Fix the neuron resampler approach (#141)

This was normalising across an incorrect dimension and not firing at the correct point.
ai-safety-foundation · Dec 9, 2023 · be08405 · be08405
1 parent 15378ab
commit be08405
Show file tree

Hide file tree

Showing 30 changed files with 1,083 additions and 738 deletions.
diff --git a/docs/content/demo.ipynb b/docs/content/demo.ipynb
@@ -38,9 +38,18 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": 5,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "The autoreload extension is already loaded. To reload it, use:\n",
+      "  %reload_ext autoreload\n"
+     ]
+    }
+   ],
    "source": [
     "# Check if we're in Colab\n",
     "try:\n",
@@ -62,22 +71,23 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 6,
    "metadata": {},
    "outputs": [],
    "source": [
     "import os\n",
     "\n",
     "from sparse_autoencoder import (\n",
-    "    sweep,\n",
-    "    SweepConfig,\n",
+    "    ActivationResamplerHyperparameters,\n",
     "    Hyperparameters,\n",
-    "    SourceModelHyperparameters,\n",
-    "    Parameter,\n",
-    "    SourceDataHyperparameters,\n",
-    "    Method,\n",
     "    LossHyperparameters,\n",
+    "    Method,\n",
     "    OptimizerHyperparameters,\n",
+    "    Parameter,\n",
+    "    SourceDataHyperparameters,\n",
+    "    SourceModelHyperparameters,\n",
+    "    sweep,\n",
+    "    SweepConfig,\n",
     ")\n",
     "import wandb\n",
     "\n",
@@ -103,7 +113,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": 7,
    "metadata": {},
    "outputs": [
     {
@@ -112,28 +122,31 @@
        "SweepConfig(parameters=Hyperparameters(\n",
        "    source_data=SourceDataHyperparameters(dataset_path=Parameter(value=NeelNanda/c4-code-tokenized-2b), context_size=Parameter(value=128))\n",
        "    source_model=SourceModelHyperparameters(name=Parameter(value=gelu-2l), hook_site=Parameter(value=mlp_out), hook_layer=Parameter(value=0), hook_dimension=Parameter(value=512), dtype=Parameter(value=float32))\n",
-       "    activation_resampler=ActivationResamplerHyperparameters(resample_interval=Parameter(value=200000000), max_resamples=Parameter(value=4), n_steps_collate=Parameter(value=100000000), resample_dataset_size=Parameter(value=819200), dead_neuron_threshold=Parameter(value=0.0))\n",
-       "    autoencoder=AutoencoderHyperparameters(expansion_factor=Parameter(value=4))\n",
-       "    loss=LossHyperparameters(l1_coefficient=Parameter(values=[0.001, 0.0001, 1e-05]))\n",
-       "    optimizer=OptimizerHyperparameters(lr=Parameter(values=[0.001, 0.0001, 1e-05]), adam_beta_1=Parameter(value=0.9), adam_beta_2=Parameter(value=0.99), adam_weight_decay=Parameter(value=0.0), amsgrad=Parameter(value=False), fused=Parameter(value=False))\n",
-       "    pipeline=PipelineHyperparameters(log_frequency=Parameter(value=100), source_data_batch_size=Parameter(value=12), train_batch_size=Parameter(value=4096), max_store_size=Parameter(value=3145728), max_activations=Parameter(value=2000000000), checkpoint_frequency=Parameter(value=100000000), validation_frequency=Parameter(value=314572800), validation_number_activations=Parameter(value=1024))\n",
+       "    activation_resampler=ActivationResamplerHyperparameters(resample_interval=Parameter(value=197885952), max_n_resamples=Parameter(value=4), n_activations_activity_collate=Parameter(value=98942976), resample_dataset_size=Parameter(value=819200), threshold_is_dead_portion_fires=Parameter(value=1e-06))\n",
+       "    autoencoder=AutoencoderHyperparameters(expansion_factor=Parameter(value=2))\n",
+       "    loss=LossHyperparameters(l1_coefficient=Parameter(max=0.01, min=0.004))\n",
+       "    optimizer=OptimizerHyperparameters(lr=Parameter(max=0.001, min=1e-05), adam_beta_1=Parameter(value=0.9), adam_beta_2=Parameter(value=0.99), adam_weight_decay=Parameter(value=0.0), amsgrad=Parameter(value=False), fused=Parameter(value=False))\n",
+       "    pipeline=PipelineHyperparameters(log_frequency=Parameter(value=100), source_data_batch_size=Parameter(value=16), train_batch_size=Parameter(value=8192), max_store_size=Parameter(value=2998272), max_activations=Parameter(value=1999847424), checkpoint_frequency=Parameter(value=47972352), validation_frequency=Parameter(value=99999744), validation_number_activations=Parameter(value=8192))\n",
        "    random_seed=Parameter(value=49)\n",
-       "), method=<Method.RANDOM: 'random'>, metric=Metric(name=total_loss, goal=minimize), command=None, controller=None, description=None, earlyterminate=None, entity=None, imageuri=None, job=None, kind=None, name=None, program=None, project=None, runcap=None)"
+       "), method=<Method.RANDOM: 'random'>, metric=Metric(name=train/loss/total_loss, goal=minimize), command=None, controller=None, description=None, earlyterminate=None, entity=None, imageuri=None, job=None, kind=None, name=None, program=None, project=None)"
       ]
      },
-     "execution_count": 3,
+     "execution_count": 7,
      "metadata": {},
      "output_type": "execute_result"
     }
    ],
    "source": [
     "sweep_config = SweepConfig(\n",
     "    parameters=Hyperparameters(\n",
+    "        activation_resampler=ActivationResamplerHyperparameters(\n",
+    "            threshold_is_dead_portion_fires=Parameter(1e-6),\n",
+    "        ),\n",
     "        loss=LossHyperparameters(\n",
-    "            l1_coefficient=Parameter(values=[1e-3, 1e-4, 1e-5]),\n",
+    "            l1_coefficient=Parameter(max=1e-2, min=4e-3),\n",
     "        ),\n",
     "        optimizer=OptimizerHyperparameters(\n",
-    "            lr=Parameter(values=[1e-3, 1e-4, 1e-5]),\n",
+    "            lr=Parameter(max=1e-3, min=1e-5),\n",
     "        ),\n",
     "        source_model=SourceModelHyperparameters(\n",
     "            name=Parameter(\"gelu-2l\"),\n",

diff --git a/docs/content/flexible_demo.ipynb b/docs/content/flexible_demo.ipynb
@@ -379,7 +379,10 @@
    "outputs": [],
    "source": [
     "activation_resampler = ActivationResampler(\n",
-    "    resample_interval=10_000, n_steps_collate=10_000, max_resamples=5\n",
+    "    resample_interval=10_000,\n",
+    "    n_activations_activity_collate=10_000,\n",
+    "    max_n_resamples=5,\n",
+    "    n_learned_features=autoencoder.n_learned_features,\n",
     ")"
    ]
   },
@@ -400,9 +403,24 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 9,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "2fe4955deca9463dbed606c9452d518e",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Resolving data files:   0%|          | 0/28 [00:00<?, ?it/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
    "source": [
     "source_data = PreTokenizedDataset(\n",
     "    dataset_path=\"NeelNanda/c4-code-tokenized-2b\", context_size=int(hyperparameters[\"context_size\"])\n",
@@ -429,7 +447,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 10,
    "metadata": {},
    "outputs": [],
    "source": [