> ## Documentation Index > Fetch the complete documentation index at: https://docs.devin.ai/llms.txt > Use this file to discover all available pages before exploring further. # Post-Train Custom Models With Devin and Modal > Post-train a custom model with Devin on a Modal GPU outpost: fine-tune a FLUX LoRA, deploy it as a Modal endpoint, and ship it in a PR Devin post-training a FLUX LoRA on a Modal GPU, deploying the best checkpoint, and adding it to a chat app Devin runs wherever the job needs to run. With [Devin Outposts](/cloud/outposts/overview), a session can run on your own infrastructure or a partner platform; here it runs in a [Modal](https://modal.com/docs/devin) GPU sandbox, so the same Devin that writes your code can train, evaluate, and deploy a model end to end. In this tutorial Devin fine-tunes a FLUX LoRA on 18 otter images so that "Devin the software engineer" renders as an otter, deploys it as a Modal endpoint, and adds it to a chat app's model picker. ## Before you start * **Create a Modal outpost**: run `uvx modal-devin init` and pick a GPU for the generated Modal app. See [Run Devin on Modal](https://modal.com/docs/devin) and the [Outposts quickstart](/cloud/outposts/quickstart.md). * **Connect your repository**: the app you want the model wired into. The video uses a private demo chat app with an image model picker, checked out onto a persistent Modal Volume. Don't have an app? Ask Devin to build one first (for example, `build a simple chat app with a model picker that calls an image generation endpoint`), or skip the app and have Devin only train and deploy the model. * **Prepare your training data**: add captioned images to the repository or a Modal Volume the outpost mounts. The example uses 18 cartoon otter images with captions like "Devin the software engineer in a data center". * **Add secrets**: store your Weights & Biases key under [Settings > Secrets](https://app.devin.ai/settings/secrets). The Modal sandbox already carries Modal credentials, so Devin can deploy endpoints without extra setup. * **Set rules for long runs**: add [organization rules](/product-guides/plugins.md#rules) (or put them in the prompt) such as: post a progress update every 15 minutes, render validation images at every checkpoint, and include the preview link in every PR. * **Connect Slack** (optional): training runs take hours, so [sync the session to a Slack thread](/integrations/slack.md#sync-sessions-with-slack-threads) to follow along. ## Step 1: Start a session on the outpost Open [Devin](https://app.devin.ai/), choose your Modal outpost from the **Virtual environment** menu below the prompt box, and turn on the Slack toggle in the composer. Check the hardware: ```text theme={null} tell me more about your environment ``` Devin reports an NVIDIA B200 with PyTorch and diffusers installed, and the project on a persistent Modal Volume. ## Step 2: Check the training data ```text theme={null} hey devin show me the images im using for supervised learning ``` Devin finds the dataset in the project and posts the 18 images, with their captions, to the session and the Slack thread. ## Step 3: Kick off the training run ```text theme={null} post train flux to create the otterimage 1.0 model. once you are done, add otterimage to the model picker in otterchat ``` `otterchat` is the demo app in the video; use your own repository's name. Devin runs the job on its own, following the rules above: 1. **Trains the LoRA**: about 1,000 steps, logged to Weights & Biases, with validation renders every 100 steps. 2. **Reports to Slack**: progress every 15 minutes with the latest renders. Early renders are humans; the otter appears as training continues. 3. **Picks the best checkpoint**: step 700, since steps 900 and 1,000 showed signs of overfitting. 4. **Deploys and ships**: serves step 700 as a serverless Modal endpoint, adds `otterimage 1.0` to the chat app's model picker, [records](/work-with-devin/testing-and-recordings.md) an end-to-end test, and opens a PR with a preview link. ![Grid of validation images at each checkpoint](https://mintcdn.com/cognitionai/4bg4JipkvpdMh6Vh/images/tutorial-library/post-training-validation.png?fit=max\&auto=format\&n=4bg4JipkvpdMh6Vh\&q=85\&s=7da45a563e1c6d538643483e75601a74) Validation renders from step 500 to step 1,000 ## Step 4: Try the model Open the preview link from the PR and ask for an image: ```text theme={null} generate an image of devin the software engineer eating spaghetti ``` ![The chat app returning a cartoon otter eating spaghetti](https://mintcdn.com/cognitionai/4bg4JipkvpdMh6Vh/images/tutorial-library/post-training-otterchat.png?fit=max\&auto=format\&n=4bg4JipkvpdMh6Vh\&q=85\&s=e7a6d33991a5cd2089bebc64857f98ec) The same prompt on a general-purpose image model returns a human. ## Step 5: Iterate with child sessions Post-training is rarely one run. Hand Devin a paper and let it try the alternatives in parallel: ```text theme={null} https://arxiv.org/pdf/2309.14859v2 spin up subsession that train otterimage on the other techniques listed in this paper other than lora ``` Devin adds the new methods to the training script in a PR and starts three [child sessions](/work-with-devin/advanced-capabilities.md#managed-devins) on the outpost, one each for LoHa, LoKr, and native fine-tuning. ![Devin's session showing three child sessions for LoHa, LoKr, and native fine-tuning](https://mintcdn.com/cognitionai/4bg4JipkvpdMh6Vh/images/tutorial-library/post-training-child-sessions.png?fit=max\&auto=format\&n=4bg4JipkvpdMh6Vh\&q=85\&s=8892cdce447e15e95fc4a01187f6af09) ## Next steps > **[Run Devin on Modal](https://modal.com/docs/devin)** > > Create a Modal-backed outpost with GPUs, Volumes, and Secrets > **[Devin Outposts](/cloud/outposts/overview)** > > Run sessions on your own infrastructure or a partner platform > **[Slack integration](/integrations/slack.md)** > > Follow long-running sessions from a Slack thread > **[Managed Devins](/work-with-devin/advanced-capabilities.md#managed-devins)** > > Spin up and monitor child sessions for parallel experiments This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.