Upload phase5b_llama2/code_static_r04_p20_s2/crosseval.log with huggingface_hub

phase5b_llama2/code_static_r04_p20_s2/crosseval.log

@@ -0,0 +1,37 @@
+2026-05-22 22:09:44.579731: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
+2026-05-22 22:09:44.598504: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered
+WARNING: All log messages before absl::InitializeLog() is called are written to STDERR
+E0000 00:00:1779487784.614790 3031383 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
+E0000 00:00:1779487784.619974 3031383 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
+W0000 00:00:1779487784.633611 3031383 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
+W0000 00:00:1779487784.633638 3031383 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
+W0000 00:00:1779487784.633642 3031383 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
+W0000 00:00:1779487784.633644 3031383 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
+2026-05-22 22:09:44.637140: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
+To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
+AttributeError: 'MessageFactory' object has no attribute 'GetPrototype'
+AttributeError: 'MessageFactory' object has no attribute 'GetPrototype'
+AttributeError: 'MessageFactory' object has no attribute 'GetPrototype'
+AttributeError: 'MessageFactory' object has no attribute 'GetPrototype'
+AttributeError: 'MessageFactory' object has no attribute 'GetPrototype'
+
+=== code_static_r04_p20_s2 (outputs/phase5b_llama2/code_static_r04_p20_s2/v8_hardened_step1000.pt) ===
+[2026-05-22 22:09:46,585] [INFO] [real_accelerator.py:254:get_accelerator] Setting ds_accelerator to cuda (auto detect)
+[2026-05-22 22:09:47,555] [INFO] [logging.py:107:log_dist] [Rank -1] [TorchCheckpointEngine] Initialized with serialization = False
+
+  ckpt val_loss=0.8700  r_sum=329120  r_min=1228  r_max=1791
+  p_v42    (64 windows): loss=3.4193
+  p_v1     (64 windows): loss=3.4177
+  p_wt2v   (299964 tokens): loss=3.3899  PPL=29.66
+  p_wt2t   (341469 tokens): loss=3.3383  PPL=28.17
+
+=== SUMMARY TABLE (cross-entropy, nats/token; test also shown as PPL) ===
+run                         p_v42       p_v1     p_wt2v     p_wt2t   test_ppl
+code_static_r04_p20_s2     3.4193     3.4177     3.3899     3.3383      28.17
+
+=== DIAGNOSTIC READING ===
+  If seed=42 run wins p_v42 but loses p_v1 (and vice versa)  -> val-window overfit (kill target)
+  If all J5+J1 beat static on p_wt2v but not p_wt2t         -> WT2 val/test gap, need multi-seed
+  If J5+J1 both lose to static on p_wt2v                    -> dynamic allocation not generalizable at this recipe
+
+Saved to models/dynamic-rank-allocation/phase5b_llama2/code_static_r04_p20_s2_crosseval.json