Upload folder using huggingface_hub

final/batch_top_k_0/google_gemma-2-2b_batch_top_k/resid_post_layer_12/trainer_0/ae.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c0dcdf066864a8c03ec88af579371d748be50f0a1a616578ee15cafbae2371f
+size 1208232982

final/batch_top_k_0/google_gemma-2-2b_batch_top_k/resid_post_layer_12/trainer_0/config.json

@@ -0,0 +1,32 @@
+{
+    "trainer": {
+        "trainer_class": "BatchTopKTrainer",
+        "dict_class": "BatchTopKSAE",
+        "lr": 0.0002,
+        "steps": 244140,
+        "auxk_alpha": 0.03125,
+        "warmup_steps": 1000,
+        "decay_start": 195312,
+        "threshold_beta": 0.999,
+        "threshold_start_step": 1000,
+        "top_k_aux": 1152,
+        "seed": 0,
+        "activation_dim": 2304,
+        "dict_size": 65536,
+        "k": 25,
+        "device": "cuda:5",
+        "layer": 12,
+        "lm_name": "google/gemma-2-2b",
+        "wandb_name": "BatchTopKTrainer-google/gemma-2-2b-resid_post_layer_12_trainer_0",
+        "submodule_name": "resid_post_layer_12"
+    },
+    "buffer": {
+        "d_submodule": 2304,
+        "io": "out",
+        "n_ctxs": 244,
+        "ctx_len": 1024,
+        "refresh_batch_size": 32,
+        "out_batch_size": 2048,
+        "device": "cuda:5"
+    }
+}

final/batch_top_k_0/google_gemma-2-2b_batch_top_k/resid_post_layer_12/trainer_0/eval_results.json

@@ -0,0 +1 @@
+{"l2_loss": 57.993975903614455, "l1_loss": 300.4578313253012, "l0": 24.746550307216413, "frac_variance_explained": 0.7531297063253012, "cossim": 0.9140625, "l2_ratio": 0.9140625, "relative_reconstruction_bias": 1.0003294427710843, "loss_original": 2.2974868222891565, "loss_reconstructed": 2.872176204819277, "loss_zero": 12.4375, "frac_recovered": 0.9432181852409639, "frac_alive": 0.9982452392578125, "hyperparameters": {"n_inputs": 1000, "context_length": 1024}}