Fix default image on runpod and vllm execution as well#1438
Conversation
deep1401
commented
Mar 3, 2026
deep1401
commented
- This fixes a bug with vllm where versions < 0.13 dont work
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
Paragon SummaryThis pull request review identified 1 issue across 1 category in 2 files. The review analyzed code changes, potential bugs, security vulnerabilities, performance issues, and code quality concerns using automated analysis tools. This PR fixes a compatibility bug where vLLM versions earlier than 0.13 failed to work properly on RunPod, correcting the default image configuration for vLLM execution. Key changes:
Confidence score: 3/5
2 files reviewed, 1 comment Severity breakdown: High: 1 Tip: |
![paragon-review[bot]](https://avatars.githubusercontent.com/in/1544258?s=60&v=4){kind=small}
|
|
||
| # Use GPU-enabled image | ||
| default_image = "runpod/pytorch:2.1.0-py3.10-cuda11.8.0-devel-ubuntu22.04" | ||
| default_image = "runpod/pytorch:1.0.3-cu1281-torch290-ubuntu2204" |
There was a problem hiding this comment.
Bug: Docker image tag does not exist on Docker Hub
Docker image tag does not exist on Docker Hub. GPU pod launches will fail for all users. Verify tag exists or use a valid one like 1.0.3-cu1290-torch290-ubuntu2204.
View Details
Location: api/transformerlab/compute_providers/runpod.py (lines 356)
Analysis
Docker image tag does not exist on Docker Hub. GPU pod launches will fail for all users
| What fails | RunPod GPU pod creation fails because the Docker image tag runpod/pytorch:1.0.3-cu1281-torch290-ubuntu2204 does not exist on Docker Hub. |
| Result | Pod fails to create with an image-not-found error because cu1281 variant only exists under v1.0.2, not v1.0.3. |
| Expected | Pod should launch successfully using a valid, published Docker image tag. |
| Impact | Complete regression of the RunPod GPU pod launch feature — every user without a custom template_id is affected. |
How to reproduce
1. Launch a RunPod GPU pod without a custom template_id so it hits the default_image fallback.
2. Observe pod creation error due to image-not-found.Patch Details
- default_image = "runpod/pytorch:1.0.3-cu1281-torch290-ubuntu2204"
+ default_image = "runpod/pytorch:1.0.3-cu1290-torch290-ubuntu2204"AI Fix Prompt
Fix this issue: Docker image tag does not exist on Docker Hub. GPU pod launches will fail for all users. Verify tag exists or use a valid one like 1.0.3-cu1290-torch290-ubuntu2204.
Location: api/transformerlab/compute_providers/runpod.py (lines 356)
Problem: RunPod GPU pod creation fails because the Docker image tag runpod/pytorch:1.0.3-cu1281-torch290-ubuntu2204 does not exist on Docker Hub.
Current behavior: Pod fails to create with an image-not-found error because cu1281 variant only exists under v1.0.2, not v1.0.3.
Expected: Pod should launch successfully using a valid, published Docker image tag.
Steps to reproduce: 1. Launch a RunPod GPU pod without a custom template_id so it hits the default_image fallback.
2. Observe pod creation error due to image-not-found.
Provide a code fix.
Tip: Reply with @paragon-run to automatically fix this issue
2 participants
