initial commit

.gitignore

@@ -0,0 +1,2 @@
+.venv/
+.DS_Store

sparse_probing/requirements.txt

@@ -0,0 +1,12 @@
+sae_lens
+transformer-lens
+torch==2.3.0
+einops==0.8.0
+matplotlib==3.8.4
+numpy==1.26.4
+pandas==2.1.2
+plotly==5.22.0
+tqdm==4.66.4
+pytest==8.3.2
+nbformat==5.10.4
+ipykernel==6.29.5

sparse_probing/src/probe_training.ipynb

@@ -0,0 +1,156 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "%load_ext autoreload\n",
+    "%autoreload 2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "<torch.autograd.grad_mode.set_grad_enabled at 0x13f1b51d0>"
+      ]
+     },
+     "execution_count": 2,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# Standard imports\n",
+    "import os\n",
+    "import torch\n",
+    "from tqdm import tqdm\n",
+    "import plotly.express as px\n",
+    "\n",
+    "# Imports for displaying vis in Colab / notebook\n",
+    "import webbrowser\n",
+    "import http.server\n",
+    "import socketserver\n",
+    "import threading\n",
+    "PORT = 8000\n",
+    "\n",
+    "torch.set_grad_enabled(False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Device: mps\n"
+     ]
+    }
+   ],
+   "source": [
+    "if torch.backends.mps.is_available():\n",
+    "    device = \"mps\"\n",
+    "else:\n",
+    "    device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
+    "\n",
+    "\n",
+    "print(f\"Device: {device}\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/Users/adamkarvonen/SAE_bench_template/.venv/lib/python3.11/site-packages/tqdm/auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
+      "  from .autonotebook import tqdm as notebook_tqdm\n",
+      "/Users/adamkarvonen/SAE_bench_template/.venv/lib/python3.11/site-packages/transformers/tokenization_utils_base.py:1617: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be deprecated in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884\n",
+      "  warnings.warn(\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Loaded pretrained model pythia-70m-deduped into HookedTransformer\n"
+     ]
+    }
+   ],
+   "source": [
+    "from datasets import load_dataset  \n",
+    "from transformer_lens import HookedTransformer\n",
+    "from sae_lens import SAE\n",
+    "\n",
+    "model = HookedTransformer.from_pretrained(\"pythia-70m-deduped\", device = device)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "SAEConfig(architecture='standard', d_in=512, d_sae=16384, activation_fn_str='topk', apply_b_dec_to_input=True, finetuning_scaling_factor=False, context_size=128, model_name='pythia-70m-deduped', hook_name='blocks.4.hook_resid_post', hook_layer=4, hook_head_index=None, prepend_bos=True, dataset_path='monology/pile-uncopyrighted', dataset_trust_remote_code=False, normalize_activations='none', dtype='float32', device='mps', sae_lens_training_version=None, activation_fn_kwargs={'k': 80}, neuronpedia_id=None, model_from_pretrained_kwargs={})\n"
+     ]
+    }
+   ],
+   "source": [
+    "# the cfg dict is returned alongside the SAE since it may contain useful information for analysing the SAE (eg: instantiating an activation store)\n",
+    "# Note that this is not the same as the SAEs config dict, rather it is whatever was in the HF repo, from which we can extract the SAE config dict\n",
+    "# We also return the feature sparsities which are stored in HF for convenience. \n",
+    "sae, cfg_dict, sparsity = SAE.from_pretrained(\n",
+    "    release = \"sae_bench_pythia70m_sweep_topk_ctx128_0730\",\n",
+    "    sae_id = \"blocks.4.hook_resid_post__trainer_10\",\n",
+    "    device = device\n",
+    ")\n",
+    "sae = sae.to(device=device)\n",
+    "\n",
+    "print(sae.cfg)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": ".venv",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.11.8"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}