Instructions to use suchitg/testrun-saes with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- SAELens
How to use suchitg/testrun-saes 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 gen_000/blocks.4.hook_resid_post
Browse files
gen_000/blocks.4.hook_resid_post/cfg.json
CHANGED
|
@@ -1 +1 @@
|
|
| 1 |
-
{"d_in": 2304, "d_sae": 16384, "dtype": "float32", "device": "cuda:0", "apply_b_dec_to_input": true, "normalize_activations": "none", "reshape_activations": "none", "metadata": {"sae_lens_version": "6.36.2", "sae_lens_training_version": "6.36.2", "dataset_path": "monology/pile-uncopyrighted", "hook_name": "blocks.4.hook_resid_post", "model_name": "gemma-2-2b", "model_class_name": "HookedTransformer", "hook_head_index": null, "context_size": 1024, "seqpos_slice": [null], "model_from_pretrained_kwargs": {"center_writing_weights": false}, "prepend_bos": true, "exclude_special_tokens": false, "sequence_separator_token": "bos", "disable_concat_sequences": false}, "decoder_init_norm": 0.1, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.05, "jumprelu_sparsity_loss_mode": "step", "l0_coefficient": 0.
|
|
|
|
| 1 |
+
{"d_in": 2304, "d_sae": 16384, "dtype": "float32", "device": "cuda:0", "apply_b_dec_to_input": true, "normalize_activations": "none", "reshape_activations": "none", "metadata": {"sae_lens_version": "6.36.2", "sae_lens_training_version": "6.36.2", "dataset_path": "monology/pile-uncopyrighted", "hook_name": "blocks.4.hook_resid_post", "model_name": "gemma-2-2b", "model_class_name": "HookedTransformer", "hook_head_index": null, "context_size": 1024, "seqpos_slice": [null], "model_from_pretrained_kwargs": {"center_writing_weights": false}, "prepend_bos": true, "exclude_special_tokens": false, "sequence_separator_token": "bos", "disable_concat_sequences": false}, "decoder_init_norm": 0.1, "jumprelu_init_threshold": 0.001, "jumprelu_bandwidth": 0.05, "jumprelu_sparsity_loss_mode": "step", "l0_coefficient": 0.1, "l0_warm_up_steps": 1000, "pre_act_loss_coefficient": null, "jumprelu_tanh_scale": 4.0, "architecture": "jumprelu"}
|
gen_000/blocks.4.hook_resid_post/sae_weights.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 302130576
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:13dd8b5fe62006d51f46bcb66d0feb9522ba016de63bf04415cc48ccc243782b
|
| 3 |
size 302130576
|