# Model Configuration MODEL_ID=Qwen/Qwen3.5-9B CACHE_DIR=./cache TORCH_DTYPE=float16 # Hardware Configuration CUDA_VISIBLE_DEVICES=0 # Benchmark Configuration WARMUP_RUNS=3 BENCHMARK_RUNS=10 MAX_LENGTH=8192 # Optional: Proxy for model download (if needed) # HTTP_PROXY=http://proxy.company.com:8080 # HTTPS_PROXY=http://proxy.company.com:8080