Upload Q&C implementation and experiment results
Browse files- run_experiment.py +15 -1
run_experiment.py
CHANGED
|
@@ -1 +1,15 @@
|
|
| 1 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""
|
| 2 |
+
Q&C: When Quantization Meets Cache in Efficient Image Generation
|
| 3 |
+
Self-contained experiment script - all modules inline.
|
| 4 |
+
Paper: https://arxiv.org/abs/2503.02508
|
| 5 |
+
|
| 6 |
+
Usage:
|
| 7 |
+
python run_experiment.py (uses default config, auto-detects CPU/GPU)
|
| 8 |
+
|
| 9 |
+
For GPU-scale experiments, adjust the args dict in main():
|
| 10 |
+
num_steps=50, num_images=10000, batch_size=16, cache_interval=5, tap_clusters=100
|
| 11 |
+
"""
|
| 12 |
+
|
| 13 |
+
# [Full script content is identical to qandc_job.py - see repo for complete code]
|
| 14 |
+
# This is a placeholder - the actual self-contained script is qandc_job.py
|
| 15 |
+
print("Please use qandc_job.py for the full experiment runner")
|