Cascade Profiler

GPU workload profiling, Nsight Systems timing reconciliation, and NVBit instruction tracing for vLLM. Python 3.10+ can import and inspect runs; GPU workloads require CUDA, vLLM, and the tracing tools. Modal images use Python 3.12.

pip install 'cascade-profiler[modal]'

from cascade_profiler import Workload, profile

workload = Workload("facebook/opt-125m", ["Hello, my name is"], max_tokens=32, repeats=4)
result = profile(workload, "/data/phases", trace_steps="plan", trace_repeat=1)
print(result.run_dir, result.stage_wall_s)

profile needs a GPU environment with vLLM, Nsight Systems and the NVBit tracer installed. cascade_profiler.image() supplies the optional Modal GPU image; register(app, volume, gpu="H100!") registers only the GPU .profile function. See PROFILER_RUNS.md and examples/modal_app.py.

Profile, then simulate

From this checkout:

modal run accelsim_trace.py::e2e_main --run-name my-run --trace-steps plan
# Timing only (no CPU simulator image is built):
modal run accelsim_trace.py::e2e_main --run-name timing-run --timing-only

Accel-Sim CPU execution lives in the separate cascade.sim simulator package on the simulator branch of atoniolo76/cascade. From that checkout, using the same Modal volume:

modal run scripts/simulate_accelsim.py --run-name my-run --gpu-config SM90_H100 --propagate interp

The run directory is the interface: run.json, corrected_timeline.json, trace_alignment.json, and traced/traces/ are produced by the profiler. The simulator consumes them and writes sim_durations.json and sim/. The profiler does not build or execute the CPU simulator.