fix: add missing av dependency in contanier for multimodal examples (#1881) #1882

biswapanda · 2025-07-11T00:20:14Z

Overview:

nvbug: 5383740

Adds missing dep av for video model.

PR from #1881

Where should the reviewer start?

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

closes GitHub issue: #xxx

TODO:

separate Dockerfile: DYN-700

Summary by CodeRabbit

New Features
- Added multiple Kubernetes deployment manifests for multimodal applications, enabling easy deployment of complex multimodal graphs with distinct service roles and resource configurations.
Bug Fixes
- Updated GPU resource specification in deployment files to use a generic gpu key instead of vendor-specific keys, improving compatibility.
Documentation
- Enhanced and clarified Kubernetes deployment instructions for multimodal examples.
- Updated vLLM installation and multi-node deployment instructions for improved clarity and accuracy.
Chores
- Added a new dependency (av==15.0.0) for video processing.
- Updated default service port for Dynamo from 3000 to 8000.
- Removed in-cluster MinIO Helm chart and related configuration.
Refactor
- Consolidated Python package installation steps and improved comment consistency.
- Replaced hardcoded port numbers in tests with a shared constant.
- Deferred certain imports to avoid runtime errors in non-GPU environments.
Style
- Fixed minor typos and standardized comment formatting.

…1881)

copy-pr-bot · 2025-07-11T00:20:18Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

coderabbitai · 2025-07-11T00:30:13Z

Caution

Review failed

Failed to post review comments.

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 26d7a61 and 78d7f1a.

📒 Files selected for processing (28)

container/Dockerfile.tensorrt_llm (3 hunks)
container/deps/requirements.txt (1 hunks)
deploy/cloud/helm/platform/Chart.yaml (0 hunks)
deploy/cloud/helm/platform/values.yaml (0 hunks)
deploy/cloud/operator/internal/consts/consts.go (1 hunks)
deploy/cloud/operator/internal/controller/dynamocomponentdeployment_controller_test.go (4 hunks)
deploy/cloud/operator/internal/dynamo/graph_test.go (4 hunks)
examples/llm/deploy/agg.yaml (1 hunks)
examples/llm/deploy/agg_router.yaml (1 hunks)
examples/llm/deploy/disagg.yaml (2 hunks)
examples/llm/deploy/disagg_router.yaml (2 hunks)
examples/multimodal/README.md (1 hunks)
examples/multimodal/deploy/k8s/agg-llava.yaml (1 hunks)
examples/multimodal/deploy/k8s/agg-phi3v.yaml (1 hunks)
examples/multimodal/deploy/k8s/agg-qwen.yaml (1 hunks)
examples/multimodal/deploy/k8s/agg-video.yaml (1 hunks)
examples/multimodal/deploy/k8s/disagg-video.yaml (1 hunks)
examples/multimodal/deploy/k8s/disagg.yaml (1 hunks)
examples/multimodal/utils/model.py (1 hunks)
examples/sglang/multinode-examples.md (1 hunks)
examples/sglang/slurm_jobs/job_script_template.j2 (1 hunks)
examples/vllm_v0/deploy/agg.yaml (3 hunks)
examples/vllm_v0/deploy/disagg.yaml (5 hunks)
examples/vllm_v0/deploy/disagg_planner.yaml (6 hunks)
examples/vllm_v1/README.md (1 hunks)
examples/vllm_v1/deploy/agg.yaml (4 hunks)
examples/vllm_v1/deploy/disagg.yaml (6 hunks)
examples/vllm_v1/deploy/disagg_planner.yaml (8 hunks)

💤 Files with no reviewable changes (2)

deploy/cloud/helm/platform/Chart.yaml
deploy/cloud/helm/platform/values.yaml

🧰 Additional context used

🧠 Learnings (17)

deploy/cloud/operator/internal/consts/consts.go (2)

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1365
File: deploy/cloud/operator/api/v1alpha1/dynamocomponentdeployment_types.go:171-178
Timestamp: 2025-06-04T13:09:53.416Z
Learning: The `DYN_DEPLOYMENT_CONFIG` environment variable (commonconsts.DynamoDeploymentConfigEnvVar) in the Dynamo operator will never be set via ValueFrom (secrets/config maps), only via direct Value assignment. The GetDynamoDeploymentConfig method correctly only checks env.Value for this specific environment variable.

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1474
File: deploy/cloud/operator/internal/controller/dynamocomponent_controller.go:1302-1306
Timestamp: 2025-06-11T21:18:00.425Z
Learning: In the Dynamo operator, the project’s preferred security posture is to set a Pod-level `PodSecurityContext` with `runAsUser`, `runAsGroup`, and `fsGroup` all set to `1000`, and then selectively override the user at the individual container level (e.g., `RunAsUser: 0` for Kaniko) when root is required.

examples/llm/deploy/agg_router.yaml (1)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

examples/llm/deploy/agg.yaml (1)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

examples/multimodal/utils/model.py (1)

Learnt from: krishung5
PR: ai-dynamo/dynamo#1388
File: examples/multimodal/utils/model.py:47-53
Timestamp: 2025-06-09T17:52:06.761Z
Learning: The current get_vision_embeddings_size() function in examples/multimodal/utils/model.py uses a hardcoded fallback of 4096 for hidden_size, which assumes all VLMs follow the LLaVA architecture pattern. This is problematic because different VLMs like Qwen2-VL, MiniCPM-V, and others have different hidden dimensions, making the fallback unreliable and potentially causing tensor shape mismatches.

examples/llm/deploy/disagg.yaml (1)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

examples/sglang/slurm_jobs/job_script_template.j2 (1)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/scenarios.py:57-57
Timestamp: 2025-07-01T15:39:56.789Z
Learning: The fault tolerance tests in tests/fault_tolerance/ are designed to run only in the mounted container environment, so hardcoded absolute paths with `/workspace/` prefix are intentional and should not be changed to relative paths.

examples/llm/deploy/disagg_router.yaml (1)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

deploy/cloud/operator/internal/dynamo/graph_test.go (2)

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1365
File: deploy/cloud/operator/api/v1alpha1/dynamocomponentdeployment_types.go:171-178
Timestamp: 2025-06-04T13:09:53.416Z
Learning: The `DYN_DEPLOYMENT_CONFIG` environment variable (commonconsts.DynamoDeploymentConfigEnvVar) in the Dynamo operator will never be set via ValueFrom (secrets/config maps), only via direct Value assignment. The GetDynamoDeploymentConfig method correctly only checks env.Value for this specific environment variable.

Learnt from: GuanLuo
PR: ai-dynamo/dynamo#1371
File: examples/llm/benchmarks/vllm_multinode_setup.sh:18-25
Timestamp: 2025-06-05T01:46:15.509Z
Learning: In multi-node setups with head/worker architecture, the head node typically doesn't need environment variables pointing to its own services (like NATS_SERVER, ETCD_ENDPOINTS) because local processes can access them via localhost. Only worker nodes need these environment variables to connect to the head node's external IP address.

container/Dockerfile.tensorrt_llm (1)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/utils/metrics.py:30-32
Timestamp: 2025-07-01T13:55:03.940Z
Learning: The `@dynamo_worker()` decorator in the dynamo codebase returns a wrapper that automatically injects the `runtime` parameter before calling the wrapped function. This means callers only need to provide the non-runtime parameters, while the decorator handles injecting the runtime argument automatically. For example, a function with signature `async def get_metrics(runtime, log_dir)` decorated with `@dynamo_worker()` can be called as `get_metrics(log_dir)` because the decorator wrapper injects the runtime parameter.

examples/vllm_v0/deploy/disagg_planner.yaml (2)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1337
File: deploy/cloud/helm/platform/components/operator/templates/image-builer-serviceaccount.yaml:0-0
Timestamp: 2025-06-03T15:26:55.732Z
Learning: The image-builder ServiceAccount in deploy/cloud/helm/platform/components/operator/templates/image-builer-serviceaccount.yaml does not need imagePullSecrets, unlike the component ServiceAccount.

examples/vllm_v1/deploy/agg.yaml (1)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

deploy/cloud/operator/internal/controller/dynamocomponentdeployment_controller_test.go (1)

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1365
File: deploy/cloud/operator/api/v1alpha1/dynamocomponentdeployment_types.go:171-178
Timestamp: 2025-06-04T13:09:53.416Z
Learning: The `DYN_DEPLOYMENT_CONFIG` environment variable (commonconsts.DynamoDeploymentConfigEnvVar) in the Dynamo operator will never be set via ValueFrom (secrets/config maps), only via direct Value assignment. The GetDynamoDeploymentConfig method correctly only checks env.Value for this specific environment variable.

examples/vllm_v0/deploy/disagg.yaml (2)

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1365
File: deploy/cloud/operator/api/v1alpha1/dynamocomponentdeployment_types.go:171-178
Timestamp: 2025-06-04T13:09:53.416Z
Learning: The `DYN_DEPLOYMENT_CONFIG` environment variable (commonconsts.DynamoDeploymentConfigEnvVar) in the Dynamo operator will never be set via ValueFrom (secrets/config maps), only via direct Value assignment. The GetDynamoDeploymentConfig method correctly only checks env.Value for this specific environment variable.

examples/vllm_v1/deploy/disagg_planner.yaml (3)

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1365
File: deploy/cloud/operator/api/v1alpha1/dynamocomponentdeployment_types.go:171-178
Timestamp: 2025-06-04T13:09:53.416Z
Learning: The `DYN_DEPLOYMENT_CONFIG` environment variable (commonconsts.DynamoDeploymentConfigEnvVar) in the Dynamo operator will never be set via ValueFrom (secrets/config maps), only via direct Value assignment. The GetDynamoDeploymentConfig method correctly only checks env.Value for this specific environment variable.

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1337
File: deploy/cloud/helm/platform/components/operator/templates/image-builer-serviceaccount.yaml:0-0
Timestamp: 2025-06-03T15:26:55.732Z
Learning: The image-builder ServiceAccount in deploy/cloud/helm/platform/components/operator/templates/image-builer-serviceaccount.yaml does not need imagePullSecrets, unlike the component ServiceAccount.

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

examples/vllm_v1/deploy/disagg.yaml (2)

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1365
File: deploy/cloud/operator/api/v1alpha1/dynamocomponentdeployment_types.go:171-178
Timestamp: 2025-06-04T13:09:53.416Z
Learning: The `DYN_DEPLOYMENT_CONFIG` environment variable (commonconsts.DynamoDeploymentConfigEnvVar) in the Dynamo operator will never be set via ValueFrom (secrets/config maps), only via direct Value assignment. The GetDynamoDeploymentConfig method correctly only checks env.Value for this specific environment variable.

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

examples/vllm_v0/deploy/agg.yaml (2)

Learnt from: julienmancuso
PR: ai-dynamo/dynamo#1365
File: deploy/cloud/operator/api/v1alpha1/dynamocomponentdeployment_types.go:171-178
Timestamp: 2025-06-04T13:09:53.416Z
Learning: The `DYN_DEPLOYMENT_CONFIG` environment variable (commonconsts.DynamoDeploymentConfigEnvVar) in the Dynamo operator will never be set via ValueFrom (secrets/config maps), only via direct Value assignment. The GetDynamoDeploymentConfig method correctly only checks env.Value for this specific environment variable.

Learnt from: nnshah1
PR: ai-dynamo/dynamo#1444
File: tests/fault_tolerance/configs/agg_tp_1_dp_8.yaml:31-38
Timestamp: 2025-07-01T15:33:53.262Z
Learning: In fault tolerance test configurations, the `resources` section under `ServiceArgs` specifies resources per individual worker, not total resources for all workers. So `workers: 8` with `gpu: '1'` means 8 workers × 1 GPU each = 8 GPUs total.

examples/sglang/multinode-examples.md (3)

Learnt from: fsaady
PR: ai-dynamo/dynamo#1730
File: examples/sglang/slurm_jobs/job_script_template.j2:59-59
Timestamp: 2025-07-02T13:20:28.800Z
Learning: In the SLURM job script template at examples/sglang/slurm_jobs/job_script_template.j2, the `--total_nodes` parameter represents the total nodes per worker type (prefill or decode), not the total nodes in the entire cluster. Each worker type needs to know its own group size for distributed coordination.

Learnt from: GuanLuo
PR: ai-dynamo/dynamo#1371
File: examples/llm/benchmarks/vllm_multinode_setup.sh:18-25
Timestamp: 2025-06-05T01:46:15.509Z
Learning: In multi-node setups with head/worker architecture, the head node typically doesn't need environment variables pointing to its own services (like NATS_SERVER, ETCD_ENDPOINTS) because local processes can access them via localhost. Only worker nodes need these environment variables to connect to the head node's external IP address.

Learnt from: fsaady
PR: ai-dynamo/dynamo#1730
File: examples/sglang/slurm_jobs/scripts/worker_setup.py:230-244
Timestamp: 2025-07-03T10:14:30.570Z
Learning: In examples/sglang/slurm_jobs/scripts/worker_setup.py, background processes (like nats-server, etcd) are intentionally left running even if later processes fail. This design choice allows users to manually connect to nodes and debug issues without having to restart the entire SLURM job from scratch, providing operational flexibility for troubleshooting in cluster environments.

🧬 Code Graph Analysis (2)

deploy/cloud/operator/internal/dynamo/graph_test.go (1)

deploy/cloud/operator/internal/consts/consts.go (1)

DynamoServicePort (9-9)

deploy/cloud/operator/internal/controller/dynamocomponentdeployment_controller_test.go (1)

deploy/cloud/operator/internal/consts/consts.go (1)

DynamoServicePort (9-9)

🪛 YAMLlint (1.37.1)

examples/multimodal/deploy/k8s/disagg-video.yaml

[error] 162-162: no new line character at the end of file

(new-line-at-end-of-file)

examples/multimodal/deploy/k8s/agg-llava.yaml

[error] 133-133: no new line character at the end of file

(new-line-at-end-of-file)

examples/multimodal/deploy/k8s/agg-phi3v.yaml

[error] 133-133: no new line character at the end of file

(new-line-at-end-of-file)

examples/multimodal/deploy/k8s/disagg.yaml

[error] 162-162: no new line character at the end of file

(new-line-at-end-of-file)

examples/multimodal/deploy/k8s/agg-video.yaml

[error] 133-133: no new line character at the end of file

(new-line-at-end-of-file)

examples/multimodal/deploy/k8s/agg-qwen.yaml

[error] 133-133: no new line character at the end of file

(new-line-at-end-of-file)

🔇 Additional comments (13)

examples/sglang/multinode-examples.md (2)

17-17: dynamo run invocation lacks configuration context

The bare dynamo run in=http out=dyn --http-port=8000 & presupposes a default config. In most recent examples the command is dynamo run -f ./configs/ingress.yaml …. If a config file is required, include it; otherwise call out that default values are sufficient.

24-26: Ensure variable names from Step 1 match later usage

The worker commands rely on ${HEAD_PREFILL_NODE_IP} and ${HEAD_DECODE_NODE_IP}. Please double-check the helper script actually emits variables with exactly these names (including the _IP suffix and the head / decode distinction) so new users don’t hit a silent env-var expansion failure.

Also applies to: 43-45, 62-64, 81-83

examples/multimodal/utils/model.py (1)

32-34: LGTM - Good defensive programming practice.

Moving the Worker import to a lazy import pattern is a smart optimization that prevents CUDA initialization errors when the code is imported but not executed on GPU hardware. This change maintains the same functionality while improving compatibility across different environments.

examples/llm/deploy/disagg.yaml (1)

82-82: LGTM - Standardizes GPU resource keys.

The change from vendor-specific nvidia.com/gpu to generic gpu improves portability and aligns with the broader effort to standardize GPU resource specifications across deployment manifests. The updates are applied consistently to both worker types and resource constraints.

Also applies to: 86-86, 109-109, 113-113

examples/sglang/slurm_jobs/job_script_template.j2 (1)

57-57: LGTM - Improves container isolation.

Changing from --container-mount-home to --no-container-mount-home enhances container security by preventing access to the host's home directory. This follows container best practices for isolation and prevents potential conflicts between host and container environments.

examples/vllm_v1/README.md (1)

44-53: LGTM - Improved installation process with better reproducibility.

The updated installation instructions provide better control and reproducibility by:

Using a specific git commit reference for consistency

Leveraging precompiled builds for better performance

Including cleanup instructions for existing installations

The multi-step process is more complex but ensures a more reliable and predictable installation.

deploy/cloud/operator/internal/dynamo/graph_test.go (2)

22-22: LGTM - Necessary import for string formatting.

The fmt package import is required to support the fmt.Sprintf calls used for parameterizing port values in the test data.

517-517: LGTM - Excellent refactoring to eliminate hardcoded port values.

Replacing hardcoded port numbers with commonconsts.DynamoServicePort improves test maintainability and ensures consistency with the updated port configuration (8000). The use of fmt.Sprintf for JSON string formatting is appropriate and makes the tests more robust to future port changes.

Also applies to: 774-776, 796-798

examples/vllm_v1/deploy/agg.yaml (1)

46-48: Verify that agg.yaml is packaged in the image or mounted at runtime

The extra -f ./configs/agg.yaml arg is great for explicit config, but make sure the file is actually present under /workspace/examples/vllm_v1/configs/ inside the runtime image (or mounted via a ConfigMap). Otherwise the entry-point will fail on a “file not found”.

examples/vllm_v0/deploy/disagg.yaml (1)

46-48: Explicit config flag looks good
The service now self-declares its config file – nice improvement.

container/Dockerfile.tensorrt_llm (1)

374-375: ARG re-declaration is necessary – good catch
Redeclaring ARG ARCH_ALT in the runtime stage prevents ${ARCH_ALT} from expanding to an empty string during COPY/ENV lines that follow.

examples/multimodal/README.md (1)

338-344: Documentation now contradicts container image

The PR title states the av dependency is baked into the container, yet the README still instructs users to pip install av manually.
Please remove or update these lines so the docs reflect the new reality.

deploy/cloud/operator/internal/controller/dynamocomponentdeployment_controller_test.go (1)

311-312: Port literal removed – good use of shared constant

Replacing magic number 3000 with commonconsts.DynamoServicePort improves maintainability and keeps tests in sync with production defaults.

Also applies to: 468-469, 901-910, 952-957

Walkthrough

This update introduces new Kubernetes deployment manifests for multimodal and vLLM examples, removes MinIO as a Helm chart dependency, and standardizes GPU resource keys in deployment YAMLs. It also updates port constants in Go code, consolidates Dockerfile commands, adds a Python dependency, and revises documentation for installation and deployment workflows.

Changes

File(s)	Change Summary
deploy/cloud/helm/platform/Chart.yaml, deploy/cloud/helm/platform/values.yaml	Removed MinIO Helm chart dependency and related configuration block.
container/Dockerfile.tensorrt_llm, container/deps/requirements.txt	Added `ARG ARCH_ALT`, consolidated pip installs, fixed comments, and added `av==15.0.0` dependency.
deploy/cloud/operator/internal/consts/consts.go	Changed `DynamoServicePort` constant from 3000 to 8000.
deploy/cloud/operator/internal/controller/dynamocomponentdeployment_controller_test.go, deploy/cloud/operator/internal/dynamo/graph_test.go	Replaced hardcoded port 3000 with `commonconsts.DynamoServicePort` in tests.
examples/llm/deploy/agg.yaml, examples/llm/deploy/agg_router.yaml, examples/llm/deploy/disagg.yaml, examples/llm/deploy/disagg_router.yaml	Changed GPU resource key from `nvidia.com/gpu` to `gpu`.
examples/vllm_v0/deploy/agg.yaml, examples/vllm_v0/deploy/disagg.yaml, examples/vllm_v0/deploy/disagg_planner.yaml, examples/vllm_v1/deploy/agg.yaml, examples/vllm_v1/deploy/disagg.yaml, examples/vllm_v1/deploy/disagg_planner.yaml	Removed `DYN_DEPLOYMENT_CONFIG` env, added `-f <config>` args, changed GPU key to `gpu`.
examples/multimodal/README.md	Replaced dependency note with detailed Kubernetes deployment instructions.
examples/multimodal/utils/model.py	Made import of `Worker` from `vllm.worker.worker` lazy inside a function.
examples/multimodal/deploy/k8s/agg-llava.yaml, examples/multimodal/deploy/k8s/agg-phi3v.yaml, examples/multimodal/deploy/k8s/agg-qwen.yaml, examples/multimodal/deploy/k8s/agg-video.yaml, examples/multimodal/deploy/k8s/disagg.yaml, examples/multimodal/deploy/k8s/disagg-video.yaml	Added new multimodal Kubernetes deployment manifests.
examples/sglang/multinode-examples.md	Improved SGLang multinode deployment documentation and example commands.
examples/sglang/slurm_jobs/job_script_template.j2	Changed enroot argument to `--no-container-mount-home`.
examples/vllm_v1/README.md	Updated vLLM installation instructions with manual steps and uninstall note.

Sequence Diagram(s)

sequenceDiagram
    participant User
    participant Kubernetes
    participant Container
    participant Secret
    participant ConfigFile

    User->>Kubernetes: Apply new multimodal deployment manifest
    Kubernetes->>Container: Launches service pod(s)
    Container->>Secret: Loads HuggingFace token (if needed)
    Container->>ConfigFile: Reads config via -f <config>.yaml
    Container->>Container: Runs dynamo serve with specified graph component

Possibly related PRs

ai-dynamo/dynamo#1833: Removes the MinIO dependency and configuration from Helm files, directly matching the MinIO removal in this PR.
ai-dynamo/dynamo#1820: Removes DYN_DEPLOYMENT_CONFIG from example YAMLs and replaces it with explicit config file arguments, similar to changes in this PR.
ai-dynamo/dynamo#1881: Adds the av==15.0.0 dependency to container/deps/requirements.txt, matching the addition in this PR.

Poem

In the warren, YAMLs grow and change,
GPU keys now less strange.
MinIO hops away,
Multimodal bunnies here to stay.
Ports align and docs are neat,
With every merge, our carrots taste sweet!
🥕🐇

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

Review comments: Directly reply to a review comment made by CodeRabbit. Example:
- I pushed a fix in commit <commit_id>, please review it.
- Explain this complex logic.
- Open a follow-up GitHub issue for this discussion.
Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query. Examples:
- @coderabbitai explain this code block.
- @coderabbitai modularize this function.
PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
- @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
- @coderabbitai read src/utils.ts and explain its main purpose.
- @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.
- @coderabbitai help me debug CodeRabbit configuration file.

Support

Need help? Create a ticket on our support page for assistance with any issues or questions.

Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments.

CodeRabbit Commands (Invoked using PR comments)

@coderabbitai pause to pause the reviews on a PR.
@coderabbitai resume to resume the paused reviews.
@coderabbitai review to trigger an incremental review. This is useful when automatic reviews are disabled for the repository.
@coderabbitai full review to do a full review from scratch and review all the files again.
@coderabbitai summary to regenerate the summary of the PR.
@coderabbitai generate sequence diagram to generate a sequence diagram of the changes in this PR.
@coderabbitai resolve resolve all the CodeRabbit review comments.
@coderabbitai configuration to show the current CodeRabbit configuration for the repository.
@coderabbitai help to get help.

Other keywords and placeholders

Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
Add @coderabbitai anywhere in the PR title to generate the title automatically.

Documentation and Community

Visit our Documentation for detailed information on how to use CodeRabbit.
Join our Discord Community to get help, request features, and share feedback.
Follow us on X/Twitter for updates and announcements.

whoisj · 2025-07-11T14:20:08Z

container/deps/requirements.txt

@@ -14,6 +14,7 @@
 # limitations under the License.

 accelerate==1.6.0
+av==15.0.0


so we've have an av dependency for every version of use of Dynamo?

starting to seems like a "and the kitchen sink" approach to engineering.

Agreed, another alternative is we dont alter the base vllm image and create multimodal specific dockerfile and publish the container.

@whoisj, @nv-anants: do you have any other suggestions?

I think that makes more sense. Might be worth keeping the containerfile in the example folder?

@biswapanda let's get this merged for now and please create a work item for separating the dockerfiles. Thanks.

whoisj

Please create a work item to separate the container files.

biswapanda · 2025-07-14T23:20:52Z

Please create a work item to separate the container files.

sounds good, I've created a ticket DYN-700 to address the comment.

fix: add missing av dependency in contanier for multimodal examples (#…

78d7f1a

…1881)

pull-request-size bot added the size/XXL label Jul 11, 2025

github-actions bot added the fix label Jul 11, 2025

biswapanda changed the base branch from main to release/0.3.2 July 11, 2025 00:20

pull-request-size bot added size/XS and removed size/XXL labels Jul 11, 2025

biswapanda enabled auto-merge (squash) July 11, 2025 00:21

biswapanda self-assigned this Jul 11, 2025

krishung5 approved these changes Jul 11, 2025

View reviewed changes

whoisj reviewed Jul 11, 2025

View reviewed changes

whoisj approved these changes Jul 14, 2025

View reviewed changes

nv-anants disabled auto-merge July 15, 2025 00:15

nv-anants merged commit 4c3751f into release/0.3.2 Jul 15, 2025
8 checks passed

nv-anants deleted the bis/cp-missing-dep branch July 15, 2025 00:15

coderabbitai bot mentioned this pull request Jul 17, 2025

feat: add vLLM v1 multi-modal example. Add llama4 Maverick example #1990

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix: add missing av dependency in contanier for multimodal examples (#1881) #1882

fix: add missing av dependency in contanier for multimodal examples (#1881) #1882

Uh oh!

biswapanda commented Jul 11, 2025 •

edited

Loading

Uh oh!

copy-pr-bot bot commented Jul 11, 2025

Uh oh!

coderabbitai bot commented Jul 11, 2025 •

edited

Loading

Review failed

Chat

Support

CodeRabbit Commands (Invoked using PR comments)

Other keywords and placeholders

Documentation and Community

Uh oh!

whoisj Jul 11, 2025

Uh oh!

biswapanda Jul 11, 2025

Uh oh!

whoisj Jul 11, 2025

Uh oh!

whoisj Jul 14, 2025

Uh oh!

whoisj left a comment

Uh oh!

biswapanda commented Jul 14, 2025

Uh oh!

Uh oh!

Uh oh!

fix: add missing av dependency in contanier for multimodal examples (#1881) #1882

fix: add missing av dependency in contanier for multimodal examples (#1881) #1882

Uh oh!

Conversation

biswapanda commented Jul 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Overview:

Where should the reviewer start?

Related Issues: (use one of the action keywords Closes / Fixes / Resolves / Relates to)

Summary by CodeRabbit

Uh oh!

copy-pr-bot bot commented Jul 11, 2025

Uh oh!

coderabbitai bot commented Jul 11, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review failed

Walkthrough

Changes

Sequence Diagram(s)

Possibly related PRs

Poem

Chat

Support

CodeRabbit Commands (Invoked using PR comments)

Other keywords and placeholders

Documentation and Community

Uh oh!

whoisj Jul 11, 2025

Choose a reason for hiding this comment

Uh oh!

biswapanda Jul 11, 2025

Choose a reason for hiding this comment

Uh oh!

whoisj Jul 11, 2025

Choose a reason for hiding this comment

Uh oh!

whoisj Jul 14, 2025

Choose a reason for hiding this comment

Uh oh!

whoisj left a comment

Choose a reason for hiding this comment

Uh oh!

biswapanda commented Jul 14, 2025

Uh oh!

Uh oh!

Uh oh!

biswapanda commented Jul 11, 2025 •

edited

Loading

coderabbitai bot commented Jul 11, 2025 •

edited

Loading