File size: 2,120 Bytes
a2d6f5f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 | [16:41:10] longeval: T=16384 prompt_len=4096 max_new=24 n=100 greedy=1 gpus=0,1
[16:41:10] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b2val_haystack.bin
[16:41:10] fan-out: 2 GPU(s), NP=1 -> 4 concurrent job(s)
[16:41:10] armA_c/niah_single already generated
[16:41:10] l1/niah_single already generated
[16:41:10] l2/niah_single already generated
[16:41:10] l3/niah_single already generated
[16:41:10] l3long/niah_single already generated
[16:41:10] 1 job(s) to run
[16:41:10] [gpu 0] l3long2k / niah_single
[16:42:35] scoring l3long2k / niah_single
=== niah_single (prompt_len=4096, n=100) ===
accuracy: 68/100 = 68.0%
depth acc (needle position in context)
0.00 11/20 55.0%
0.25 15/20 75.0%
0.50 10/20 50.0%
0.75 14/20 70.0%
1.00 18/20 90.0%
--- 2 sample generations ---
[HIT ] expect=8056020
got= 805602020.\n
[HIT ] expect=5343902
got= 5343902.\n
wrote decide_len_4096/scores/l3long2k_niah_single.csv
[16:42:35] building summary
=== per-task accuracy % (secondary/descriptive: ~10 per depth cell) ===
task armA_c l1 l2 l3 l3long l3long2k
niah_single 73.0 75.0 80.0 74.0 65.0 68.0
=== PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) ===
depth armA_c l1 l2 l3 l3long l3long2k
0.00 55.0+-20.0 50.0+-20.1 75.0+-17.8 65.0+-19.3 35.0+-19.3 55.0+-20.0
0.25 70.0+-18.7 80.0+-16.8 85.0+-15.4 85.0+-15.4 75.0+-17.8 75.0+-17.8
0.50 50.0+-20.1 60.0+-19.7 60.0+-19.7 50.0+-20.1 55.0+-20.0 50.0+-20.1
0.75 95.0+-11.4 95.0+-11.4 90.0+-13.7 80.0+-16.8 85.0+-15.4 70.0+-18.7
1.00 95.0+-11.4 90.0+-13.7 90.0+-13.7 90.0+-13.7 75.0+-17.8 90.0+-13.7
wrote decide_len_4096/summary.csv and decide_len_4096/depth_summary.csv
[16:42:35] DONE -> decide_len_4096/summary.csv
|