Upload baselines/asvd_ppl.log with huggingface_hub

baselines/asvd_ppl.log

@@ -0,0 +1,162 @@
+Loading model...
+`torch_dtype` is deprecated! Use `dtype` instead!
+2026-04-21 06:45:33.028642: 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`.
+2026-04-21 06:45:33.043469: 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 00:00:1776753933.059345     804 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered
+E0000 00:00:1776753933.064594     804 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered
+W0000 00:00:1776753933.078016     804 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
+W0000 00:00:1776753933.078038     804 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
+W0000 00:00:1776753933.078042     804 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
+W0000 00:00:1776753933.078044     804 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once.
+2026-04-21 06:45:33.081513: 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'
+
+Collecting covariances...
+  Calibration: 16/256 samples
+  Calibration: 32/256 samples
+  Calibration: 48/256 samples
+  Calibration: 64/256 samples
+  Calibration: 80/256 samples
+  Calibration: 96/256 samples
+  Calibration: 112/256 samples
+  Calibration: 128/256 samples
+  Calibration: 144/256 samples
+  Calibration: 160/256 samples
+  Calibration: 176/256 samples
+  Calibration: 192/256 samples
+  Calibration: 208/256 samples
+  Calibration: 224/256 samples
+  Calibration: 240/256 samples
+  Calibration: 256/256 samples
+
+============================================================
+ASVD PPL-mode at ratio=0.4
+[ASVD-PPL] Measuring sensitivity: 224 layers × 6 ratios
+  [1/224] self_attn.q_proj layer=0 pr=0.4 ppl=5.34
+  [1/224] self_attn.q_proj layer=0 pr=0.5 ppl=5.34
+  [1/224] self_attn.q_proj layer=0 pr=0.6 ppl=5.34
+  [1/224] self_attn.q_proj layer=0 pr=0.7 ppl=5.34
+  [1/224] self_attn.q_proj layer=0 pr=0.8 ppl=5.34
+  [1/224] self_attn.q_proj layer=0 pr=0.9 ppl=5.34
+  [2/224] self_attn.k_proj layer=0 pr=0.4 ppl=5.34
+  [2/224] self_attn.k_proj layer=0 pr=0.5 ppl=5.34
+  [2/224] self_attn.k_proj layer=0 pr=0.6 ppl=5.34
+  [2/224] self_attn.k_proj layer=0 pr=0.7 ppl=5.34
+  [2/224] self_attn.k_proj layer=0 pr=0.8 ppl=5.34
+  [2/224] self_attn.k_proj layer=0 pr=0.9 ppl=5.34
+  [3/224] self_attn.v_proj layer=0 pr=0.4 ppl=5.34
+  [3/224] self_attn.v_proj layer=0 pr=0.5 ppl=5.34
+  [3/224] self_attn.v_proj layer=0 pr=0.6 ppl=5.34
+  [3/224] self_attn.v_proj layer=0 pr=0.7 ppl=5.34
+  [3/224] self_attn.v_proj layer=0 pr=0.8 ppl=5.34
+  [3/224] self_attn.v_proj layer=0 pr=0.9 ppl=5.34
+  [4/224] self_attn.o_proj layer=0 pr=0.4 ppl=5.34
+  [4/224] self_attn.o_proj layer=0 pr=0.5 ppl=5.34
+  [4/224] self_attn.o_proj layer=0 pr=0.6 ppl=5.34
+  [4/224] self_attn.o_proj layer=0 pr=0.7 ppl=5.34
+  [4/224] self_attn.o_proj layer=0 pr=0.8 ppl=5.34
+  [4/224] self_attn.o_proj layer=0 pr=0.9 ppl=5.34
+  [5/224] mlp.gate_proj layer=0 pr=0.4 ppl=5.34
+  [5/224] mlp.gate_proj layer=0 pr=0.5 ppl=5.34
+  [5/224] mlp.gate_proj layer=0 pr=0.6 ppl=5.34
+  [5/224] mlp.gate_proj layer=0 pr=0.7 ppl=5.34
+  [5/224] mlp.gate_proj layer=0 pr=0.8 ppl=5.34
+  [5/224] mlp.gate_proj layer=0 pr=0.9 ppl=5.34
+  [6/224] mlp.up_proj layer=0 pr=0.4 ppl=5.34
+  [6/224] mlp.up_proj layer=0 pr=0.5 ppl=5.34
+  [6/224] mlp.up_proj layer=0 pr=0.6 ppl=5.34
+  [6/224] mlp.up_proj layer=0 pr=0.7 ppl=5.34
+  [6/224] mlp.up_proj layer=0 pr=0.8 ppl=5.34
+  [6/224] mlp.up_proj layer=0 pr=0.9 ppl=5.34
+  [7/224] mlp.down_proj layer=0 pr=0.4 ppl=5.35
+  [7/224] mlp.down_proj layer=0 pr=0.5 ppl=5.34
+  [7/224] mlp.down_proj layer=0 pr=0.6 ppl=5.34
+  [7/224] mlp.down_proj layer=0 pr=0.7 ppl=5.34
+  [7/224] mlp.down_proj layer=0 pr=0.8 ppl=5.34
+  [7/224] mlp.down_proj layer=0 pr=0.9 ppl=5.34
+  [8/224] self_attn.q_proj layer=1 pr=0.4 ppl=5.34
+  [8/224] self_attn.q_proj layer=1 pr=0.5 ppl=5.34
+  [8/224] self_attn.q_proj layer=1 pr=0.6 ppl=5.34
+  [8/224] self_attn.q_proj layer=1 pr=0.7 ppl=5.34
+  [8/224] self_attn.q_proj layer=1 pr=0.8 ppl=5.34
+  [8/224] self_attn.q_proj layer=1 pr=0.9 ppl=5.34
+  [9/224] self_attn.k_proj layer=1 pr=0.4 ppl=5.34
+  [9/224] self_attn.k_proj layer=1 pr=0.5 ppl=5.34
+  [9/224] self_attn.k_proj layer=1 pr=0.6 ppl=5.34
+  [9/224] self_attn.k_proj layer=1 pr=0.7 ppl=5.34
+  [9/224] self_attn.k_proj layer=1 pr=0.8 ppl=5.34
+  [9/224] self_attn.k_proj layer=1 pr=0.9 ppl=5.34
+  [10/224] self_attn.v_proj layer=1 pr=0.4 ppl=5.34
+  [10/224] self_attn.v_proj layer=1 pr=0.5 ppl=5.34
+  [10/224] self_attn.v_proj layer=1 pr=0.6 ppl=5.34
+  [10/224] self_attn.v_proj layer=1 pr=0.7 ppl=5.34
+  [10/224] self_attn.v_proj layer=1 pr=0.8 ppl=5.34
+  [10/224] self_attn.v_proj layer=1 pr=0.9 ppl=5.34
+  [11/224] self_attn.o_proj layer=1 pr=0.4 ppl=5.34
+  [11/224] self_attn.o_proj layer=1 pr=0.5 ppl=5.34
+  [11/224] self_attn.o_proj layer=1 pr=0.6 ppl=5.34
+  [11/224] self_attn.o_proj layer=1 pr=0.7 ppl=5.34
+  [11/224] self_attn.o_proj layer=1 pr=0.8 ppl=5.34
+  [11/224] self_attn.o_proj layer=1 pr=0.9 ppl=5.34
+  [12/224] mlp.gate_proj layer=1 pr=0.4 ppl=5.34
+  [12/224] mlp.gate_proj layer=1 pr=0.5 ppl=5.34
+  [12/224] mlp.gate_proj layer=1 pr=0.6 ppl=5.34
+  [12/224] mlp.gate_proj layer=1 pr=0.7 ppl=5.34
+  [12/224] mlp.gate_proj layer=1 pr=0.8 ppl=5.34
+  [12/224] mlp.gate_proj layer=1 pr=0.9 ppl=5.34
+  [13/224] mlp.up_proj layer=1 pr=0.4 ppl=5.35
+  [13/224] mlp.up_proj layer=1 pr=0.5 ppl=5.34
+  [13/224] mlp.up_proj layer=1 pr=0.6 ppl=5.34
+  [13/224] mlp.up_proj layer=1 pr=0.7 ppl=5.34
+  [13/224] mlp.up_proj layer=1 pr=0.8 ppl=5.34
+  [13/224] mlp.up_proj layer=1 pr=0.9 ppl=5.34
+  [14/224] mlp.down_proj layer=1 pr=0.4 ppl=5.36
+  [14/224] mlp.down_proj layer=1 pr=0.5 ppl=5.35
+  [14/224] mlp.down_proj layer=1 pr=0.6 ppl=5.35
+  [14/224] mlp.down_proj layer=1 pr=0.7 ppl=5.34
+  [14/224] mlp.down_proj layer=1 pr=0.8 ppl=5.34
+  [14/224] mlp.down_proj layer=1 pr=0.9 ppl=5.34
+  [15/224] self_attn.q_proj layer=2 pr=0.4 ppl=5.34
+  [15/224] self_attn.q_proj layer=2 pr=0.5 ppl=5.34
+  [15/224] self_attn.q_proj layer=2 pr=0.6 ppl=5.34
+  [15/224] self_attn.q_proj layer=2 pr=0.7 ppl=5.34
+  [15/224] self_attn.q_proj layer=2 pr=0.8 ppl=5.34
+  [15/224] self_attn.q_proj layer=2 pr=0.9 ppl=5.34
+  [16/224] self_attn.k_proj layer=2 pr=0.4 ppl=5.34
+  [16/224] self_attn.k_proj layer=2 pr=0.5 ppl=5.34
+  [16/224] self_attn.k_proj layer=2 pr=0.6 ppl=5.34
+  [16/224] self_attn.k_proj layer=2 pr=0.7 ppl=5.34
+  [16/224] self_attn.k_proj layer=2 pr=0.8 ppl=5.34
+  [16/224] self_attn.k_proj layer=2 pr=0.9 ppl=5.34
+  [17/224] self_attn.v_proj layer=2 pr=0.4 ppl=5.34
+  [17/224] self_attn.v_proj layer=2 pr=0.5 ppl=5.34
+  [17/224] self_attn.v_proj layer=2 pr=0.6 ppl=5.34
+  [17/224] self_attn.v_proj layer=2 pr=0.7 ppl=5.34
+  [17/224] self_attn.v_proj layer=2 pr=0.8 ppl=5.34
+  [17/224] self_attn.v_proj layer=2 pr=0.9 ppl=5.34
+  [18/224] self_attn.o_proj layer=2 pr=0.4 ppl=5.34
+  [18/224] self_attn.o_proj layer=2 pr=0.5 ppl=5.34
+  [18/224] self_attn.o_proj layer=2 pr=0.6 ppl=5.34
+  [18/224] self_attn.o_proj layer=2 pr=0.7 ppl=5.34
+  [18/224] self_attn.o_proj layer=2 pr=0.8 ppl=5.34
+  [18/224] self_attn.o_proj layer=2 pr=0.9 ppl=5.34
+  [19/224] mlp.gate_proj layer=2 pr=0.4 ppl=5.35
+  [19/224] mlp.gate_proj layer=2 pr=0.5 ppl=5.34
+  [19/224] mlp.gate_proj layer=2 pr=0.6 ppl=5.34
+  [19/224] mlp.gate_proj layer=2 pr=0.7 ppl=5.34
+  [19/224] mlp.gate_proj layer=2 pr=0.8 ppl=5.34
+  [19/224] mlp.gate_proj layer=2 pr=0.9 ppl=5.34
+  [20/224] mlp.up_proj layer=2 pr=0.4 ppl=5.35
+  [20/224] mlp.up_proj layer=2 pr=0.5 ppl=5.34
+  [20/224] mlp.up_proj layer=2 pr=0.6 ppl=5.34
+  [20/224] mlp.up_proj layer=2 pr=0.7 ppl=5.34
+  [20/224] mlp.up_proj layer=2 pr=0.8 ppl=5.34
+  [20/224] mlp.up_proj layer=2 pr=0.9 ppl=5.34
+  [21/224] mlp.down_proj layer=2 pr=0.4 ppl=5.37
+  [21/224] mlp.down_proj layer=2 pr=0.5 ppl=5.36