TPRD-1710: Update default branches post-25.09 (#99)

Co-authored-by: Yingge He <[email protected]>

Author: mc-nv

ci/L0_additional_outputs_vllm/test.sh

@@ -30,7 +30,6 @@ source ../common/util.sh
 
 pip3 install pytest==8.1.1
 pip3 install tritonclient[grpc]
-pip3 install "transformers<=4.53.3" # TODO:DLIS-8441 remove this dependency
 
 # Prepare Model
 rm -rf models vllm_baseline_output.pkl && mkdir -p models

ci/L0_backend_vllm/test.sh

@@ -29,7 +29,6 @@ RET=0
 SUBTESTS="accuracy_test request_cancellation enabled_stream vllm_backend metrics_test"
 
 python3 -m pip install tritonclient[grpc]
-python3 -m pip install "transformers<=4.53.3" # TODO:DLIS-8441 remove this dependency
 
 for TEST in ${SUBTESTS}; do
     (cd ${TEST} && bash -ex test.sh && cd ..)

ci/L0_check_health_vllm/test.sh

@@ -30,7 +30,6 @@ source ../common/util.sh
 
 pip3 install pytest==8.1.1
 pip3 install tritonclient[grpc]
-pip3 install "transformers<=4.53.3" # TODO:DLIS-8441 remove this dependency
 
 RET=0
 

ci/L0_multi_gpu_vllm/multi_lora/test.sh

@@ -41,6 +41,9 @@ EXPECTED_NUM_TESTS=2
 GENERATE_ENDPOINT="localhost:8000/v2/models/vllm_llama_multi_lora/generate"
 CHECK_FOR_ERROR=true
 
+export C_INCLUDE_PATH=/usr/local/cuda/include:$C_INCLUDE_PATH
+export TRITON_PTXAS_PATH=/usr/local/cuda/bin/ptxas
+
 make_api_call() {
     local endpoint="$1"
     local data="$2"

ci/L0_multi_gpu_vllm/test.sh

@@ -29,7 +29,6 @@ RET=0
 SUBTESTS="vllm_backend multi_lora"
 
 python3 -m pip install tritonclient[grpc]
-python3 -m pip install "transformers<=4.53.3" # TODO:DLIS-8441 remove this dependency
 
 for TEST in ${SUBTESTS}; do
     (cd ${TEST} && bash -ex test.sh && cd ..)

ci/L0_multi_gpu_vllm/vllm_backend/test.sh

@@ -36,6 +36,9 @@ CLIENT_PY="./vllm_multi_gpu_test.py"
 SAMPLE_MODELS_REPO="../../../samples/model_repository"
 EXPECTED_NUM_TESTS=1
 
+export C_INCLUDE_PATH=/usr/local/cuda/include:$C_INCLUDE_PATH
+export TRITON_PTXAS_PATH=/usr/local/cuda/bin/ptxas
+
 ### Helpers
 function validate_file_contains() {
     local KEY="${1}"