E2E CloudModel Repository FAQ
For step-by-step fixes, see Troubleshooting Model Repository.
Choosing and creating
Section titled “Choosing and creating”What is the difference between a Model Repository and a Model Endpoint?
Section titled “What is the difference between a Model Repository and a Model Endpoint?”A repository stores model files. A Model Endpoint is a running service that answers requests using files from a repository. See Use a repository from a Model Endpoint.
Can I use a Model Repository without a Model Endpoint?
Section titled “Can I use a Model Repository without a Model Endpoint?”Yes. You can use a repository only to store, version, and share model files, and you can use it from instances and notebooks. See Use a repository from an instance or notebook.
Which file formats can I store?
Section titled “Which file formats can I store?”Any. The upload panel says “All file types accepted”. You can store PyTorch models (.pth, .pt), Hugging Face models, TensorRT engines, ONNX models, Triton repositories, or any custom format. TensorFlow SavedModel and Keras files can be stored too. See Upload model files.
Which Model Type do I pick for a TensorFlow model?
Section titled “Which Model Type do I pick for a TensorFlow model?”Any type. Model Type has no TensorFlow entry, and a repository takes any file. Pick PyTorch for PyTorch models and Hugging Face models, Triton for Triton Inference Server model repositories, and Custom for other formats. See Create a repository.
Which storage option should I choose?
Section titled “Which storage option should I choose?”Choose New EOS Bucket when you are starting fresh: TIR creates and manages the bucket and its keys. Choose Existing EOS Bucket to reuse storage you already have, so that model files can sit with data you keep in one bucket. Choose External EOS Bucket to use files from an E2E bucket that another account owns. For that, you need the bucket name and a key pair that can read and write it. See Use an existing or external bucket.
Can I use a bucket from another storage provider?
Section titled “Can I use a bucket from another storage provider?”No. An External EOS Bucket must be an E2E Object Storage bucket in the same region as the repository. See Use an existing or external bucket.
Can I change the storage type after I create a repository?
Section titled “Can I change the storage type after I create a repository?”No. Create a new repository with the storage you want. See Create a repository.
Can I rename a repository?
Section titled “Can I rename a repository?”No. You set a repository’s name once, so choose one you want to keep. See Create a repository.
Can I encrypt a repository?
Section titled “Can I encrypt a repository?”No. The Create Model Repository form has no encryption choice. See Create a repository.
Which regions have Model Repository?
Section titled “Which regions have Model Repository?”Delhi and Chennai. A repository belongs to one project and one region, so you see the repositories of the project and region you are in. See Create a repository.
Can another project use my repository?
Section titled “Can another project use my repository?”No. Each project creates its own repository. See Create a repository.
Access and keys
Section titled “Access and keys”Who in my project can use a repository?
Section titled “Who in my project can use a repository?”The Owner, an Admin, a Project Manager, and a Project Lead can manage repositories. A Member needs a policy that allows the Inference service. Repository access comes with those Inference service permissions. For a Member, each permission opens up one kind of action:
- Reading opens the list, the tabs, and the files.
- Creating makes repositories and shows Deploy Model.
- Updating shows Upload.
- Deleting removes repositories and files.
The portal shows only the actions your role allows. Choosing Existing EOS Bucket also needs the permission to read datasets. See Create a repository.
Can I restrict a repository to some people in the project?
Section titled “Can I restrict a repository to some people in the project?”No. Anyone who can open a repository can use its files and keys. Permissions cover every repository in the project at once. See Find a repository’s access keys and endpoint.
Who can see the repository’s keys?
Section titled “Who can see the repository’s keys?”Anyone who can open the repository. Save them somewhere safe and keep them private. You can show the secret key again whenever you need it. See Find a repository’s access keys and endpoint.
Where do I find the access keys and endpoint?
Section titled “Where do I find the access keys and endpoint?”On the repository’s Overview tab, under Bucket Details, and inside the commands on its Using CLI tab. The EOS Endpoint is different in Delhi and Chennai. See Find a repository’s access keys and endpoint.
Files and versions
Section titled “Files and versions”How do I get files in and out of a repository?
Section titled “How do I get files in and out of a repository?”Use the portal’s Model Files tab, the MinIO CLI (mc), s3cmd, or the Python SDK. The files are also reachable over an S3-compatible API. See Upload model files and Browse, download, and delete files.
How do I upload a large model?
Section titled “How do I upload a large model?”Use the MinIO CLI, s3cmd, or the Python SDK. The portal uploads up to 10 files at a time and recommends the MinIO CLI or s3cmd for files over 5 GB. Compressing files before you upload also helps. See Connect to a repository with the MinIO CLI or s3cmd.
Does mc cp have a –parallel option?
Section titled “Does mc cp have a –parallel option?”No. To copy several files at once, use mc cp with --max-workers. See Upload model files.
How do I version my models?
Section titled “How do I version my models?”Use folders in the bucket: /v1 and /v2 for numbered versions, /production and /staging for environments, and /experiments/exp-001 for experiments. The portal has no version screen. See Upload model files.
How many versions should I keep?
Section titled “How many versions should I keep?”Keep the two or three previous versions for rollback, and archive or delete older ones to save storage. See Upload model files.
Can I delete a version?
Section titled “Can I delete a version?”Yes. Delete its folder or its files with the portal, the MinIO CLI, or the SDK. Deleting is permanent. See Browse, download, and delete files.
Endpoints and training
Section titled “Endpoints and training”How do I link a repository to a Model Endpoint?
Section titled “How do I link a repository to a Model Endpoint?”You have two ways. Select Deploy Model on the repository’s row and pick a framework. Or select Link with Model Repository while you create an endpoint, choose the repository, and give the Model Path if the files are in a folder, such as /v1. The endpoint downloads the model from the repository when its container starts or restarts. See Use a repository from a Model Endpoint.
Why can my endpoint not find the model?
Section titled “Why can my endpoint not find the model?”Check three things. The Model Path must match the folder layout, the files must be in the repository, and the bucket name must be right. Then read the endpoint’s logs. See Use a repository from a Model Endpoint.
Can I serve a LoRA adapter from a repository?
Section titled “Can I serve a LoRA adapter from a repository?”Yes, on vLLM. Link the repository, then choose PEFT Model (for LoRA-based inference) under Model Type and give the Base Model Name. The repository must hold the adapter’s configuration file, named adapter_config.json, and a file ending in .safetensors. See Use a repository from a Model Endpoint.
Can I mount a repository on an instance?
Section titled “Can I mount a repository on an instance?”No. Instead, use the Python SDK or the MinIO CLI from the instance to move files in and out. See Use a repository from an instance or notebook.
Can I keep a fine-tuned model in a repository?
Section titled “Can I keep a fine-tuned model in a repository?”Yes. A repository can store base models for fine-tuning and fine-tuned checkpoints. You can also link it to a Model Endpoint to serve it, as Can I deploy my own or a fine-tuned model? explains. See Store fine-tuned model output.
Deleting and billing
Section titled “Deleting and billing”What happens to my files when I delete a repository?
Section titled “What happens to my files when I delete a repository?”For a New EOS Bucket, the repository is deleted, together with the bucket and every file in it. For an Existing EOS Bucket or an External EOS Bucket, copy any files you need before you delete the repository. Deleting is permanent. See Delete a repository.
Can I delete a repository that has an endpoint?
Section titled “Can I delete a repository that has an endpoint?”No. Delete the endpoint first. See Delete a repository.
How is a repository billed?
Section titled “How is a repository billed?”The Model Repository screens show no price. A repository itself is free: you pay for the storage of its bucket at the Object Storage rate, by the month’s peak. The peak is the most storage it held at any point in the month, even for a short time. Data transfer is charged as Object Storage charges it when you upload or download model files. An existing bucket is charged as before, with the repository’s files counting toward its peak. An endpoint that uses a repository is billed on its own, for its workers; see How is an endpoint billed?. For rates, see E2E’s pricing page. See Create a repository.
Do I set a size or pay in advance?
Section titled “Do I set a size or pay in advance?”No. A repository grows with its files, and its storage is billed by the month’s peak. See Create a repository.