Upload folder using huggingface_hub

Llama3_1Base-L4A-8x/checkpoints/final.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c9afcf686b00959a10f4c9e19e8144ea96e488d7cfea4820f76602bc449244d
+size 537076272

Llama3_1Base-L4A-8x/hyperparams.json

@@ -0,0 +1,23 @@
+{
+    "hook_point_in": "blocks.4.hook_attn_out",
+    "hook_point_out": "blocks.4.hook_attn_out",
+    "use_decoder_bias": true,
+    "apply_decoder_bias_to_pre_encoder": false,
+    "expansion_factor": 8,
+    "d_model": 4096,
+    "d_sae": 32768,
+    "norm_activation": "token-wise",
+    "dataset_average_activation_norm": null,
+    "decoder_exactly_fixed_norm": false,
+    "sparsity_include_decoder_norm": true,
+    "use_glu_encoder": false,
+    "init_decoder_norm": null,
+    "init_encoder_norm": null,
+    "init_encoder_with_decoder_transpose": true,
+    "l1_coefficient": 1.6e-05,
+    "l1_coefficient_warmup_steps": 14648,
+    "lp": 1,
+    "use_ghost_grads": false,
+    "tp_size": 1,
+    "ddp_size": 1
+}

Llama3_1Base-L4A-8x/lm_config.json

@@ -0,0 +1,7 @@
+{
+    "model_name": "meta-llama/Meta-Llama-3.1-8B",
+    "use_flash_attn": false,
+    "cache_dir": null,
+    "d_model": 4096,
+    "local_files_only": false
+}