Instructions to use Coaster41/patchtst-sae-test with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- SAELens
How to use Coaster41/patchtst-sae-test with SAELens:
# pip install sae-lens from sae_lens import SAE sae, cfg_dict, sparsity = SAE.from_pretrained( release = "RELEASE_ID", # e.g., "gpt2-small-res-jb". See other options in https://github.com/jbloomAus/SAELens/blob/main/sae_lens/pretrained_saes.yaml sae_id = "SAE_ID", # e.g., "blocks.8.hook_resid_pre". Won't always be a hook point ) - Notebooks
- Google Colab
- Kaggle
Upload SAE blocks.0.hook_mlp_out
Browse files
blocks.0.hook_mlp_out/cfg.json
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"d_in": 256, "d_sae": 4096, "dtype": "float32", "device": "cuda", "apply_b_dec_to_input": false, "normalize_activations": "none", "reshape_activations": "none", "metadata": {"sae_lens_version": "6.5.3", "sae_lens_training_version": "6.5.3", "dataset_path": "autogluon/chronos_datasets", "hook_name": "blocks.0.hook_mlp_out", "model_name": "patchtst_relu", "model_class_name": "HookedTransformer", "hook_head_index": null, "context_size": 512, "num_patches": 32, "seqpos_slice": [null], "model_from_pretrained_kwargs": {"center_writing_weights": false}, "prepend_bos": false, "sequence_separator_token": null, "disable_concat_sequences": false}, "decoder_init_norm": 0.1, "l1_coefficient": 5, "lp_norm": 1.0, "l1_warm_up_steps": 1500, "architecture": "standard"}
|
blocks.0.hook_mlp_out/sae_weights.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:93cec67e50e1930889990e2928f5a8836e9250f1c3005590239006e014174ec7
|
| 3 |
+
size 8406320
|