Hi, I'm encountering an issue when trying to update the configuration of a serverless endpoint.
Even if I don't change any fields, attempting to save/update the config results in the following error:
Invalid GPU Pool ID: "NVIDIA H200". The gpuIds field accepts a comma-separated list of GPU pool IDs (e.g. "ADA_24,AMPERE_80"). To exclude specific GPU types within a pool, prefix them with "-" (e.g. "AMPERE_24,-NVIDIA L4"). Valid pool IDs: AMPERE_16, AMPERE_24, ADA_24, AMPERE_48, ADA_48_PRO, AMPERE_80, ADA_80_PRO, HOPPER_141, ADA_32_PRO, BLACKWELL_96, BLACKWELL_180.
https://docs.runpod.io/references/gpu-types#gpu-pools
Because of this error, I'm also unable to modify the GPU pool (for example, remove H200 from the pool), as the update action always fails.
Endpoint ID: 258wgsnsb82bme
Expected behavior:
- I should be able to update the endpoint configuration without errors if no invalid changes are made
- I should be able to remove unsupported GPU types (like H200) from the pool
Actual behavior:
- The update fails with an "Invalid GPU Pool ID" error
- The error persists even when no changes are made
- The UI/API seems to be stuck with an invalid GPU configuration
Additional context:
It looks like "NVIDIA H200" is being treated as a GPU pool ID, while only pool IDs like HOPPER_141 are valid. Possibly the system saved an invalid value earlier or there's a mismatch between UI and backend validation.