working with gpt2

.gitignore

@@ -0,0 +1,2 @@
+__pycache__
+flagged/

app.py

@@ -0,0 +1,50 @@
+import gradio as gr
+
+# A sequence of lines both those typed in and the line so far
+# when save is clicked the txt file is downloaded
+lines = []
+
+
+def greet(name):
+    return "Hello " + name + "!"
+
+
+def add_to_lines(new_line):
+    # TODO: add new_line to the array
+    # TODO: send the full text to the language model generator
+    lines.append(new_line)
+    return "this is the next line in the poem"
+
+
+def downloadtext():
+    # somehow print the values from the list
+    pass
+
+
+# TODO: somehow loop and create all of the text added so far
+
+with gr.Blocks() as demo:
+    gr.Markdown("Start typing below and then click **Run** to see the output.")
+    # Need to render a group of these
+    with gr.Group():
+        with gr.Row():
+            inp = gr.Textbox(placeholder="What is your name?")
+            out = gr.Textbox()
+        btn = gr.Button("Run")
+        btn.click(fn=add_to_lines, inputs=inp, outputs=out)
+
+
+# demo = gr.Interface(
+#     fn=getnextline,
+#     inputs=gr.Textbox(lines=1, placeholder="..."),
+#     outputs=gr.Markdown(
+#         """
+#         text as output
+#         """
+#     ),
+#     allow_flagging="never",
+# )
+
+
+if __name__ == "__main__":
+    demo.launch()

fine-tune-llm.ipynb

@@ -103,6 +103,183 @@
     "classifier(\"We are very happy to show you the 🤗 Transformers library.\")"
    ]
   },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Take a prompt and generate a line of text"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "6f80cb24ef764bd192e5d3af79f9f5f1",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Downloading:   0%|          | 0.00/665 [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "4e64c3a035b54f0c90ca5cc6e341ad21",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Downloading:   0%|          | 0.00/475M [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "All model checkpoint layers were used when initializing TFGPT2LMHeadModel.\n",
+      "\n",
+      "All the layers of TFGPT2LMHeadModel were initialized from the model checkpoint at gpt2.\n",
+      "If your task is similar to the task the model of the checkpoint was trained on, you can already use TFGPT2LMHeadModel for predictions without further training.\n"
+     ]
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "4f442dd13c5747e1811c2199423de0c9",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Downloading:   0%|          | 0.00/0.99M [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "a88572455b744b18b99c5bd775944d77",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Downloading:   0%|          | 0.00/446k [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "d31a3e7e53a7422eabfcb61ff5248b8b",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Downloading:   0%|          | 0.00/1.29M [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Setting `pad_token_id` to 50256 (first `eos_token_id`) to generate sequence\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "[{'generated_text': 'Hello, I\\'m a language model for the world of design,\" explained the senior designer. \"In JavaScript, each line represents a block of code that'},\n",
+       " {'generated_text': \"Hello, I'm a language modeler extraordinaire. So if you're looking for an elegant and flexible way to express your language or for an\"},\n",
+       " {'generated_text': \"Hello, I'm a language modeler for Ruby using R, and as a newbie to Rails, I've been very interested in these two techniques\"}]"
+      ]
+     },
+     "execution_count": 3,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "from transformers import pipeline\n",
+    "generator = pipeline('text-generation', model = 'gpt2')\n",
+    "generator(\"Hello, I'm a language model\", max_length = 30, num_return_sequences=3)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "application/vnd.jupyter.widget-view+json": {
+       "model_id": "af1c7c8ed0a84d74823d961f6bdb1e0d",
+       "version_major": 2,
+       "version_minor": 0
+      },
+      "text/plain": [
+       "Downloading:   0%|          | 0.00/523M [00:00<?, ?B/s]"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "Setting `pad_token_id` to `eos_token_id`:50256 for open-end generation.\n"
+     ]
+    },
+    {
+     "data": {
+      "text/plain": [
+       "[{'generated_text': 'something to start with, they say, but even if it was just to add an element of humor to the recipe, she said it could become a'},\n",
+       " {'generated_text': \"something to start with. You don't have to have a real connection with the people in this building to have any sort of connection with them. And\"},\n",
+       " {'generated_text': \"something to start with, as I've seen several years to come. You're supposed to be a good, loving parent, and your kids are supposed\"}]"
+      ]
+     },
+     "execution_count": 3,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "from transformers import pipeline\n",
+    "from transformers import AutoTokenizer, AutoModelForCausalLM\n",
+    "\n",
+    "# tokenizer = AutoTokenizer.from_pretrained(\"BritishLibraryLabs/bl-books-genre\")\n",
+    "# model = AutoModelForCausalLM.from_pretrained(\"BritishLibraryLabs/bl-books-genre\")\n",
+    "# \"BritishLibraryLabs/bl-books-genre\"\n",
+    "\n",
+    "tokenizer = AutoTokenizer.from_pretrained(\"gpt2\")\n",
+    "model = AutoModelForCausalLM.from_pretrained(\"gpt2\")\n",
+    "\n",
+    "# generator = pipeline('text-generation', model = \"BritishLibraryLabs/bl-books-genre\")\n",
+    "# generator(\"Hello, I'm a language model\", max_length = 30, num_return_sequences=3)\n",
+    "\n",
+    "generator = pipeline(task=\"text-generation\", model=model, tokenizer=tokenizer)\n",
+    "generator('something to start with', max_length = 30, num_return_sequences=3)\n"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,