add deepseek-r1-distill-qwen-1.5b-rtx4070 model (#376)

fixes #374 

note that this requires new vllm 0.7.1 image

charts/models/values.yaml

@@ -307,6 +307,19 @@ catalog:
       - --disable-log-requests
     resourceProfile: nvidia-gpu-gh200:1
     targetRequests: 200
+  deepseek-r1-distill-qwen-1.5b-rtx4070:
+    enabled: false
+    features: ["TextGeneration"]
+    url: "hf://deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B"
+    engine: VLLM
+    env:
+      VLLM_USE_V1: "1"
+    args:
+    - --max-model-len=2048
+    - --max-num-batched-token=2048
+    - --max-num-seqs=8
+    - --kv-cache-dtype=fp8
+    resourceProfile: nvidia-gpu-rtx4070-8gb:1
   deepseek-r1-mi300x:
     enabled: false
     features: [TextGeneration]

manifests/models/deepseek-r1-distill-qwen-1.5b-rtx4070.yaml

@@ -0,0 +1,17 @@
+# Source: models/templates/models.yaml
+apiVersion: kubeai.org/v1
+kind: Model
+metadata:
+  name: deepseek-r1-distill-qwen-1.5b-rtx4070
+spec:
+  features: [TextGeneration]
+  url: hf://deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B
+  engine: VLLM
+  args:
+    - --max-model-len=2048
+    - --max-num-batched-token=2048
+    - --max-num-seqs=8
+    - --kv-cache-dtype=fp8
+  env:
+    VLLM_USE_V1: "1"
+  resourceProfile: nvidia-gpu-rtx4070-8gb:1