Upgraded llm-d-inference-sim to v0.10.2 (#2030)

* Upgraded llm-d-inference-sim to v0.10.2

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

* Upgraded llm-d-inference-sim to v0.10.2

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

---------

Signed-off-by: Shmuel Kallner <kallner@il.ibm.com>

Author: shmuelk

.github/workflows/ci-pr-checks.yaml

@@ -222,7 +222,7 @@ jobs:
             archive: llm-d-router-endpoint-picker
           - name: simulator
             source: pull
-            image: ghcr.io/llm-d/llm-d-inference-sim:v0.9.2
+            image: ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
             archive: llm-d-inference-sim
     name: e2e-image-common (${{ matrix.image.name }})
     steps:

DEVELOPMENT.md

@@ -85,7 +85,7 @@ Creates a new `kind` cluster (or reuses an existing one) in the `default` namesp
 > [!NOTE]
 > You can pre-pull external images to avoid slow downloads:
 > ```
-> docker pull ghcr.io/llm-d/llm-d-inference-sim:v0.9.2
+> docker pull ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
 > docker pull vllm/vllm-openai-cpu:v0.21.0
 > ```
 
@@ -528,7 +528,7 @@ a shared PVC (`ec-cache-pvc`) for encoder embeddings transfer.
 
 | Variable | Default | Description |
 |---|---|---|
-| `VLLM_IMAGE` | `ghcr.io/llm-d/llm-d-inference-sim:v0.9.2` | vLLM container image to deploy. Can be a simulator or a real vLLM image (e.g., `vllm/vllm-openai:v0.16.0`). Defaults to the simulator image. |
+| `VLLM_IMAGE` | `ghcr.io/llm-d/llm-d-inference-sim:v0.10.2` | vLLM container image to deploy. Can be a simulator or a real vLLM image (e.g., `vllm/vllm-openai:v0.16.0`). Defaults to the simulator image. |
 | `VLLM_SIM_MODE` | `echo` | Simulator response mode. `echo` returns the input prompt as the response (useful for routing validation). `random` returns random sentences from a pre-defined bank. Only applies when using the simulator overlay. |
 
 ### Cleanup
@@ -622,7 +622,7 @@ kubectl --context kind-e2e-tests get pods
 | `VLLM_EXTRA_ARGS_E` | _(empty)_ | Additional flags for the Encoder vLLM container (e.g. `--mm-processor-kwargs={}`) |
 | `VLLM_EXTRA_ARGS_P` | _(empty)_ | Additional flags for the Prefill vLLM container (e.g. `--gpu-memory-utilization=0.9`) |
 | `VLLM_EXTRA_ARGS_D` | _(empty)_ | Additional flags for the Decode vLLM container (e.g. `--tensor-parallel-size=2`) |
-| `VLLM_IMAGE` | `ghcr.io/llm-d/llm-d-inference-sim:v0.9.2` | vLLM container image to deploy. Can be a simulator or a real vLLM image (e.g., `vllm/vllm-openai:v0.16.0`) |
+| `VLLM_IMAGE` | `ghcr.io/llm-d/llm-d-inference-sim:v0.10.2` | vLLM container image to deploy. Can be a simulator or a real vLLM image (e.g., `vllm/vllm-openai:v0.16.0`) |
 | `VLLM_SIM_MODE` | `echo` | Simulator response mode. Supported values: `echo` (returns the input prompt as the response), `random` (returns a random sentence from a pre-defined bank) |
 | `SIDECAR_IMAGE` | `ghcr.io/llm-d/llm-d-router-disagg-sidecar:dev` | Routing sidecar image loaded into the Kind cluster |
 | `VLLM_RENDER_IMAGE` | `vllm/vllm-openai-cpu:v0.21.0` | vLLM renderer image loaded into the Kind cluster |

Makefile

@@ -33,7 +33,7 @@ SIDECAR_IMAGE_TAG_BASE ?= $(IMAGE_REGISTRY)/$(SIDECAR_IMAGE_NAME)
 SIDECAR_TAG ?= dev
 export SIDECAR_IMAGE ?= $(SIDECAR_IMAGE_TAG_BASE):$(SIDECAR_TAG)
 
-VLLM_SIMULATOR_TAG ?= v0.9.2
+VLLM_SIMULATOR_TAG ?= v0.10.2
 VLLM_SIMULATOR_TAG_BASE ?= $(IMAGE_REGISTRY)/$(VLLM_SIMULATOR_IMAGE_NAME)
 export VLLM_IMAGE ?= $(VLLM_SIMULATOR_TAG_BASE):$(VLLM_SIMULATOR_TAG)
 

config/manifests/vllm/sim-deployment.yaml

@@ -15,7 +15,7 @@ spec:
     spec:
       containers:
       - name: vllm-sim
-        image: ghcr.io/llm-d/llm-d-inference-sim:v0.9.2
+        image: ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
         imagePullPolicy: Always
         args:
         - --model

config/manifests/vllm/sim-grpc-deployment.yaml

@@ -15,7 +15,7 @@ spec:
     spec:
       containers:
       - name: vllm-sim
-        image: ghcr.io/llm-d/llm-d-inference-sim:v0.9.2
+        image: ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
         imagePullPolicy: Always
         args:
         - --model

deploy/environments/dev/README.md

@@ -67,7 +67,7 @@ Variables substituted at deploy time via `envsubst` or Go test `substituteMany`:
 | `DISAGG_E` | Deploy a separate Encoder pod (`true`/`false`) | `false` |
 | `DISAGG_P` | Deploy a separate Prefill pod (`true`/`false`) | `false` |
 | `VLLM_SIM_MODE` | Simulator response mode: `echo` (returns input) or `random` (random sentences) | `echo` |
-| `VLLM_IMAGE` | vLLM container image (simulator or real) | `ghcr.io/llm-d/llm-d-inference-sim:v0.9.2` |
+| `VLLM_IMAGE` | vLLM container image (simulator or real) | `ghcr.io/llm-d/llm-d-inference-sim:v0.10.2` |
 | `SIDECAR_IMAGE` | Routing sidecar image | `ghcr.io/llm-d/llm-d-router-disagg-sidecar:dev` |
 | `VLLM_RENDER_IMAGE` | vLLM render sidecar image | `vllm/vllm-openai-cpu:v0.21.0` |
 | `MODEL_NAME` | Model name passed to vLLM. Can be a real HuggingFace model (e.g. `TinyLlama/TinyLlama-1.1B-Chat-v1.0`, `Qwen/Qwen3-VL-2B-Instruct`) or an arbitrary name when using the simulator (e.g. `food-review`) | `food-review` |

scripts/kind-dev-env.sh

@@ -23,7 +23,7 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 : "${IMAGE_REGISTRY:=ghcr.io/llm-d}"
 
 # Set a default VLLM_SIMULATOR_TAG if not provided
-export VLLM_SIMULATOR_TAG="${VLLM_SIMULATOR_TAG:-v0.9.2}"
+export VLLM_SIMULATOR_TAG="${VLLM_SIMULATOR_TAG:-v0.10.2}"
 
 # VLLM_IMAGE: the vLLM container image to deploy. Can be a simulator or real vLLM image
 # (e.g., vllm/vllm-openai:v0.16.0 for production). Defaults to the simulator image.

scripts/pull_images.sh

@@ -7,7 +7,7 @@ echo "Using container tool: ${CONTAINER_RUNTIME}"
 # Set a default EPP_TAG if not provided
 EPP_TAG="${EPP_TAG:-dev}"
 # Set a default VLLM_SIMULATOR_TAG if not provided
-VLLM_SIMULATOR_TAG="${VLLM_SIMULATOR_TAG:-v0.9.2}"
+VLLM_SIMULATOR_TAG="${VLLM_SIMULATOR_TAG:-v0.10.2}"
 # Set the default routing side car image tag
 SIDECAR_TAG="${SIDECAR_TAG:-dev}"
 

test/e2e/e2e_suite_test.go

@@ -76,7 +76,7 @@ var (
 
 	containerRuntime = env.GetEnvString("CONTAINER_RUNTIME", "docker", ginkgo.GinkgoLogr)
 	eppImage         = env.GetEnvString("EPP_IMAGE", "ghcr.io/llm-d/llm-d-router-endpoint-picker:dev", ginkgo.GinkgoLogr)
-	vllmSimImage     = env.GetEnvString("VLLM_IMAGE", "ghcr.io/llm-d/llm-d-inference-sim:v0.9.2", ginkgo.GinkgoLogr)
+	vllmSimImage     = env.GetEnvString("VLLM_IMAGE", "ghcr.io/llm-d/llm-d-inference-sim:v0.10.2", ginkgo.GinkgoLogr)
 	sideCarImage     = env.GetEnvString("SIDECAR_IMAGE", "ghcr.io/llm-d/llm-d-router-disagg-sidecar:dev", ginkgo.GinkgoLogr)
 	vllmRenderImage  = env.GetEnvString("VLLM_RENDER_IMAGE", "vllm/vllm-openai-cpu:v0.21.0", ginkgo.GinkgoLogr)
 	loadRenderImage  = env.GetEnvBool("LOAD_VLLM_RENDER_IMAGE", true, ginkgo.GinkgoLogr)

test/perf/config/llm-d-sim-deployment.yaml

@@ -62,7 +62,7 @@ spec:
           valueFrom:
             fieldRef:
               fieldPath: status.podIP
-        image: ghcr.io/llm-d/llm-d-inference-sim:v0.9.0
+        image: ghcr.io/llm-d/llm-d-inference-sim:v0.10.2
         imagePullPolicy: IfNotPresent
         livenessProbe:
           httpGet: