Upload finetune.ipynb

finetune.ipynb

@@ -143,7 +143,13 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "Epoch 1/7, Train Batch 9/82, Train Loss: 0.071097498259893286"
+      "Epoch 1/7, Train Loss: 0.5704409927129745, Valid Loss: 0.3897556330009205, lr: 0.0005\n",
+      "Epoch 2/7, Train Loss: 0.403159585849541, Valid Loss: 0.24347291268953464, lr: 0.0001\n",
+      "Epoch 3/7, Train Loss: 0.27275778398644634, Valid Loss: 0.13453135721203757, lr: 2e-05\n",
+      "Epoch 4/7, Train Loss: 0.1717792261482739, Valid Loss: 0.07003633981775038, lr: 4.000000000000001e-06\n",
+      "Epoch 5/7, Train Loss: 0.10565993448764813, Valid Loss: 0.03993069042065522, lr: 8.000000000000002e-07\n",
+      "Epoch 6/7, Train Loss: 0.0703794076675322, Valid Loss: 0.02531332855408148, lr: 1.6000000000000003e-07\n",
+      "Epoch 7/7, Train Batch 7/82, Train Loss: 0.0036747022645502556"
      ]
     }
    ],
@@ -183,8 +189,18 @@
     "\n",
     "    print(f'\\rEpoch {epoch+1}/{num_epochs}, Train Loss: {train_loss/len(train_loader)}, Valid Loss: {valid_loss/len(valid_loader)}, lr: {learning_rate}',end=\"\\n\")\n",
     "    learning_rate /= 5\n",
-    "model.save_pretrained('fine_tuned_model')\n",
-    "tokenizer.save_pretrained('fine_tuned_tokenizer')"
+    "model.save_pretrained('brunosan/GPT2-impactscience')\n",
+    "tokenizer.save_pretrained('brunosan/GPT2-impactscience')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "model.save_pretrained('brunosan/GPT2-impactscience')\n",
+    "tokenizer.save_pretrained('brunosan/GPT2-impactscience')"
    ]
   },
   {
@@ -218,9 +234,9 @@
     "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
     "\n",
     "# Load the fine-tuned model\n",
-    "model_path = 'fine_tuned_model'\n",
-    "tokenizer = GPT2Tokenizer.from_pretrained('fine_tuned_tokenizer')\n",
-    "model = GPT2LMHeadModel.from_pretrained('fine_tuned_model').to(device)\n",
+    "model_path = 'brunosan/GPT2-impactscience'\n",
+    "tokenizer = GPT2Tokenizer.from_pretrained(model_path)\n",
+    "model = GPT2LMHeadModel.from_pretrained(model_path).to(device)\n",
     "\n",
     "# Set the pad_token_id to the same value as the unk_token_id\n",
     "#model.config.pad_token_id = tokenizer.unk_token_id\n",
@@ -232,7 +248,7 @@
     "temperature = 1.0\n",
     "\n",
     "# Generate text using beam search, n-grams, and other techniques\n",
-    "prompt = \"Antonio,\"\n",
+    "prompt = \"The impact of climate change on \"\n",
     "\n",
     "def generate(prompt):\n",
     "    input_ids = tokenizer.encode(prompt, return_tensors='pt').to(device)\n",