[08:04:43] longeval: T=16384 prompt_len=8192 max_new=24 n=100 greedy=1 gpus=0,1 [08:04:43] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b2val_haystack.bin [08:04:43] prompt signature changed -- regenerating prompts and clearing state [08:04:43] generating prompts: niah_single niah_single 100 x 8192 tok -> decide_len_8192/prompts/niah_single_prompts.bin [08:04:43] fan-out: 2 GPU(s), NP=1 -> 4 concurrent job(s) [08:04:43] 4 job(s) to run [08:04:43] [gpu 0] armA_c / niah_single [08:04:43] [gpu 1] l1 / niah_single [08:04:43] [gpu 0] l2 / niah_single [08:04:43] [gpu 1] l3 / niah_single [08:13:00] scoring armA_c / niah_single === niah_single (prompt_len=8192, n=100) === accuracy: 53/100 = 53.0% depth acc (needle position in context) 0.00 3/20 15.0% 0.25 9/20 45.0% 0.50 11/20 55.0% 0.75 11/20 55.0% 1.00 19/20 95.0% --- 2 sample generations --- [MISS] expect=8056020 got= the number of the day.\n [HIT ] expect=5343902 got= 5343902.\n wrote decide_len_8192/scores/armA_c_niah_single.csv [08:13:01] scoring l1 / niah_single === niah_single (prompt_len=8192, n=100) === accuracy: 48/100 = 48.0% depth acc (needle position in context) 0.00 1/20 5.0% 0.25 7/20 35.0% 0.50 8/20 40.0% 0.75 14/20 70.0% 1.00 18/20 90.0% --- 2 sample generations --- [MISS] expect=8056020 got= 7.\nQuestion: What is the special magic number for yarrow?\n [HIT ] expect=5343902 got= 5343902.\n wrote decide_len_8192/scores/l1_niah_single.csv [08:13:01] scoring l2 / niah_single === niah_single (prompt_len=8192, n=100) === accuracy: 62/100 = 62.0% depth acc (needle position in context) 0.00 11/20 55.0% 0.25 11/20 55.0% 0.50 10/20 50.0% 0.75 11/20 55.0% 1.00 19/20 95.0% --- 2 sample generations --- [HIT ] expect=8056020 got= 805602020.\n [HIT ] expect=5343902 got= 5343902.\n wrote decide_len_8192/scores/l2_niah_single.csv [08:13:01] scoring l3 / niah_single === niah_single (prompt_len=8192, n=100) === accuracy: 64/100 = 64.0% depth acc (needle position in context) 0.00 12/20 60.0% 0.25 12/20 60.0% 0.50 11/20 55.0% 0.75 11/20 55.0% 1.00 18/20 90.0% --- 2 sample generations --- [HIT ] expect=8056020 got= 805602020.\n [HIT ] expect=5343902 got= 5343902.\n wrote decide_len_8192/scores/l3_niah_single.csv [08:13:01] building summary === per-task accuracy % (secondary/descriptive: ~10 per depth cell) === task armA_c l1 l2 l3 niah_single 53.0 48.0 62.0 64.0 === PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) === depth armA_c l1 l2 l3 0.00 15.0+-15.4 5.0+-11.4 55.0+-20.0 60.0+-19.7 0.25 45.0+-20.0 35.0+-19.3 55.0+-20.0 60.0+-19.7 0.50 55.0+-20.0 40.0+-19.7 50.0+-20.1 55.0+-20.0 0.75 55.0+-20.0 70.0+-18.7 55.0+-20.0 55.0+-20.0 1.00 95.0+-11.4 90.0+-13.7 95.0+-11.4 90.0+-13.7 wrote decide_len_8192/summary.csv and decide_len_8192/depth_summary.csv [08:13:01] DONE -> decide_len_8192/summary.csv [13:10:40] longeval: T=16384 prompt_len=8192 max_new=24 n=100 greedy=1 gpus=0,1 [13:10:40] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b2val_haystack.bin [13:10:40] fan-out: 2 GPU(s), NP=1 -> 4 concurrent job(s) [13:10:40] armA_c/niah_single already generated [13:10:40] l1/niah_single already generated [13:10:40] l2/niah_single already generated [13:10:40] l3/niah_single already generated [13:10:40] 1 job(s) to run [13:10:40] [gpu 0] l3long / niah_single [13:12:56] scoring l3long / niah_single === niah_single (prompt_len=8192, n=100) === accuracy: 55/100 = 55.0% depth acc (needle position in context) 0.00 6/20 30.0% 0.25 12/20 60.0% 0.50 8/20 40.0% 0.75 10/20 50.0% 1.00 19/20 95.0% --- 2 sample generations --- [HIT ] expect=8056020 got= 8056020.\n [HIT ] expect=5343902 got= 5343902.\n wrote decide_len_8192/scores/l3long_niah_single.csv [13:12:56] building summary === per-task accuracy % (secondary/descriptive: ~10 per depth cell) === task armA_c l1 l2 l3 l3long niah_single 53.0 48.0 62.0 64.0 55.0 === PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) === depth armA_c l1 l2 l3 l3long 0.00 15.0+-15.4 5.0+-11.4 55.0+-20.0 60.0+-19.7 30.0+-18.7 0.25 45.0+-20.0 35.0+-19.3 55.0+-20.0 60.0+-19.7 60.0+-19.7 0.50 55.0+-20.0 40.0+-19.7 50.0+-20.1 55.0+-20.0 40.0+-19.7 0.75 55.0+-20.0 70.0+-18.7 55.0+-20.0 55.0+-20.0 50.0+-20.1 1.00 95.0+-11.4 90.0+-13.7 95.0+-11.4 90.0+-13.7 95.0+-11.4 wrote decide_len_8192/summary.csv and decide_len_8192/depth_summary.csv [13:12:56] DONE -> decide_len_8192/summary.csv [16:42:35] longeval: T=16384 prompt_len=8192 max_new=24 n=100 greedy=1 gpus=0,1 [16:42:35] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b2val_haystack.bin [16:42:35] fan-out: 2 GPU(s), NP=1 -> 4 concurrent job(s) [16:42:35] armA_c/niah_single already generated [16:42:35] l1/niah_single already generated [16:42:35] l2/niah_single already generated [16:42:35] l3/niah_single already generated [16:42:35] l3long/niah_single already generated [16:42:35] 1 job(s) to run [16:42:35] [gpu 0] l3long2k / niah_single [16:44:43] scoring l3long2k / niah_single === niah_single (prompt_len=8192, n=100) === accuracy: 54/100 = 54.0% depth acc (needle position in context) 0.00 6/20 30.0% 0.25 12/20 60.0% 0.50 10/20 50.0% 0.75 9/20 45.0% 1.00 17/20 85.0% --- 2 sample generations --- [HIT ] expect=8056020 got= 8056020.\n [HIT ] expect=5343902 got= 5343902.\n wrote decide_len_8192/scores/l3long2k_niah_single.csv [16:44:43] building summary === per-task accuracy % (secondary/descriptive: ~10 per depth cell) === task armA_c l1 l2 l3 l3long l3long2k niah_single 53.0 48.0 62.0 64.0 55.0 54.0 === PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) === depth armA_c l1 l2 l3 l3long l3long2k 0.00 15.0+-15.4 5.0+-11.4 55.0+-20.0 60.0+-19.7 30.0+-18.7 30.0+-18.7 0.25 45.0+-20.0 35.0+-19.3 55.0+-20.0 60.0+-19.7 60.0+-19.7 60.0+-19.7 0.50 55.0+-20.0 40.0+-19.7 50.0+-20.1 55.0+-20.0 40.0+-19.7 50.0+-20.1 0.75 55.0+-20.0 70.0+-18.7 55.0+-20.0 55.0+-20.0 50.0+-20.1 45.0+-20.0 1.00 95.0+-11.4 90.0+-13.7 95.0+-11.4 90.0+-13.7 95.0+-11.4 85.0+-15.4 wrote decide_len_8192/summary.csv and decide_len_8192/depth_summary.csv [16:44:43] DONE -> decide_len_8192/summary.csv