Upload folder using huggingface_hub

logs/lenbucket/RESULTS.md

@@ -0,0 +1,102 @@
+# Training document length vs 16k retrieval -- 2026-09-08
+
+All arms: 400 steps, global batch 524,288, from run7' (run 6 step 750),
+YARN_ORIG_MAXPOS=4096 matched, Muon+ZeRO-1 bf16, 4xH100.
+Only the training data differs. NIAH scored generatively (substring match),
+greedy, N=100 per cell, one shared haystack held out from every arm.
+
+| arm | training docs |
+|---|---|
+| armA_c | flux plain text, median 826 / p90 2,721 tok |
+| l1 | b1_4096_8192, docs in [4096, 8192) |
+| l2 | b2_8192_16384, docs in [8192, 16384) |
+
+## PROMPT_LEN=1024
+```
+task,armA_c,l1,l2
+niah_single,94.0,95.0,96.0
+val_backward,1.0,1.0,1.0
+```
+
+## PROMPT_LEN=2048
+```
+task,armA_c,l1,l2
+niah_single,94.0,96.0,94.0
+val_backward,2.0,2.0,1.0
+```
+
+## PROMPT_LEN=4096
+```
+task,armA_c,l1,l2
+niah_single,84.0,88.0,92.0
+val_backward,0.0,0.0,0.0
+```
+
+## PROMPT_LEN=8192
+```
+task,armA_c,l1,l2
+niah_single,62.0,67.0,78.0
+val_backward,2.0,1.0,2.0
+```
+
+## PROMPT_LEN=16360
+```
+task,armA_c,l1,l2
+niah_single,33.0,39.0,48.0
+val_backward,1.0,1.0,0.0
+```
+
+## Training losses
+```
+l1: loss: 2.458802
+l2: loss: 2.794149
+```
+
+NOTE: losses are NOT comparable across arms -- each is measured on its own
+data distribution. Longer documents are a different prediction problem, so
+a higher loss on b2 than on flux says nothing about model quality. Only the
+retrieval numbers compare.
+
+
+┌─────────┬────────────────────────┬────────────────┬─────────────────┐
+│ context │ armA_c (flux, med 826) │ l1 (docs 4-8k) │ l2 (docs 8-16k) │
+├─────────┼────────────────────────┼────────────────┼─────────────────┤
+│ 1k      │ 94.0 ±4.8              │ 95.0 ±4.5      │ 96.0 ±4.1       │
+├─────────┼────────────────────────┼────────────────┼─────────────────┤
+│ 2k      │ 94.0 ±4.8              │ 96.0 ±4.1      │ 94.0 ±4.8       │
+├─────────┼────────────────────────┼────────────────┼─────────────────┤
+│ 4k      │ 84.0 ±7.2              │ 88.0 ±6.4      │ 92.0 ±5.4       │
+├─────────┼────────────────────────┼────────────────┼─────────────────┤
+│ 8k      │ 62.0 ±9.3              │ 67.0 ±9.1      │ 78.0 ±8.0       │
+├─────────┼────────────────────────┼────────────────┼─────────────────┤
+│ 16360   │ 33.0 ±9.1              │ 39.0 ±9.4      │ 48.0 ±9.6       │
+└─────────┴────────────────────────┴────────────────┴─────────────────┘
+
+┌─────────┬───────────┬───────────┬─────────┬─────────────────┐
+│ context │ l1 − armA │ l2 − armA │ l2 − l1 │   resolvable    │
+├─────────┼───────────┼───────────┼─────────┼─────────────────┤
+│ 1k      │ +1.0      │ +2.0      │ +1.0    │ no              │
+├─────────┼───────────┼───────────┼─────────┼─────────────────┤
+│ 2k      │ +2.0      │ +0.0      │ −2.0    │ no              │
+├─────────┼───────────┼───────────┼─────────┼─────────────────┤
+│ 4k      │ +4.0      │ +8.0      │ +4.0    │ no (bar ±9.0)   │
+├─────────┼───────────┼───────────┼─────────┼─────────────────┤
+│ 8k      │ +5.0      │ +16.0     │ +11.0   │ YES (bar ±12.3) │
+├─────────┼───────────┼───────────┼─────────┼─────────────────┤
+│ 16360   │ +6.0      │ +15.0     │ +9.0    │ YES (bar ±13.2) │
+└─────────┴───────────┴─��─────────┴─────────┴─────────────────┘
+
+
+The effect is monotone in training document length and grows with eval context. Nothing distinguishable at 1-2k, +8pp at 4k, then +16pp at 8k and +15pp at 16k. Ordering is l2 > l1 > armA_c at every length where there's signal — three independent lengths agreeing on the same ordering, with the two long ones individually resolvable.
+
+Your hypothesis is supported. The flux corpus's short documents (median 826, p90 2,721) were a genuine ceiling on long-context retrieval, not just the YaRN cache. Training on documents that actually span the window recovers a large chunk: at the full 16,384 window, 33% → 48%, a +45% relative improvement from changing nothing but document length.
+
+Backward retrieval is untouched — val_backward sits at 0-2% everywhere against ~2.9% chance, identical across all three arms. Document length does nothing for value→key retrieval. That failure mode is independent of both the cache fix and the data fix.
+
+Two caveats that matter for how you read this
+
+The control's absolute numbers moved. armA_c scored 25% at 16k on the flux haystack; here it's 33%. Same checkpoint — the haystack changed (b3's 16-32k documents, held out from all arms). That's exactly why I re-scored the control rather than reusing its published numbers; all three columns above share one haystack, so they compare, but they don't compare to the 2026-09-03 table.
+
+Training losses do not compare across arms (l1 2.459, l2 2.794). Each is measured on its own distribution, and predicting inside 8-16k documents is a harder problem than inside 4-8k ones. l2 has the worst loss and the best retrieval — the same divergence as the cache experiment. Only the retrieval numbers compare.
+
+Results are in lenbucket_logs/RESULTS.md.

logs/lenbucket/eval_1024.log

@@ -0,0 +1,162 @@
+[08:17:23] longeval: T=16384 prompt_len=1024 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[08:17:23] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[08:17:23] prompt signature changed -- regenerating prompts and clearing state
+[08:17:23] generating prompts: niah_single
+[08:17:28] generating prompts: val_backward
+[08:17:28] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[08:17:28] 6 job(s) to run
+[08:17:28]   [gpu 0] armA_c / niah_single
+[08:17:28]   [gpu 1] armA_c / val_backward
+[08:17:28]   [gpu 2] l1 / niah_single
+[08:17:28]   [gpu 3] l1 / val_backward
+[08:17:28]   [gpu 0] l2 / niah_single
+[08:17:28]   [gpu 1] l2 / val_backward
+[08:20:56]   NO OUTPUT armA_c/niah_single -- see lenb_len_1024/gen/armA_c_niah_single.log
+[08:20:56]   NO OUTPUT armA_c/val_backward -- see lenb_len_1024/gen/armA_c_val_backward.log
+[08:20:56]   NO OUTPUT l1/niah_single -- see lenb_len_1024/gen/l1_niah_single.log
+[08:20:56]   NO OUTPUT l1/val_backward -- see lenb_len_1024/gen/l1_val_backward.log
+[08:20:56]   NO OUTPUT l2/niah_single -- see lenb_len_1024/gen/l2_niah_single.log
+[08:20:56]   NO OUTPUT l2/val_backward -- see lenb_len_1024/gen/l2_val_backward.log
+[08:20:56] building summary
+  no scores yet
+[08:20:56] DONE -> lenb_len_1024/summary.csv
+[08:56:22] longeval: T=16384 prompt_len=1024 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[08:56:22] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[08:56:22] prompt signature changed -- regenerating prompts and clearing state
+[08:56:22] generating prompts: niah_single
+[08:56:22] generating prompts: val_backward
+[08:56:23] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[08:56:23] 6 job(s) to run
+[08:56:23]   [gpu 0] armA_c / niah_single
+[08:56:23]   [gpu 1] armA_c / val_backward
+[08:56:23]   [gpu 2] l1 / niah_single
+[08:56:23]   [gpu 3] l1 / val_backward
+[08:56:23]   [gpu 0] l2 / niah_single
+[08:56:23]   [gpu 1] l2 / val_backward
+[09:01:58]   scoring armA_c / niah_single
+
+=== niah_single  (prompt_len=1024, n=100) ===
+  accuracy: 94/100 = 94.0%
+    depth      acc   (needle position in context)
+     0.00  17/20   85.0%
+     0.25  20/20  100.0%
+     0.50  19/20   95.0%
+     0.75  20/20  100.0%
+     1.00  18/20   90.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 8056020.\n
+  [HIT ] expect=5343902
+         got= 5343902.  shaped, pink physician raised his hand to his mouth in a gesture of silence to the awkward
+
+  wrote lenb_len_1024/scores/armA_c_niah_single.csv
+[09:01:58]   scoring armA_c / val_backward
+
+=== val_backward  (prompt_len=1024, n=100) ===
+  accuracy: 1/100 = 1.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   1/20    5.0%
+     1.00   0/20    0.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the one that has the special magic number 8056020.\n
+  [MISS] expect=cactus
+         got= the Zangoose.\n
+
+  wrote lenb_len_1024/scores/armA_c_val_backward.csv
+[09:01:58]   scoring l1 / niah_single
+
+=== niah_single  (prompt_len=1024, n=100) ===
+  accuracy: 95/100 = 95.0%
+    depth      acc   (needle position in context)
+     0.00  18/20   90.0%
+     0.25  20/20  100.0%
+     0.50  19/20   95.0%
+     0.75  20/20  100.0%
+     1.00  18/20   90.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 805602020.\n
+  [HIT ] expect=5343902
+         got= 5343902.  shaped, pink physician raised his hand to his mouth in a gesture of silence to the awkward
+
+  wrote lenb_len_1024/scores/l1_niah_single.csv
+[09:01:59]   scoring l1 / val_backward
+
+=== val_backward  (prompt_len=1024, n=100) ===
+  accuracy: 1/100 = 1.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   1/20    5.0%
+     1.00   0/20    0.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the item with the special magic number 8056020.\n
+  [MISS] expect=cactus
+         got= the Zangoose.\n
+
+  wrote lenb_len_1024/scores/l1_val_backward.csv
+[09:01:59]   scoring l2 / niah_single
+
+=== niah_single  (prompt_len=1024, n=100) ===
+  accuracy: 96/100 = 96.0%
+    depth      acc   (needle position in context)
+     0.00  19/20   95.0%
+     0.25  19/20   95.0%
+     0.50  20/20  100.0%
+     0.75  19/20   95.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 805602020.\n
+  [HIT ] expect=5343902
+         got= 5343902.  shaped, pink physician raised his hand to his mouth in a gesture of silence to the awkward
+
+  wrote lenb_len_1024/scores/l2_niah_single.csv
+[09:01:59]   scoring l2 / val_backward
+
+=== val_backward  (prompt_len=1024, n=100) ===
+  accuracy: 1/100 = 1.0%
+    depth      acc   (needle position in context)
+     0.00   1/20    5.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   0/20    0.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the item with the special magic number 8056020.\n
+  [MISS] expect=cactus
+         got= the Zangoose.\n
+
+  wrote lenb_len_1024/scores/l2_val_backward.csv
+[09:01:59] building summary
+
+=== per-task accuracy %  (secondary/descriptive: ~10 per depth cell) ===
+  task                 armA_c       l1       l2
+  niah_single            94.0     95.0     96.0
+  val_backward            1.0      1.0      1.0
+
+  degenerate (<=1% for every model): ['val_backward']
+
+=== PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) ===
+  depth             armA_c              l1              l2
+  0.00          85.0+-15.4      90.0+-13.7      95.0+-11.4
+  0.25         100.0+- 8.1     100.0+- 8.1      95.0+-11.4
+  0.50          95.0+-11.4      95.0+-11.4     100.0+- 8.1
+  0.75         100.0+- 8.1     100.0+- 8.1      95.0+-11.4
+  1.00          90.0+-13.7      90.0+-13.7      95.0+-11.4
+
+  wrote lenb_len_1024/summary.csv and lenb_len_1024/depth_summary.csv
+[09:01:59] DONE -> lenb_len_1024/summary.csv

logs/lenbucket/eval_16360.log

@@ -0,0 +1,162 @@
+[08:31:22] longeval: T=16384 prompt_len=16360 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[08:31:22] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[08:31:22] prompt signature changed -- regenerating prompts and clearing state
+[08:31:22] generating prompts: niah_single
+[08:31:22] generating prompts: val_backward
+[08:31:23] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[08:31:23] 6 job(s) to run
+[08:31:23]   [gpu 0] armA_c / niah_single
+[08:31:23]   [gpu 1] armA_c / val_backward
+[08:31:23]   [gpu 2] l1 / niah_single
+[08:31:23]   [gpu 3] l1 / val_backward
+[08:31:23]   [gpu 0] l2 / niah_single
+[08:31:23]   [gpu 1] l2 / val_backward
+[08:34:50]   NO OUTPUT armA_c/niah_single -- see lenb_len_16360/gen/armA_c_niah_single.log
+[08:34:50]   NO OUTPUT armA_c/val_backward -- see lenb_len_16360/gen/armA_c_val_backward.log
+[08:34:50]   NO OUTPUT l1/niah_single -- see lenb_len_16360/gen/l1_niah_single.log
+[08:34:50]   NO OUTPUT l1/val_backward -- see lenb_len_16360/gen/l1_val_backward.log
+[08:34:50]   NO OUTPUT l2/niah_single -- see lenb_len_16360/gen/l2_niah_single.log
+[08:34:50]   NO OUTPUT l2/val_backward -- see lenb_len_16360/gen/l2_val_backward.log
+[08:34:50] building summary
+  no scores yet
+[08:34:50] DONE -> lenb_len_16360/summary.csv
+[09:27:23] longeval: T=16384 prompt_len=16360 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[09:27:23] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[09:27:23] prompt signature changed -- regenerating prompts and clearing state
+[09:27:23] generating prompts: niah_single
+[09:27:24] generating prompts: val_backward
+[09:27:24] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[09:27:24] 6 job(s) to run
+[09:27:24]   [gpu 0] armA_c / niah_single
+[09:27:24]   [gpu 1] armA_c / val_backward
+[09:27:24]   [gpu 2] l1 / niah_single
+[09:27:24]   [gpu 3] l1 / val_backward
+[09:27:24]   [gpu 0] l2 / niah_single
+[09:27:24]   [gpu 1] l2 / val_backward
+[09:45:18]   scoring armA_c / niah_single
+
+=== niah_single  (prompt_len=16360, n=100) ===
+  accuracy: 33/100 = 33.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   4/20   20.0%
+     0.75  11/20   55.0%
+     1.00  18/20   90.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=8056020
+         got= the number of the Ghost-Type Pokemon that has been summoned to the Zangoose's lair.\n
+  [MISS] expect=5343902
+         got= the number of the “original sin” of the so-called primitive accumulation, which is the so-called
+
+  wrote lenb_len_16360/scores/armA_c_niah_single.csv
+[09:45:18]   scoring armA_c / val_backward
+
+=== val_backward  (prompt_len=16360, n=100) ===
+  accuracy: 1/100 = 1.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   1/20    5.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Ghost Dog Pokemon.\n
+  [MISS] expect=cactus
+         got= the “West”.\n
+
+  wrote lenb_len_16360/scores/armA_c_val_backward.csv
+[09:45:18]   scoring l1 / niah_single
+
+=== niah_single  (prompt_len=16360, n=100) ===
+  accuracy: 39/100 = 39.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   6/20   30.0%
+     0.75  15/20   75.0%
+     1.00  18/20   90.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=8056020
+         got= the number of the Ghost Dog Pokemon that can be summoned to the Zangoose's lair.\n
+  [MISS] expect=5343902
+         got= the number of the “centripetal” or “internal” elements of the “West
+
+  wrote lenb_len_16360/scores/l1_niah_single.csv
+[09:45:19]   scoring l1 / val_backward
+
+=== val_backward  (prompt_len=16360, n=100) ===
+  accuracy: 1/100 = 1.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   1/20    5.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Ghost Dog Pokemon.\n
+  [MISS] expect=cactus
+         got= the “exact number” of the “exact number” of the “exact number
+
+  wrote lenb_len_16360/scores/l1_val_backward.csv
+[09:45:19]   scoring l2 / niah_single
+
+=== niah_single  (prompt_len=16360, n=100) ===
+  accuracy: 48/100 = 48.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   6/20   30.0%
+     0.50  10/20   50.0%
+     0.75  14/20   70.0%
+     1.00  18/20   90.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=8056020
+         got= the number of the Zangoose's soul.\n
+  [MISS] expect=5343902
+         got= the number of the “West” and the “Rest” in the world.\n
+
+  wrote lenb_len_16360/scores/l2_niah_single.csv
+[09:45:19]   scoring l2 / val_backward
+
+=== val_backward  (prompt_len=16360, n=100) ===
+  accuracy: 0/100 = 0.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   0/20    0.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Seviper's treasure bag.\n
+  [MISS] expect=cactus
+         got= the “national question”.\n
+
+  wrote lenb_len_16360/scores/l2_val_backward.csv
+[09:45:19] building summary
+
+=== per-task accuracy %  (secondary/descriptive: ~10 per depth cell) ===
+  task                 armA_c       l1       l2
+  niah_single            33.0     39.0     48.0
+  val_backward            1.0      1.0      0.0
+
+  degenerate (<=1% for every model): ['val_backward']
+
+=== PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) ===
+  depth             armA_c              l1              l2
+  0.00           0.0+- 8.1       0.0+- 8.1       0.0+- 8.1
+  0.25           0.0+- 8.1       0.0+- 8.1      30.0+-18.7
+  0.50          20.0+-16.8      30.0+-18.7      50.0+-20.1
+  0.75          55.0+-20.0      75.0+-17.8      70.0+-18.7
+  1.00          90.0+-13.7      90.0+-13.7      90.0+-13.7
+
+  wrote lenb_len_16360/summary.csv and lenb_len_16360/depth_summary.csv
+[09:45:19] DONE -> lenb_len_16360/summary.csv

logs/lenbucket/eval_2048.log

@@ -0,0 +1,160 @@
+[08:20:56] longeval: T=16384 prompt_len=2048 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[08:20:56] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[08:20:56] prompt signature changed -- regenerating prompts and clearing state
+[08:20:56] generating prompts: niah_single
+[08:20:57] generating prompts: val_backward
+[08:20:57] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[08:20:57] 6 job(s) to run
+[08:20:57]   [gpu 0] armA_c / niah_single
+[08:20:57]   [gpu 1] armA_c / val_backward
+[08:20:57]   [gpu 2] l1 / niah_single
+[08:20:57]   [gpu 3] l1 / val_backward
+[08:20:57]   [gpu 0] l2 / niah_single
+[08:20:57]   [gpu 1] l2 / val_backward
+[08:24:24]   NO OUTPUT armA_c/niah_single -- see lenb_len_2048/gen/armA_c_niah_single.log
+[08:24:24]   NO OUTPUT armA_c/val_backward -- see lenb_len_2048/gen/armA_c_val_backward.log
+[08:24:24]   NO OUTPUT l1/niah_single -- see lenb_len_2048/gen/l1_niah_single.log
+[08:24:24]   NO OUTPUT l1/val_backward -- see lenb_len_2048/gen/l1_val_backward.log
+[08:24:24]   NO OUTPUT l2/niah_single -- see lenb_len_2048/gen/l2_niah_single.log
+[08:24:24]   NO OUTPUT l2/val_backward -- see lenb_len_2048/gen/l2_val_backward.log
+[08:24:24] building summary
+  no scores yet
+[08:24:24] DONE -> lenb_len_2048/summary.csv
+[09:01:59] longeval: T=16384 prompt_len=2048 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[09:01:59] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[09:01:59] prompt signature changed -- regenerating prompts and clearing state
+[09:01:59] generating prompts: niah_single
+[09:02:00] generating prompts: val_backward
+[09:02:00] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[09:02:00] 6 job(s) to run
+[09:02:00]   [gpu 0] armA_c / niah_single
+[09:02:00]   [gpu 1] armA_c / val_backward
+[09:02:00]   [gpu 2] l1 / niah_single
+[09:02:00]   [gpu 3] l1 / val_backward
+[09:02:00]   [gpu 0] l2 / niah_single
+[09:02:00]   [gpu 1] l2 / val_backward
+[09:08:17]   scoring armA_c / niah_single
+
+=== niah_single  (prompt_len=2048, n=100) ===
+  accuracy: 94/100 = 94.0%
+    depth      acc   (needle position in context)
+     0.00  18/20   90.0%
+     0.25  18/20   90.0%
+     0.50  20/20  100.0%
+     0.75  19/20   95.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 80560202020202020202020202020202020202020202020
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_2048/scores/armA_c_niah_single.csv
+[09:08:17]   scoring armA_c / val_backward
+
+=== val_backward  (prompt_len=2048, n=100) ===
+  accuracy: 2/100 = 2.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   1/20    5.0%
+     0.50   1/20    5.0%
+     0.75   0/20    0.0%
+     1.00   0/20    0.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Zangoose.\n
+  [MISS] expect=cactus
+         got= the Zangoose.\n
+
+  wrote lenb_len_2048/scores/armA_c_val_backward.csv
+[09:08:17]   scoring l1 / niah_single
+
+=== niah_single  (prompt_len=2048, n=100) ===
+  accuracy: 96/100 = 96.0%
+    depth      acc   (needle position in context)
+     0.00  19/20   95.0%
+     0.25  18/20   90.0%
+     0.50  20/20  100.0%
+     0.75  20/20  100.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 80560202020202020202020202020202020202020202020
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_2048/scores/l1_niah_single.csv
+[09:08:18]   scoring l1 / val_backward
+
+=== val_backward  (prompt_len=2048, n=100) ===
+  accuracy: 2/100 = 2.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   1/20    5.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   1/20    5.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the one that has the special magic number 8056020.\n
+  [MISS] expect=cactus
+         got= the Zangoose.\n
+
+  wrote lenb_len_2048/scores/l1_val_backward.csv
+[09:08:18]   scoring l2 / niah_single
+
+=== niah_single  (prompt_len=2048, n=100) ===
+  accuracy: 94/100 = 94.0%
+    depth      acc   (needle position in context)
+     0.00  20/20  100.0%
+     0.25  18/20   90.0%
+     0.50  19/20   95.0%
+     0.75  18/20   90.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 80560202020202020202020202020202020202020202020
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_2048/scores/l2_niah_single.csv
+[09:08:18]   scoring l2 / val_backward
+
+=== val_backward  (prompt_len=2048, n=100) ===
+  accuracy: 1/100 = 1.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   1/20    5.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Zangoose.\n
+  [MISS] expect=cactus
+         got= the Zangoose.\n
+
+  wrote lenb_len_2048/scores/l2_val_backward.csv
+[09:08:18] building summary
+
+=== per-task accuracy %  (secondary/descriptive: ~10 per depth cell) ===
+  task                 armA_c       l1       l2
+  niah_single            94.0     96.0     94.0
+  val_backward            2.0      2.0      1.0
+
+=== PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) ===
+  depth             armA_c              l1              l2
+  0.00          45.0+-14.7      47.5+-14.8      50.0+-14.8
+  0.25          47.5+-14.8      47.5+-14.8      45.0+-14.7
+  0.50          52.5+-14.8      50.0+-14.8      47.5+-14.8
+  0.75          47.5+-14.8      50.0+-14.8      45.0+-14.7
+  1.00          47.5+-14.8      50.0+-14.8      50.0+-14.8
+
+  wrote lenb_len_2048/summary.csv and lenb_len_2048/depth_summary.csv
+[09:08:18] DONE -> lenb_len_2048/summary.csv

logs/lenbucket/eval_4096.log

@@ -0,0 +1,162 @@
+[08:24:24] longeval: T=16384 prompt_len=4096 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[08:24:24] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[08:24:24] prompt signature changed -- regenerating prompts and clearing state
+[08:24:24] generating prompts: niah_single
+[08:24:25] generating prompts: val_backward
+[08:24:25] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[08:24:25] 6 job(s) to run
+[08:24:25]   [gpu 0] armA_c / niah_single
+[08:24:25]   [gpu 1] armA_c / val_backward
+[08:24:25]   [gpu 2] l1 / niah_single
+[08:24:25]   [gpu 3] l1 / val_backward
+[08:24:25]   [gpu 0] l2 / niah_single
+[08:24:25]   [gpu 1] l2 / val_backward
+[08:27:53]   NO OUTPUT armA_c/niah_single -- see lenb_len_4096/gen/armA_c_niah_single.log
+[08:27:53]   NO OUTPUT armA_c/val_backward -- see lenb_len_4096/gen/armA_c_val_backward.log
+[08:27:53]   NO OUTPUT l1/niah_single -- see lenb_len_4096/gen/l1_niah_single.log
+[08:27:53]   NO OUTPUT l1/val_backward -- see lenb_len_4096/gen/l1_val_backward.log
+[08:27:53]   NO OUTPUT l2/niah_single -- see lenb_len_4096/gen/l2_niah_single.log
+[08:27:53]   NO OUTPUT l2/val_backward -- see lenb_len_4096/gen/l2_val_backward.log
+[08:27:53] building summary
+  no scores yet
+[08:27:53] DONE -> lenb_len_4096/summary.csv
+[09:08:18] longeval: T=16384 prompt_len=4096 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[09:08:18] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[09:08:18] prompt signature changed -- regenerating prompts and clearing state
+[09:08:18] generating prompts: niah_single
+[09:08:19] generating prompts: val_backward
+[09:08:19] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[09:08:19] 6 job(s) to run
+[09:08:19]   [gpu 0] armA_c / niah_single
+[09:08:19]   [gpu 1] armA_c / val_backward
+[09:08:19]   [gpu 2] l1 / niah_single
+[09:08:19]   [gpu 3] l1 / val_backward
+[09:08:19]   [gpu 0] l2 / niah_single
+[09:08:19]   [gpu 1] l2 / val_backward
+[09:16:07]   scoring armA_c / niah_single
+
+=== niah_single  (prompt_len=4096, n=100) ===
+  accuracy: 84/100 = 84.0%
+    depth      acc   (needle position in context)
+     0.00  15/20   75.0%
+     0.25  17/20   85.0%
+     0.50  17/20   85.0%
+     0.75  17/20   85.0%
+     1.00  18/20   90.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 8056020.\n
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_4096/scores/armA_c_niah_single.csv
+[09:16:07]   scoring armA_c / val_backward
+
+=== val_backward  (prompt_len=4096, n=100) ===
+  accuracy: 0/100 = 0.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   0/20    0.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Zangoose.\n
+  [MISS] expect=cactus
+         got= the Zangoose.\n
+
+  wrote lenb_len_4096/scores/armA_c_val_backward.csv
+[09:16:08]   scoring l1 / niah_single
+
+=== niah_single  (prompt_len=4096, n=100) ===
+  accuracy: 88/100 = 88.0%
+    depth      acc   (needle position in context)
+     0.00  14/20   70.0%
+     0.25  20/20  100.0%
+     0.50  18/20   90.0%
+     0.75  17/20   85.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 80560202020202020202020202020202020202020202020
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_4096/scores/l1_niah_single.csv
+[09:16:08]   scoring l1 / val_backward
+
+=== val_backward  (prompt_len=4096, n=100) ===
+  accuracy: 0/100 = 0.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   0/20    0.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Zangoose.\n
+  [MISS] expect=cactus
+         got= the item with the special magic number 5343902.\n
+
+  wrote lenb_len_4096/scores/l1_val_backward.csv
+[09:16:08]   scoring l2 / niah_single
+
+=== niah_single  (prompt_len=4096, n=100) ===
+  accuracy: 92/100 = 92.0%
+    depth      acc   (needle position in context)
+     0.00  18/20   90.0%
+     0.25  19/20   95.0%
+     0.50  18/20   90.0%
+     0.75  18/20   90.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 80560202020202020202020202020202020202020202020
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_4096/scores/l2_niah_single.csv
+[09:16:08]   scoring l2 / val_backward
+
+=== val_backward  (prompt_len=4096, n=100) ===
+  accuracy: 0/100 = 0.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   0/20    0.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Zangoose.\n
+  [MISS] expect=cactus
+         got= the Seviper.\n
+
+  wrote lenb_len_4096/scores/l2_val_backward.csv
+[09:16:08] building summary
+
+=== per-task accuracy %  (secondary/descriptive: ~10 per depth cell) ===
+  task                 armA_c       l1       l2
+  niah_single            84.0     88.0     92.0
+  val_backward            0.0      0.0      0.0
+
+  degenerate (<=1% for every model): ['val_backward']
+
+=== PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) ===
+  depth             armA_c              l1              l2
+  0.00          75.0+-17.8      70.0+-18.7      90.0+-13.7
+  0.25          85.0+-15.4     100.0+- 8.1      95.0+-11.4
+  0.50          85.0+-15.4      90.0+-13.7      90.0+-13.7
+  0.75          85.0+-15.4      85.0+-15.4      90.0+-13.7
+  1.00          90.0+-13.7      95.0+-11.4      95.0+-11.4
+
+  wrote lenb_len_4096/summary.csv and lenb_len_4096/depth_summary.csv
+[09:16:08] DONE -> lenb_len_4096/summary.csv

logs/lenbucket/eval_8192.log

@@ -0,0 +1,160 @@
+[08:27:53] longeval: T=16384 prompt_len=8192 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[08:27:53] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[08:27:53] prompt signature changed -- regenerating prompts and clearing state
+[08:27:53] generating prompts: niah_single
+[08:27:54] generating prompts: val_backward
+[08:27:54] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[08:27:54] 6 job(s) to run
+[08:27:54]   [gpu 0] armA_c / niah_single
+[08:27:54]   [gpu 1] armA_c / val_backward
+[08:27:54]   [gpu 2] l1 / niah_single
+[08:27:54]   [gpu 3] l1 / val_backward
+[08:27:54]   [gpu 0] l2 / niah_single
+[08:27:54]   [gpu 1] l2 / val_backward
+[08:31:22]   NO OUTPUT armA_c/niah_single -- see lenb_len_8192/gen/armA_c_niah_single.log
+[08:31:22]   NO OUTPUT armA_c/val_backward -- see lenb_len_8192/gen/armA_c_val_backward.log
+[08:31:22]   NO OUTPUT l1/niah_single -- see lenb_len_8192/gen/l1_niah_single.log
+[08:31:22]   NO OUTPUT l1/val_backward -- see lenb_len_8192/gen/l1_val_backward.log
+[08:31:22]   NO OUTPUT l2/niah_single -- see lenb_len_8192/gen/l2_niah_single.log
+[08:31:22]   NO OUTPUT l2/val_backward -- see lenb_len_8192/gen/l2_val_backward.log
+[08:31:22] building summary
+  no scores yet
+[08:31:22] DONE -> lenb_len_8192/summary.csv
+[09:16:09] longeval: T=16384 prompt_len=8192 max_new=24 n=100 greedy=1 gpus=0,1,2,3
+[09:16:09] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin
+[09:16:09] prompt signature changed -- regenerating prompts and clearing state
+[09:16:09] generating prompts: niah_single
+[09:16:09] generating prompts: val_backward
+[09:16:09] fan-out: 4 GPU(s), NP=1 -> 8 concurrent job(s)
+[09:16:09] 6 job(s) to run
+[09:16:09]   [gpu 0] armA_c / niah_single
+[09:16:09]   [gpu 1] armA_c / val_backward
+[09:16:09]   [gpu 2] l1 / niah_single
+[09:16:09]   [gpu 3] l1 / val_backward
+[09:16:09]   [gpu 0] l2 / niah_single
+[09:16:09]   [gpu 1] l2 / val_backward
+[09:27:22]   scoring armA_c / niah_single
+
+=== niah_single  (prompt_len=8192, n=100) ===
+  accuracy: 62/100 = 62.0%
+    depth      acc   (needle position in context)
+     0.00   3/20   15.0%
+     0.25  12/20   60.0%
+     0.50  14/20   70.0%
+     0.75  14/20   70.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 8056020.\n
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_8192/scores/armA_c_niah_single.csv
+[09:27:22]   scoring armA_c / val_backward
+
+=== val_backward  (prompt_len=8192, n=100) ===
+  accuracy: 2/100 = 2.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   2/20   10.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Seviper.\n
+  [MISS] expect=cactus
+         got= the Zangoose.\n
+
+  wrote lenb_len_8192/scores/armA_c_val_backward.csv
+[09:27:22]   scoring l1 / niah_single
+
+=== niah_single  (prompt_len=8192, n=100) ===
+  accuracy: 67/100 = 67.0%
+    depth      acc   (needle position in context)
+     0.00   4/20   20.0%
+     0.25  12/20   60.0%
+     0.50  15/20   75.0%
+     0.75  17/20   85.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 8056020.\n
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_8192/scores/l1_niah_single.csv
+[09:27:23]   scoring l1 / val_backward
+
+=== val_backward  (prompt_len=8192, n=100) ===
+  accuracy: 1/100 = 1.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   0/20    0.0%
+     1.00   1/20    5.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the item with the special magic number 8056020.\n
+  [MISS] expect=cactus
+         got= the Ghost Dog Pokemon.\n
+
+  wrote lenb_len_8192/scores/l1_val_backward.csv
+[09:27:23]   scoring l2 / niah_single
+
+=== niah_single  (prompt_len=8192, n=100) ===
+  accuracy: 78/100 = 78.0%
+    depth      acc   (needle position in context)
+     0.00  14/20   70.0%
+     0.25  14/20   70.0%
+     0.50  14/20   70.0%
+     0.75  17/20   85.0%
+     1.00  19/20   95.0%
+
+  --- 2 sample generations ---
+  [HIT ] expect=8056020
+         got= 805602020.\n
+  [HIT ] expect=5343902
+         got= 5343902.\n
+
+  wrote lenb_len_8192/scores/l2_niah_single.csv
+[09:27:23]   scoring l2 / val_backward
+
+=== val_backward  (prompt_len=8192, n=100) ===
+  accuracy: 2/100 = 2.0%
+    depth      acc   (needle position in context)
+     0.00   0/20    0.0%
+     0.25   0/20    0.0%
+     0.50   0/20    0.0%
+     0.75   1/20    5.0%
+     1.00   1/20    5.0%
+
+  --- 2 sample generations ---
+  [MISS] expect=yarrow
+         got= the Seviper's special magic number.\n
+  [MISS] expect=cactus
+         got= the Seviper's treasure bag.\n
+
+  wrote lenb_len_8192/scores/l2_val_backward.csv
+[09:27:23] building summary
+
+=== per-task accuracy %  (secondary/descriptive: ~10 per depth cell) ===
+  task                 armA_c       l1       l2
+  niah_single            62.0     67.0     78.0
+  val_backward            2.0      1.0      2.0
+
+=== PRIMARY: accuracy % pooled by depth across tasks (+-95% Wilson) ===
+  depth             armA_c              l1              l2
+  0.00           7.5+- 8.6      10.0+- 9.5      35.0+-14.2
+  0.25          30.0+-13.7      30.0+-13.7      35.0+-14.2
+  0.50          35.0+-14.2      37.5+-14.4      35.0+-14.2
+  0.75          35.0+-14.2      42.5+-14.6      45.0+-14.7
+  1.00          52.5+-14.8      50.0+-14.8      50.0+-14.8
+
+  wrote lenb_len_8192/summary.csv and lenb_len_8192/depth_summary.csv
+[09:27:23] DONE -> lenb_len_8192/summary.csv

logs/lenbucket/l1.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:62fb62744e37bed667bd11a6860c34eebe887e4d9070514cad9b7e33fb7d4ae7
+size 61

logs/lenbucket/l2.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6522f5bd1f3ca4c0bb03917c17873cf40717741c0bcb3b641e48bedb3bb8587a
+size 61

logs/lenbucket/master.log

@@ -0,0 +1,39 @@
+[06:41:49] === length-bucket experiment ===
+[06:41:49] free disk 113G
+[06:41:49] l1: 4 train shards, 1 val shards in lendata/b1_4096_8192
+[06:41:49] l1: attempting CP_B=2 (grad_accum=16, global 524288)
+[07:29:38] l1 -> run 1, blumodelcp_run1_step_400.ckpt, B=2, final loss loss: 2.458802
+[07:29:38] l1  peak mem mem: 68769MB
+[07:29:38] l2: 4 train shards, 1 val shards in lendata/b2_8192_16384
+[07:29:38] l2: attempting CP_B=2 (grad_accum=16, global 524288)
+[08:17:23] l2 -> run 2, blumodelcp_run2_step_400.ckpt, B=2, final loss loss: 2.794149
+[08:17:23] l2  peak mem mem: 68769MB
+[08:17:23] eval models: armA_c l1 l2 
+[08:17:23] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin (195370858 B)
+[08:17:23] eval PROMPT_LEN=1024 N=100
+[08:20:56] eval PROMPT_LEN=2048 N=100
+[08:24:24] eval PROMPT_LEN=4096 N=100
+[08:27:53] eval PROMPT_LEN=8192 N=100
+[08:31:22] eval PROMPT_LEN=16360 N=100
+[08:34:50] eval done
+[08:34:50] wrote lenbucket_logs/RESULTS.md
+[08:34:50] === LENBUCKET DONE ===
+[08:56:22] === length-bucket experiment ===
+[08:56:22] free disk 70G
+[08:56:22] l1 already trained -> checkpoints_bluscriptcp/rank_0/blumodelcp_run1_step_400.ckpt
+[08:56:22] l2 already trained -> checkpoints_bluscriptcp/rank_0/blumodelcp_run2_step_400.ckpt
+[08:56:22] eval models: armA_c l1 l2 
+[08:56:22] haystack: /root/Adhi/BluTrain/dist/Context_Parallelism/lendata/haystack/b3val_haystack.bin (195370858 B)
+[08:56:22] eval PROMPT_LEN=1024 N=100
+[09:01:59]   -> niah_single,94.0,95.0,96.0
+[09:01:59] eval PROMPT_LEN=2048 N=100
+[09:08:18]   -> niah_single,94.0,96.0,94.0
+[09:08:18] eval PROMPT_LEN=4096 N=100
+[09:16:08]   -> niah_single,84.0,88.0,92.0
+[09:16:08] eval PROMPT_LEN=8192 N=100
+[09:27:23]   -> niah_single,62.0,67.0,78.0
+[09:27:23] eval PROMPT_LEN=16360 N=100
+[09:45:19]   -> niah_single,33.0,39.0,48.0
+[09:45:19] eval done
+[09:45:19] wrote lenbucket_logs/RESULTS.md
+[09:45:19] === LENBUCKET DONE ===

logs/lenbucket/upload.log

@@ -0,0 +1,22 @@
+url=https://huggingface.co/unparallelled/BluScriptCP/commit/43d5bfa9963b524fdf46f5e1309e3033708e503c
+url=https://huggingface.co/unparallelled/BluScriptCP/commit/fde21af468d74468a4acb06b19326a46fcd75caa
+Found 2 files to upload
+Uploading... 2/2 files checked, 0/0 uploaded (0.00B transferred), 0 committed in 0 commit(s)
+url=https://huggingface.co/unparallelled/BluScriptCP/commit/a1487708468da197733ba5487f809f9cdf52e850
+Found 38 files to upload
+Uploading... 38/38 files checked, 0/0 uploaded (0.00B transferred), 0 committed in 0 commit(s)
+url=https://huggingface.co/unparallelled/BluScriptCP/commit/3163896796f2e098e456e951aa35f8e3383d0864
+Found 38 files to upload
+Uploading... 38/38 files checked, 0/2 uploaded (0.00B transferred), 0 committed in 0 commit(s)
+url=https://huggingface.co/unparallelled/BluScriptCP/commit/690600acec45e3fa121a74b43b5a05a9c599a3be
+Found 38 files to upload
+Uploading... 38/38 files checked, 0/2 uploaded (0.00B transferred), 0 committed in 0 commit(s)
+url=https://huggingface.co/unparallelled/BluScriptCP/commit/c4b7d584860724cd04c3153b9d0056e336f77e4e
+Found 38 files to upload
+Uploading... 38/38 files checked, 0/0 uploaded (0.00B transferred), 0 committed in 0 commit(s)
+url=https://huggingface.co/unparallelled/BluScriptCP/commit/b77c08fda2c778fe60ab027add701e35e05e6e72
+Found 38 files to upload
+Uploading... 38/38 files checked, 0/2 uploaded (0.00B transferred), 0 committed in 0 commit(s)
+url=https://huggingface.co/unparallelled/BluScriptCP/commit/c7ed28914b1b08a9239691a218c7c81bbbde0cb4
+Found 12 files to upload
+Uploading... 12/12 files checked, 0/0 uploaded (0.00B transferred), 0 committed in 0 commit(s)