Show that model did train
Browse files
GPT2_Linear_4bit_training.ipynb
CHANGED
@@ -822,6 +822,48 @@
|
|
822 |
}
|
823 |
]
|
824 |
},
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
825 |
{
|
826 |
"cell_type": "markdown",
|
827 |
"source": [
|
|
|
822 |
}
|
823 |
]
|
824 |
},
|
825 |
+
{
|
826 |
+
"cell_type": "code",
|
827 |
+
"source": [
|
828 |
+
"inputs = {k:v.cuda() for k,v in tokenizer(\"\"\"\n",
|
829 |
+
"You are an AI assistant. You will be given a question. You must generate a short and factual answer.\n",
|
830 |
+
"What is the capital city of France?\n",
|
831 |
+
"\"\"\", return_tensors='pt').items()}\n",
|
832 |
+
"outputs = model.generate(**inputs, max_new_tokens=16, temperature=0.5, do_sample=True)\n",
|
833 |
+
"print(tokenizer.decode(outputs[0]), \"...\")"
|
834 |
+
],
|
835 |
+
"metadata": {
|
836 |
+
"colab": {
|
837 |
+
"base_uri": "https://localhost:8080/"
|
838 |
+
},
|
839 |
+
"id": "wr6bfZ0wyk3c",
|
840 |
+
"outputId": "ca4ede1b-7456-43ea-ce52-961c1383dff8"
|
841 |
+
},
|
842 |
+
"execution_count": 16,
|
843 |
+
"outputs": [
|
844 |
+
{
|
845 |
+
"output_type": "stream",
|
846 |
+
"name": "stderr",
|
847 |
+
"text": [
|
848 |
+
"Setting `pad_token_id` to `eos_token_id`:50256 for open-end generation.\n"
|
849 |
+
]
|
850 |
+
},
|
851 |
+
{
|
852 |
+
"output_type": "stream",
|
853 |
+
"name": "stdout",
|
854 |
+
"text": [
|
855 |
+
"\n",
|
856 |
+
"You are an AI assistant. You will be given a question. You must generate a short and factual answer.\n",
|
857 |
+
"What is the capital city of France?\n",
|
858 |
+
"\n",
|
859 |
+
"\n",
|
860 |
+
"Paris\n",
|
861 |
+
"\n",
|
862 |
+
"Paris is the capital of France. The city is located ...\n"
|
863 |
+
]
|
864 |
+
}
|
865 |
+
]
|
866 |
+
},
|
867 |
{
|
868 |
"cell_type": "markdown",
|
869 |
"source": [
|