| [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 |
|
|
