add dense t-SNE scaling output scaleup7_tasks7_stage2.log

visualizations/h100/tsne_scaling_20260728_final_dense/logs/scaleup7_tasks7_stage2.log

@@ -0,0 +1,139 @@
+dataset samples 71191 selected 3500 tasks ['insert_HDMI', 'insert_hole', 'lift_bottle', 'pull_out_key', 'insert_tube', 'lift_can', 'put_bottle_in_shelf']
+
+[transformers] ViTModel LOAD REPORT from: /dev/shm/muse/src/ViTacDreamer_policy/.hf_cache/google_vit_base_patch16_224
+Key                 | Status     | 
+--------------------+------------+-
+classifier.bias     | UNEXPECTED | 
+classifier.weight   | UNEXPECTED | 
+pooler.dense.weight | MISSING    | 
+pooler.dense.bias   | MISSING    | 
+
+Notes:
+- UNEXPECTED:	can be ignored when loading from different task/architecture; not ok if you expect identical arch.
+- MISSING:	those params were newly initialized because missing from the checkpoint. Consider training on your downstream task.
+
+[transformers] ViTModel LOAD REPORT from: /dev/shm/muse/src/ViTacDreamer_policy/.hf_cache/google_vit_base_patch16_224
+Key                 | Status     | 
+--------------------+------------+-
+classifier.bias     | UNEXPECTED | 
+classifier.weight   | UNEXPECTED | 
+pooler.dense.weight | MISSING    | 
+pooler.dense.bias   | MISSING    | 
+
+Notes:
+- UNEXPECTED:	can be ignored when loading from different task/architecture; not ok if you expect identical arch.
+- MISSING:	those params were newly initialized because missing from the checkpoint. Consider training on your downstream task.
+/dev/shm/muse/src/ViTacDreamer_policy/scripts/analysis/generate_scaleup_tsne.py:82: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.
+  ckpt=torch.load(ckpt_path, map_location=device)
+load /dev/shm/muse/checkpoints/scaleup7_h32_20260726/stage2_prior512_encoder500_scaleup7_hlen5_stride5_e25_b192_from_stage1e30/stage2_v2_best.pth missing 0 unexpected 0
+batch 1 features 32
+batch 2 features 64
+batch 3 features 96
+batch 4 features 128
+batch 5 features 160
+batch 6 features 192
+batch 7 features 224
+batch 8 features 256
+batch 9 features 288
+batch 10 features 320
+batch 11 features 352
+batch 12 features 384
+batch 13 features 416
+batch 14 features 448
+batch 15 features 480
+batch 16 features 512
+batch 17 features 544
+batch 18 features 576
+batch 19 features 608
+batch 20 features 640
+batch 21 features 672
+batch 22 features 704
+batch 23 features 736
+batch 24 features 768
+batch 25 features 800
+batch 26 features 832
+batch 27 features 864
+batch 28 features 896
+batch 29 features 928
+batch 30 features 960
+batch 31 features 992
+batch 32 features 1024
+batch 33 features 1056
+batch 34 features 1088
+batch 35 features 1120
+batch 36 features 1152
+batch 37 features 1184
+batch 38 features 1216
+batch 39 features 1248
+batch 40 features 1280
+batch 41 features 1312
+batch 42 features 1344
+batch 43 features 1376
+batch 44 features 1408
+batch 45 features 1440
+batch 46 features 1472
+batch 47 features 1504
+batch 48 features 1536
+batch 49 features 1568
+batch 50 features 1600
+batch 51 features 1632
+batch 52 features 1664
+batch 53 features 1696
+batch 54 features 1728
+batch 55 features 1760
+batch 56 features 1792
+batch 57 features 1824
+batch 58 features 1856
+batch 59 features 1888
+batch 60 features 1920
+batch 61 features 1952
+batch 62 features 1984
+batch 63 features 2016
+batch 64 features 2048
+batch 65 features 2080
+batch 66 features 2112
+batch 67 features 2144
+batch 68 features 2176
+batch 69 features 2208
+batch 70 features 2240
+batch 71 features 2272
+batch 72 features 2304
+batch 73 features 2336
+batch 74 features 2368
+batch 75 features 2400
+batch 76 features 2432
+batch 77 features 2464
+batch 78 features 2496
+batch 79 features 2528
+batch 80 features 2560
+batch 81 features 2592
+batch 82 features 2624
+batch 83 features 2656
+batch 84 features 2688
+batch 85 features 2720
+batch 86 features 2752
+batch 87 features 2784
+batch 88 features 2816
+batch 89 features 2848
+batch 90 features 2880
+batch 91 features 2912
+batch 92 features 2944
+batch 93 features 2976
+batch 94 features 3008
+batch 95 features 3040
+batch 96 features 3072
+batch 97 features 3104
+batch 98 features 3136
+batch 99 features 3168
+batch 100 features 3200
+batch 101 features 3232
+batch 102 features 3264
+batch 103 features 3296
+batch 104 features 3328
+batch 105 features 3360
+batch 106 features 3392
+batch 107 features 3424
+batch 108 features 3456
+batch 109 features 3488
+batch 110 features 3500
+SAVED outputs/tsne_scaling_20260728_final_dense/scaleup7_tasks7_stage2_tsne.png