feat: nvidia nim blueprint for drug discovery #948
Conversation
|
/gcbrun |
|
|
||
| Here you will find the NVIDIA NIM Blueprints that can be provisioned to run on GKE. These are good for proof of concepts only. | ||
|
|
||
| - [ ] [Generative Virtual Screening for Drug Discovery](./drugdiscovery/README.md) uses 3 NIMs |
There was a problem hiding this comment.
Is there a reason you made this an empty checkbox? Would a numbered or bulleted list make more sense?
| * **kubectl:** kubectl command-line tool installed and configured. | ||
|
|
||
| Clone the repo before proceeding further: | ||
| `git clone https://github.com/GoogleCloudPlatform/ai-on-gke` |
There was a problem hiding this comment.
I think this would format better using the multi-line code block like you do below:
git clone https://github.com/GoogleCloudPlatform/ai-on-gke
cd ai-on-gke/tutorials-and-examples/nvidia-nim/blueprints/drugdiscovery| gcloud container clusters create "${CLUSTER_NAME}" \ | ||
| --project="${PROJECT_ID}" \ | ||
| --num-nodes=1 --location="${ZONE}" \ | ||
| --machine-type=e2-standard-16 \ |
There was a problem hiding this comment.
Do we really need a 16 core machine in the default node pool for management? Do we run anything on this node as part of the guide or is this just for system components?
There was a problem hiding this comment.
good catch. changed to e2-standard-2.
| --max-time 900 \ | ||
| -X POST \ | ||
| -i \ | ||
| "http://localhost:62921/protein-structure/alphafold2/predict-structure-from-sequence" \ |
There was a problem hiding this comment.
Where does this port number come from? In the port forward command it looks like you only set the destination port; should we set a source port too so that it is consistent for this next step in the guide?
There was a problem hiding this comment.
switched to a fixed port and instructions to update if unavailable.
| "min_similarity": 0.7, | ||
| "iterations": 10 | ||
| }' \ | ||
| "http://localhost:49530/generate" |
There was a problem hiding this comment.
Same question about the port here.
| - name: NGC_API_KEY | ||
| valueFrom: | ||
| secretKeyRef: | ||
| name: ngc-api-key # Replace with your actual secret name |
There was a problem hiding this comment.
Do we need to add a step to the guide to create this secret? It looks like you create a docker registry secret but not the secret used by the application.
I found the secret below at the bottom of this file. We either need to update step 7 to let people know to edit this file before applying it (to set the secret value) or we should update step 6 to create two secrets instead of creating the ngc-api-key secret in this file.
I also this the comment here isn't necessary, because we will be setting the secret name correctly (either in this file or in the guide) so you shouldn't have to verify the secret name as the comment suggests.
There was a problem hiding this comment.
simplified this, created the secret key from cli as it was in a env variable and removed this part.
| @@ -0,0 +1,110 @@ | |||
| import requests | |||
|
|
|||
| AF2_HOST = "http://localhost:<replace with port number>" | |||
There was a problem hiding this comment.
I think if we change the port forward command to use a consistent local port number we can just insert it here and folks wouldn't need to modify this file before running it.
|
@roberthbailey comments addressed. ready for review. |
Generative virtual screening for drug discovery NIM Blueprint on GKE