Skip to content

Commit

Permalink
Add more scaler shortcuts
Browse files Browse the repository at this point in the history
  • Loading branch information
SawyerCzupka committed Nov 8, 2024
1 parent d068278 commit 10670fc
Showing 1 changed file with 11 additions and 3 deletions.
14 changes: 11 additions & 3 deletions deployment/Makefile
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# Variables
SERVICE_URL := http://vllm-svc:8000
AUTOSCALER_URL := http://vllm-autoscaler-service:80
AUTOSCALER_URL := http://vllm-autoscaler-service:8000
CONTENT_TYPE := application/json
MODEL_NAME := mistralai/Mistral-7B-Instruct-v0.3
PROMPT := \"Give a detailed summary of the various activities to do in San Francisco.\"
Expand All @@ -15,9 +15,17 @@ list-models:
@echo "Listing available models..."
@curl -s -X GET "$(SERVICE_URL)/v1/models" | jq

auto-health:
scaler-health:
@echo "Testing autoscaler health endpoint..."
@curl -s -X GET "$(AUTOSCALER_URL)" | jq
@curl -X GET "$(AUTOSCALER_URL)/health" | jq

scaler-up:
@echo "Scaling up..."
@curl -s -X POST "$(AUTOSCALER_URL)/scale/1" | jq

scaler-down:
@echo "Scaling down..."
@curl -X POST "$(AUTOSCALER_URL)/scale/0" | jq

generate-tokens:
@echo "Generating tokens..."
Expand Down

0 comments on commit 10670fc

Please sign in to comment.