@ SkyPilot streamlines the process of running such large-scale jobs in the cloud. Please find the complete blog post [here](https://blog.skypilot.co/large-scale-vector-database/) -### Step 0: Set Up The Environment +## Step 0: Set Up The Environment Install the following Prerequisites: * SkyPilot: Make sure you have SkyPilot installed and `sky check` should succeed. Refer to [SkyPilot’s documentation](https://docs.skypilot.co/en/latest/getting-started/installation.html) for instructions. * Hugging Face Token: To download dataset from Hugging Face Hub, you will need your token. Follow the steps below to configure your token. @@ -28,7 +28,7 @@ HF_TOKEN=hf_xxxxx ``` or set up the environment variable `HF_TOKEN`. -### Step 1: Compute Vectors from Image Data with OpenAI CLIP +## Step 1: Compute Vectors from Image Data with OpenAI CLIP You need to convert images into vector representations (embeddings) so they can be stored in a vector database. Models like [CLIP by OpenAI](https://openai.com/index/clip/) learn powerful representations that map images and text into the same embedding space. This allows for semantic similarity calculations, making queries like “a photo of a cloud” match relevant images. Use the following command to launch a job that processes your image dataset and computes the CLIP embeddings: @@ -51,7 +51,7 @@ You can also use `sky jobs queue` and `sky jobs dashboard` to see the status of SkyPilot Dashboard

-### Step 2: Construct the Vector Database from Computed Embeddings +## Step 2: Construct the Vector Database from Computed Embeddings Once you have the image embeddings, you need a specialized engine to perform rapid similarity searches at scale. In this example, we use [ChromaDB](https://docs.trychroma.com/getting-started) to store and query the embeddings. This step ingests the embeddings from Step 1 into a vector database to enable real-time or near real-time search over millions of vectors. To construct the database from embeddings: @@ -68,7 +68,7 @@ Processing batches: 100%|██████████| 1/1 [00:02<00:00, 2.39 Processing files: 100%|██████████| 12/12 [00:05<00:00, 2.04it/s]/1 [00:00