Upload eval_wt_vs_kd_360bp_kdrl_20260505/run.log with huggingface_hub

eval_wt_vs_kd_360bp_kdrl_20260505/run.log

@@ -0,0 +1,30 @@
+[transformers] You are using a model of type `ntv3_posttrained` to instantiate a model of type `ntv3_conditioned`. This may be expected if you are loading a checkpoint that shares a subset of the architecture (e.g., loading a `sam2_video` checkpoint into `Sam2Model`), but is otherwise not supported and can yield errors. Please verify that the checkpoint is compatible with the model you are instantiating.
+
+[transformers] NTv3Generative LOAD REPORT from: /workspace/dnathinker/ntv3_local/100m_post
+Key                                                                               | Status     |  | 
+----------------------------------------------------------------------------------+------------+--+-
+core.bigwig_head.species_heads.{5, 6, 7, 8, 13, 16, 18, 21, 23}.head.bias         | UNEXPECTED |  | 
+core.bigwig_head.species_heads.{5, 6, 7, 8, 13, 16, 18, 21, 23}.layer_norm.weight | UNEXPECTED |  | 
+core.bigwig_head.species_heads.{5, 6, 7, 8, 13, 16, 18, 21, 23}.head.weight       | UNEXPECTED |  | 
+core.bigwig_head.species_heads.{5, 6, 7, 8, 13, 16, 18, 21, 23}.layer_norm.bias   | UNEXPECTED |  | 
+core.bed_head.layer_norm.bias                                                     | UNEXPECTED |  | 
+core.bed_head.head.bias                                                           | UNEXPECTED |  | 
+core.bed_head.layer_norm.weight                                                   | UNEXPECTED |  | 
+core.bed_head.head.weight                                                         | UNEXPECTED |  | 
+
+Notes:
+- UNEXPECTED:	can be ignored when loading from different task/architecture; not ok if you expect identical arch.
+[transformers] You are using a model of type `ntv3_posttrained` to instantiate a model of type `ntv3_conditioned`. This may be expected if you are loading a checkpoint that shares a subset of the architecture (e.g., loading a `sam2_video` checkpoint into `Sam2Model`), but is otherwise not supported and can yield errors. Please verify that the checkpoint is compatible with the model you are instantiating.
+[wt-vs-kd] trunk loaded: missing=0 unexpected=12
+[wt-vs-kd] RL-LoRA: attached + loaded 48 tensors
+[t3-rl] [§B] loaded KD tables: [('Ex', 20), ('In', 20), ('OPC', 20), ('Ast', 20), ('Oli', 20), ('Mic', 20), ('End', 20)]
+[wt-vs-kd] test rows: 200
+  [wt-vs-kd] 25/200 elapsed=145.6s
+  [wt-vs-kd] 50/200 elapsed=291.0s
+  [wt-vs-kd] 75/200 elapsed=436.2s
+  [wt-vs-kd] 100/200 elapsed=581.1s
+  [wt-vs-kd] 125/200 elapsed=725.7s
+  [wt-vs-kd] 150/200 elapsed=870.5s
+  [wt-vs-kd] 175/200 elapsed=1015.1s
+  [wt-vs-kd] 200/200 elapsed=1160.0s
+[wt-vs-kd] mean_wt=0.450 mean_kd=0.215 ratio=0.478 kd<wt=34/200