Upload training_logs/it_phase5b_subseq_h8__seed42.json with huggingface_hub

training_logs/it_phase5b_subseq_h8__seed42.json

@@ -0,0 +1,95 @@
+{
+  "loss": [
+    176646.40625,
+    83437.6640625,
+    78705.734375,
+    76847.0390625,
+    76004.984375,
+    75623.4609375,
+    74648.8046875,
+    74630.4609375,
+    74368.6875,
+    74384.96875,
+    73978.6484375,
+    73781.1484375,
+    73685.5234375,
+    73644.7578125,
+    73773.625,
+    73894.203125
+  ],
+  "l0": [
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0,
+    500.0
+  ],
+  "steps_logged": [
+    0,
+    200,
+    400,
+    600,
+    800,
+    1000,
+    1200,
+    1400,
+    1600,
+    1800,
+    2000,
+    2200,
+    2400,
+    2600,
+    2800,
+    3000
+  ],
+  "final_step": 3000,
+  "converged": true,
+  "plateau_last": 0.008688741850918075,
+  "elapsed_s": 13961.203199386597,
+  "T_max": 10,
+  "t_sample": 5,
+  "shifts": [
+    1,
+    2,
+    5
+  ],
+  "matryoshka_h_size": 3686,
+  "row": 13,
+  "arch_id": "phase5b_subseq_h8",
+  "arch": "phase5b_subseq_h8",
+  "group": 2,
+  "src_class": "SubseqH8",
+  "src_module": "src.architectures.phase5b_subseq_sampling_txcdr",
+  "T": null,
+  "n_layers": null,
+  "k_win": 500,
+  "k_pos": 100,
+  "alpha": null,
+  "gamma": null,
+  "n_scales": null,
+  "seed": 42,
+  "d_in": 2304,
+  "d_sae": 18432,
+  "subject_model": "google/gemma-2-2b-it",
+  "anchor_layer": 13,
+  "mlc_layers": [
+    11,
+    12,
+    13,
+    14,
+    15
+  ],
+  "phase": "phase7_unification",
+  "run_id": "it_phase5b_subseq_h8__seed42"
+}