[CI/Build] Add Qwen2.5-VL-7B-Instruct ChartQA Accuracy Tests in CI (#21810)

Signed-off-by: Ye (Charlotte) Qi <yeq@meta.com>
Signed-off-by: zhewenli <zhewenli@meta.com>
Co-authored-by: Ye (Charlotte) Qi <yeq@meta.com>
Co-authored-by: Ye (Charlotte) Qi <ye.charlotte.qi@gmail.com>

.buildkite/lm-eval-harness/configs/Meta-Llama-3-8B-QQQ.yaml

+# For vllm script, with -t option (tensor parallel size).
+# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m HandH1998/QQQ-Llama-3-8b-g128 -b 32 -l 1000 -f 5 -t 1
+model_name: "HandH1998/QQQ-Llama-3-8b-g128"
+tasks:
+- name: "gsm8k"
+  metrics:
+  - name: "exact_match,strict-match"
+    value: 0.419
+  - name: "exact_match,flexible-extract"
+    value: 0.416
+limit: 1000
+num_fewshot: 5

.buildkite/lm-eval-harness/configs/Meta-Llama-4-Maverick-17B-128E-Instruct-FP8-MM.yaml

+# For hf script, without -t option (tensor parallel size).
+# bash .buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh -m meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8 -b 32 -l 100 -t 8
+model_name: "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8"
+backend: "vllm-vlm"
+tasks:
+- name: "chartqa"
+  metrics:
+  - name: "relaxed_accuracy,none"
+    value: 0.90
+limit: 100
+num_fewshot: 0

.buildkite/lm-eval-harness/configs/Meta-Llama-4-Maverick-17B-128E-Instruct-FP8.yaml

+# For hf script, without -t option (tensor parallel size).
+# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8 -b 32 -l 250 -t 8 -f 5
+model_name: "meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8"
+backend: "vllm-vlm"
+tasks:
+- name: "mmlu_pro"
+  metrics:
+  - name: "exact_match,custom-extract"
+    value: 0.80
+limit: 250 # will run on 250 * 14 subjects = 3500 samples
+num_fewshot: 5

.buildkite/lm-eval-harness/configs/Qwen2.5-VL-3B-Instruct-FP8-dynamic.yaml

-# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic -b auto -l 1319 -f 5 -t 1
+# For vllm script, with -t option (tensor parallel size)
+# bash .buildkite/lm-eval-harness/run-lm-eval-gsm-vllm-baseline.sh -m RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic -l 1319 -t 1
 model_name: "RedHatAI/Qwen2.5-VL-3B-Instruct-FP8-Dynamic"
 tasks:
 - name: "gsm8k"

.buildkite/lm-eval-harness/configs/Qwen2.5-VL-7B-Instruct.yaml

+# For vllm script, with -t option (tensor parallel size).
+# bash .buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh -m Qwen/Qwen2.5-VL-7B-Instruct -l 2500 -t 1
+
+model_name: "Qwen/Qwen2.5-VL-7B-Instruct"
+backend: "vllm-vlm"
+tasks:
+- name: "chartqa"
+  metrics:
+  - name: "relaxed_accuracy,none"
+    value: 0.855
+limit: 2500
+num_fewshot: 0

.buildkite/lm-eval-harness/configs/models-large-h100.txt

+Meta-Llama-4-Maverick-17B-128E-Instruct-FP8.yaml

.buildkite/lm-eval-harness/configs/models-mm-large-h100.txt

+Meta-Llama-4-Maverick-17B-128E-Instruct-FP8-MM.yaml

.buildkite/lm-eval-harness/configs/models-mm-small.txt

+Qwen2.5-VL-7B-Instruct.yaml
\ No newline at end of file

.buildkite/lm-eval-harness/run-lm-eval-chartqa-vllm-vlm-baseline.sh

+#!/bin/bash
+# We can use this script to compute baseline accuracy on chartqa for vllm.
+#
+# Make sure you have lm-eval-harness installed:
+#   pip install lm-eval==0.4.9
+
+usage() {
+    echo``
+    echo "Runs lm eval harness on ChartQA using multimodal vllm."
+    echo "This pathway is intended to be used to create baselines for "
+    echo "our correctness tests in vllm's CI."
+    echo
+    echo "usage: ${0} <options>"
+    echo
+    echo "  -m    - huggingface stub or local directory of the model"
+    echo "  -l    - limit number of samples to run"
+    echo "  -t    - tensor parallel size to run at"
+    echo
+}
+
+while getopts "m:l:t:" OPT; do
+  case ${OPT} in
+    m ) 
+        MODEL="$OPTARG"
+        ;;
+    l ) 
+        LIMIT="$OPTARG"
+        ;;
+    t ) 
+        TP_SIZE="$OPTARG"
+        ;;
+    \? ) 
+        usage
+        exit 1
+        ;;
+  esac
+done
+
+lm_eval --model vllm-vlm \
+  --model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE" \
+  --tasks chartqa \
+  --batch_size auto \
+  --apply_chat_template \
+  --limit $LIMIT

.buildkite/lm-eval-harness/run-lm-eval-mmlupro-vllm-baseline.sh

+#!/bin/bash
+# We can use this script to compute baseline accuracy on MMLUPRO for vllm.
+# We use this for fp8, which HF does not support.
+#
+# Make sure you have lm-eval-harness installed:
+#   pip install git+https://github.com/EleutherAI/lm-evaluation-harness.git@206b7722158f58c35b7ffcd53b035fdbdda5126d#egg=lm-eval[api]
+
+usage() {
+    echo``
+    echo "Runs lm eval harness on MMLU Pro using huggingface transformers."
+    echo "This pathway is intended to be used to create baselines for "
+    echo "our automated nm-test-accuracy workflow"
+    echo
+    echo "usage: ${0} <options>"
+    echo
+    echo "  -m    - huggingface stub or local directory of the model"
+    echo "  -l    - limit number of samples to run"
+    echo "  -f    - number of fewshot samples to use"
+    echo "  -t    - tensor parallel size to run at"
+    echo
+}
+
+while getopts "m:b:l:f:t:" OPT; do
+  case ${OPT} in
+    m )
+        MODEL="$OPTARG"
+        ;;
+    b )
+        BATCH_SIZE="$OPTARG"
+        ;;
+    l )
+        LIMIT="$OPTARG"
+        ;;
+    f )
+        FEWSHOT="$OPTARG"
+        ;;
+    t )
+        TP_SIZE="$OPTARG"
+        ;;
+    \? )
+        usage
+        exit 1
+        ;;
+  esac
+done
+
+lm_eval --model vllm \
+  --model_args "pretrained=$MODEL,tensor_parallel_size=$TP_SIZE,add_bos_token=true,trust_remote_code=true,max_model_len=4096" \
+  --tasks mmlu_pro --num_fewshot "$FEWSHOT" --limit "$LIMIT" \
+  --batch_size auto

.buildkite/lm-eval-harness/test_lm_eval_correctness.py

@@ -19,21 +19,27 @@ RTOL = 0.08
 def launch_lm_eval(eval_config, tp_size):
     trust_remote_code = eval_config.get("trust_remote_code", False)
     max_model_len = eval_config.get("max_model_len", 4096)
+    batch_size = eval_config.get("batch_size", "auto")
+    backend = eval_config.get("backend", "vllm")
     model_args = (
         f"pretrained={eval_config['model_name']},"
         f"tensor_parallel_size={tp_size},"
         f"enforce_eager=true,"
         f"add_bos_token=true,"
         f"trust_remote_code={trust_remote_code},"
-        f"max_model_len={max_model_len}"
+        f"max_model_len={max_model_len},"
     )
     results = lm_eval.simple_evaluate(
-        model="vllm",
+        model=backend,
         model_args=model_args,
         tasks=[task["name"] for task in eval_config["tasks"]],
         num_fewshot=eval_config["num_fewshot"],
         limit=eval_config["limit"],
-        batch_size="auto",
+        # TODO(yeq): using chat template w/ fewshot_as_multiturn is supposed help
+        # text models. however, this is regressing measured strict-match for
+        # existing text models in CI, so only apply it for mm.
+        apply_chat_template=backend == "vllm-vlm",
+        batch_size=batch_size,
     )
     return results
 

.buildkite/test-pipeline.yaml

@@ -734,6 +734,16 @@ steps:
     - pytest -v -s models/multimodal -m core_model --ignore models/multimodal/generation/test_whisper.py --ignore models/multimodal/processing
     - cd .. && VLLM_WORKER_MULTIPROC_METHOD=spawn pytest -v -s tests/models/multimodal/generation/test_whisper.py -m core_model  # Otherwise, mp_method="spawn" doesn't work
 
+- label: Multi-Modal Accuracy Eval (Small Models) # 50min
+  timeout_in_minutes: 70
+  working_dir: "/vllm-workspace/.buildkite/lm-eval-harness"
+  source_file_dependencies:
+  - vllm/multimodal/
+  - vllm/inputs/
+  - vllm/v1/core/
+  commands:
+  - pytest -s -v test_lm_eval_correctness.py --config-list-file=configs/models-mm-small.txt --tp-size=1
+
 - label: Multi-Modal Models Test (Extended) 1
   mirror_hardwares: [amdexperimental]
   optional: true