Hugging Face's logo

AbstractPhil
/
captionbert-8192-rotary

Feature Extraction
Transformers
TensorBoard
Safetensors
English
captionbert_structured_rotary
sentence-similarity
consensus-distillation
rotary-embeddings
custom_code
Model card Files
xet
 Metrics Community
AbstractPhil commited on
Commit
cf83c50
·
verified ·
1 Parent(s): 1f0837b

Preserve complete isolated phase-control console

Browse files
Files changed (1) hide show
  1. campaigns/structured-20261106/operations/phase-control-B/20261106T201750Z/console.log +57 -0
campaigns/structured-20261106/operations/phase-control-B/20261106T201750Z/console.log ADDED
@@ -0,0 +1,57 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ GPU memory ceiling 92.00 GB total; allocator allowance 90.61 GB
2
+ Frozen development: KALLP-seed2, 4,096 rows; FP32, no optimizer
3
+
4
+ Loading weights: 0%| | 0/158 [00:00<?, ?it/s]
5
+ Loading weights: 100%|██████████| 158/158 [00:00<00:00, 18292.99it/s]
6
+ [transformers] `use_return_dict` is deprecated! Use `return_dict` instead!
7
+ Lookup: 400/4,096 development captions
8
+ Lookup: 800/4,096 development captions
9
+ Lookup: 1,200/4,096 development captions
10
+ Lookup: 1,600/4,096 development captions
11
+ Lookup: 2,000/4,096 development captions
12
+ Lookup: 2,400/4,096 development captions
13
+ Lookup: 2,800/4,096 development captions
14
+ Lookup: 3,200/4,096 development captions
15
+ Lookup: 3,600/4,096 development captions
16
+ Lookup: 4,000/4,096 development captions
17
+ Validation heldout: 5,000 rows, batch 64
18
+ Validation heldout: 1,600/5,000 rows
19
+ Validation heldout: 3,200/5,000 rows
20
+ Validation heldout: 4,800/5,000 rows
21
+ Validation heldout: 5,000/5,000 rows
22
+ Validation heldout: computing cosine, effective rank and self-cosine...
23
+ Validation heldout: n=5,000 cosine=0.76683 erank=79.04200272569783 self_cos=0.004560547461659847
24
+ Validation >128 tokens: 5,000 rows, batch 64
25
+ Validation >128 tokens: 1,600/5,000 rows
26
+ Validation >128 tokens: 3,200/5,000 rows
27
+ Validation >128 tokens: 4,800/5,000 rows
28
+ Validation >128 tokens: 5,000/5,000 rows
29
+ Validation >128 tokens: computing cosine, effective rank and self-cosine...
30
+ Validation >128 tokens: n=5,000 cosine=0.70840 erank=60.24114986186824 self_cos=0.013372006044127109
31
+ Validation complete in 2.5s
32
+ Lookup: 400/4,096 development captions
33
+ Lookup: 800/4,096 development captions
34
+ Lookup: 1,200/4,096 development captions
35
+ Lookup: 1,600/4,096 development captions
36
+ Lookup: 2,000/4,096 development captions
37
+ Lookup: 2,400/4,096 development captions
38
+ Lookup: 2,800/4,096 development captions
39
+ Lookup: 3,200/4,096 development captions
40
+ Lookup: 3,600/4,096 development captions
41
+ Lookup: 4,000/4,096 development captions
42
+ Validation heldout: 5,000 rows, batch 64
43
+ Validation heldout: 1,600/5,000 rows
44
+ Validation heldout: 3,200/5,000 rows
45
+ Validation heldout: 4,800/5,000 rows
46
+ Validation heldout: 5,000/5,000 rows
47
+ Validation heldout: computing cosine, effective rank and self-cosine...
48
+ Validation heldout: n=5,000 cosine=0.76577 erank=79.33233652719497 self_cos=0.0035207391288226758
49
+ Validation >128 tokens: 5,000 rows, batch 64
50
+ Validation >128 tokens: 1,600/5,000 rows
51
+ Validation >128 tokens: 3,200/5,000 rows
52
+ Validation >128 tokens: 4,800/5,000 rows
53
+ Validation >128 tokens: 5,000/5,000 rows
54
+ Validation >128 tokens: computing cosine, effective rank and self-cosine...
55
+ Validation >128 tokens: n=5,000 cosine=0.70645 erank=60.70284102430367 self_cos=0.011925528330081241
56
+ Validation complete in 2.3s
57
+ {"arm": "KALLP", "seed": 2, "measurements": {"intact": {"all": {"queries": 10146, "captions": 1935, "record_mass": 0.9995156526565552, "caption_mean_record_mass": 0.9996639028076053, "median_record_mass": 0.9998456239700317, "exact_record_top1": 1.0, "wrong_adjacent_mass": 1.8013324734056368e-06, "uniform_mass": 0.013713696040213108}, "ending": {"queries": 9483, "captions": 1935, "record_mass": 0.99957674741745, "caption_mean_record_mass": 0.9996939892916716, "median_record_mass": 0.9998509883880615, "exact_record_top1": 1.0, "wrong_adjacent_mass": 1.8618209196574753e-06, "uniform_mass": 0.013700349256396294}, "non_ending": {"queries": 663, "captions": 299, "record_mass": 0.9986433982849121, "caption_mean_record_mass": 0.9988771832507589, "median_record_mass": 0.9996861219406128, "exact_record_top1": 1.0, "wrong_adjacent_mass": 9.361531851936888e-07, "uniform_mass": 0.013904616236686707}, "continuation": {"queries": 663, "captions": 299, "record_mass": 0.9995196461677551, "caption_mean_record_mass": 0.9995732722075089, "median_record_mass": 0.9998263716697693, "exact_record_top1": 1.0, "wrong_adjacent_mass": 4.174129344391986e-07, "uniform_mass": 0.013904616236686707}, "above128": {"queries": 3198, "captions": 252, "record_mass": 0.9992398619651794, "caption_mean_record_mass": 0.9992921014153768, "median_record_mass": 0.9997345209121704, "exact_record_top1": 1.0, "wrong_adjacent_mass": 1.5269587265720475e-06, "uniform_mass": 0.006118615623563528}}, "phase_constant": {"all": {"queries": 10146, "captions": 1935, "record_mass": 0.0965317040681839, "caption_mean_record_mass": 0.14569690794484252, "median_record_mass": 0.06158467009663582, "exact_record_top1": 0.34131677508377684, "wrong_adjacent_mass": 0.01376841589808464, "uniform_mass": 0.013713696040213108}, "ending": {"queries": 9483, "captions": 1935, "record_mass": 0.09836108237504959, "caption_mean_record_mass": 0.14723616505056158, "median_record_mass": 0.06314440071582794, "exact_record_top1": 0.3534746388273753, "wrong_adjacent_mass": 0.013675509952008724, "uniform_mass": 0.013700349256396294}, "non_ending": {"queries": 663, "captions": 299, "record_mass": 0.07036581635475159, "caption_mean_record_mass": 0.06514541684579023, "median_record_mass": 0.041065264493227005, "exact_record_top1": 0.167420814479638, "wrong_adjacent_mass": 0.015097277238965034, "uniform_mass": 0.013904616236686707}, "continuation": {"queries": 663, "captions": 299, "record_mass": 0.09113398939371109, "caption_mean_record_mass": 0.09189729233362702, "median_record_mass": 0.058898452669382095, "exact_record_top1": 0.2895927601809955, "wrong_adjacent_mass": 0.014761113561689854, "uniform_mass": 0.013904616236686707}, "above128": {"queries": 3198, "captions": 252, "record_mass": 0.0454973541200161, "caption_mean_record_mass": 0.0471360515396569, "median_record_mass": 0.035579562187194824, "exact_record_top1": 0.2761100687929956, "wrong_adjacent_mass": 0.005715395323932171, "uniform_mass": 0.006118615623563528}}}}