Upload 2 files

matryoshka_sae_top_300/config.json

@@ -0,0 +1 @@
+{"name": "matryoshka_top_300_dict_65536", "seed": 49, "batch_size": 4096, "lr": 0.003, "l1_coeff": 0, "beta1": 0.9, "beta2": 0.99, "max_grad_norm": 100000, "num_epochs": 2, "sae_type": "matryoshka_sae", "dtype": "torch.bfloat16", "model_dtype": "torch.bfloat16", "n_batches_to_dead": 20, "act_size": 4096, "top_k": 300, "top_k_aux": 512, "aux_penalty": 0.03125, "input_unit_norm": false, "group_sizes": [1024, 1024, 2048, 4096, 8192, 16384, 32768], "dict_size": 65536, "dataset_path": "/scratch/daniel_garibi/t5_embedding_data/t5_embedding_gpu.zarr", "results_path": "/scratch/daniel_garibi/results", "perf_log_freq": 100, "checkpoint_freq": 10000, "data_loader_workers": 1, "bandwidth": 0.001}

matryoshka_sae_top_300/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:306e45c15a086065fa5724b8d98e83e51dbe79fb44112f6bd051ad660d866086
+size 2147764955