weights

checkpoint_28000.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb103ebfaef05e2a2d18fd1c147c34133dba092eee6a603fd59a7548c3442a27
+size 604198864

config.json

@@ -0,0 +1,22 @@
+{
+    "trainer": {
+        "dict_class": "CrossCoder",
+        "trainer_class": "CrossCoderTrainer",
+        "activation_dim": 768,
+        "dict_size": 49152,
+        "lr": 0.001,
+        "l1_penalty": 0.03,
+        "warmup_steps": 1000,
+        "resample_steps": null,
+        "device": "cuda",
+        "layer": 3,
+        "lm_name": "tiny-gpt-27M-mixtral-5l-active-27M",
+        "wandb_name": "L3-mu3.0e-02-lr1e-03-64-mixed",
+        "submodule_name": null,
+        "use_mse_loss": false,
+        "code_normalization": "MIXED",
+        "code_normalization_alpha_sae": 0.7,
+        "code_normalization_alpha_cc": 0.3,
+        "target_rms": 1.0
+    }
+}

model_final.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57050bf18b863d2c43f8113d53477f11f134cdbaec6966ec321b54b83ca2b23d
+size 604198789