Updates inference notebook.

model/notebooks/inference.ipynb

@@ -16,7 +16,7 @@
     "import torch\n",
     "\n",
     "from huggingface_hub import hf_hub_download\n",
-    "from transformers import AutoTokenizer\n",
+    "from transformers import AutoTokenizer, T5Tokenizer\n",
     "\n",
     "from model.distilbert import DistilBertClassificationModel\n",
     "from model.scibert import SciBertClassificationModel\n",
@@ -39,7 +39,7 @@
    "outputs": [],
    "source": [
     "# Baseline\n",
-    "repo_id = \"ppak10/defect-classification-distilbert-baseline-25-epochs\"\n",
+    "# repo_id = \"ppak10/defect-classification-distilbert-baseline-25-epochs\"\n",
     "# repo_id = \"ppak10/defect-classification-scibert-baseline-25-epochs\"\n",
     "# repo_id = \"ppak10/defect-classification-llama-baseline-25-epochs\"\n",
     "# repo_id = \"ppak10/defect-classification-t5-baseline-25-epochs\"\n",
@@ -48,7 +48,7 @@
     "# repo_id = \"ppak10/defect-classification-distilbert-prompt-02-epochs\"\n",
     "# repo_id = \"ppak10/defect-classification-scibert-prompt-02-epochs\"\n",
     "# repo_id = \"ppak10/defect-classification-llama-prompt-02-epochs\"\n",
-    "# repo_id = \"ppak10/defect-classification-t5-prompt-02-epochs\"\n",
+    "repo_id = \"ppak10/defect-classification-t5-prompt-02-epochs\"\n",
     "\n",
     "# Initialize the model\n",
     "model = DistilBertClassificationModel(repo_id)\n",
@@ -63,8 +63,11 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "# Load the tokenizer\n",
-    "tokenizer = AutoTokenizer.from_pretrained(repo_id)\n",
+    "# Uncomment for DistilBERT, SciBERT, and Llama\n",
+    "# tokenizer = AutoTokenizer.from_pretrained(repo_id)\n",
+    "\n",
+    "# Uncomment for T5\n",
+    "tokenizer = T5Tokenizer.from_pretrained(repo_id)\n",
     "\n",
     "# Loads classification head weights\n",
     "classification_head_path = hf_hub_download(\n",
@@ -84,10 +87,10 @@
    "outputs": [],
    "source": [
     "# Baseline\n",
-    "text = \"Ti-6Al-4V[SEP]280.0 W[SEP]400.0 mm/s[SEP]100.0 microns[SEP]50.0 microns[SEP]100.0 microns\"\n",
+    "# text = \"Ti-6Al-4V[SEP]280.0 W[SEP]400.0 mm/s[SEP]100.0 microns[SEP]50.0 microns[SEP]100.0 microns\"\n",
     "\n",
     "# Prompt\n",
-    "# text = \"What are the likely imperfections that occur in Ti-6Al-4V L-PBF builds at 280.0 W, given a 100.0 microns beam diameter, a 400.0 mm/s scan speed, a 100.0 microns hatch spacing, and a 50.0 microns layer height?\""
+    "text = \"What are the likely imperfections that occur in Ti-6Al-4V L-PBF builds at 280.0 W, given a 100.0 microns beam diameter, a 400.0 mm/s scan speed, a 100.0 microns hatch spacing, and a 50.0 microns layer height?\""
    ]
   },
   {
@@ -99,7 +102,7 @@
     "# Tokenize inputs \n",
     "inputs = tokenizer(text, return_tensors=\"pt\", truncation=True, padding=\"max_length\", max_length=256)\n",
     "\n",
-    "# For scibert\n",
+    "# For SciBERT specific case. \n",
     "inputs_kwargs = {}\n",
     "for key, value in inputs.items():\n",
     "    if key not in [\"token_type_ids\"]:\n",