Dear authors, I wonder why TOTAL_VIDEO_CHUNK=CHUNK_TIL_VIDEO_END_CAP=48 instead of 49?
According to env.py/fixed_env.py, video chunk counter first +=1, and then is compared to TOTAL_VIDEO_CHUNK=48, so it only tests 48 chunks out of all 49 chunks, which corresponds to the log with only 48 chunks results.
I tried to set the total chunks as 49 and find that, MPC algorithm can perform quite differently. Specifically, for normal_bus_1 trace, the average QoE/reward=0.328 when TOTAL_VIDEO_CHUNK=CHUNK_TIL_VIDEO_END_CAP=49, and QoE/reward=1.857 when TOTAL_VIDEO_CHUNK=CHUNK_TIL_VIDEO_END_CAP=48.
The attached files give detailed log for this trace, where each line denotes timestamp, bitrate, buffer_size, rebuf, video_chunk_size, delay, reward.
Thank you!
chunk_48_norway_bus_1.txtchunk_49_norway_bus_1.txt
MardenSSS
commented
1 year ago
Dear authors, I wonder why TOTAL_VIDEO_CHUNK=CHUNK_TIL_VIDEO_END_CAP=48 instead of 49? According to env.py/fixed_env.py, video chunk counter first +=1, and then is compared to TOTAL_VIDEO_CHUNK=48, so it only tests 48 chunks out of all 49 chunks, which corresponds to the log with only 48 chunks results. I tried to set the total chunks as 49 and find that, MPC algorithm can perform quite differently. Specifically, for normal_bus_1 trace, the average QoE/reward=0.328 when TOTAL_VIDEO_CHUNK=CHUNK_TIL_VIDEO_END_CAP=49, and QoE/reward=1.857 when TOTAL_VIDEO_CHUNK=CHUNK_TIL_VIDEO_END_CAP=48. The attached files give detailed log for this trace, where each line denotes timestamp, bitrate, buffer_size, rebuf, video_chunk_size, delay, reward. Thank you! chunk_48_norway_bus_1.txt chunk_49_norway_bus_1.txt