1st version of Indic Gemma

.gitignore

@@ -0,0 +1,112 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# pyenv
+.python-version
+
+# celery beat schedule file
+celerybeat-schedule
+
+# SageMath parsed files
+*.sage.py
+
+# dotenv
+.env
+
+# virtualenv
+.venv
+venv/
+ENV/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+
+# IDE settings
+.vscode/
+.idea/
+
+
+.editorconfig
+.envrc
+
+*.har

app.py

@@ -1 +1,191 @@
-test
+import typing
+import os
+import gradio as gr
+from typing import List
+
+import urllib.parse as urlparse
+import json
+from gradio_client import Client as GradioClient
+
+class GradioUserInference:
+    @staticmethod
+    def chat_interface_components(
+            sample_func: typing.Callable,
+    ):
+        """
+        The function `chat_interface_components` creates the components for a chat interface, including
+        a chat history, message box, buttons for submitting, stopping, and clearing the conversation,
+        and sliders for advanced options.
+        """
+
+        # _max_length = max_sequence_length
+        # _max_new_tokens = max_new_tokens
+        # _max_compile_tokens = max_compile_tokens
+
+        with gr.Column("100%"):
+            gr.Markdown(
+                f"# <h1><center style='color:#6600FF;'>[Indic Gemma 7B Demo](https://huggingface.co/Telugu-LLM-Labs/Indic-gemma-7b-finetuned-sft-Navarasa-2.0) -- Hosted on [Google Cloud TPU v4 Instance](https://cloud.google.com/tpu/docs/v4)</center></h1>",
+            )
+            history = gr.Chatbot(
+                elem_id="Indic",
+                label="Indic",
+                container=True,
+                height="55vh",
+            )
+            prompt = gr.Textbox(
+                show_label=False, placeholder='Enter your prompt here..', container=False
+            )
+            input = gr.Textbox(
+                show_label=False, placeholder='Provide Additional Input here..', container=False
+            )
+            with gr.Row():
+                submit = gr.Button(
+                    value="Run",
+                    variant="primary"
+                )
+                stop = gr.Button(
+                    value='Stop'
+                )
+                clear = gr.Button(
+                    value='Clear Conversation'
+                )
+            gr.Markdown(
+                "# <h4><center style='color:black;'>Hosted by "
+                "[Detoxio AI](https://detoxio.ai) for educational purpose. Thanks to Google Cloud for TPUV4 Infrastructure, and [Telugu-LLM-Labs](https://huggingface.co/Telugu-LLM-Labs) for finetuning Gemma on Indian Languages</center></h4>",
+            )
+
+        inputs = [
+            prompt,
+            input,
+            history,
+        ]
+
+        clear.click(fn=lambda: [], outputs=[history])
+        sub_event = submit.click(
+            fn=sample_func, inputs=inputs, outputs=[prompt, input, history]
+        )
+        txt_event = prompt.submit(
+            fn=sample_func, inputs=inputs, outputs=[prompt, input, history]
+        )
+        stop.click(
+            fn=None,
+            inputs=None,
+            outputs=None,
+            cancels=[txt_event, sub_event]
+        )
+
+    def _handle_gradio_input(
+            self,
+            prompt: str,
+            input: str,
+            history: List[List[str]],
+    ):
+        raise NotImplementedError()
+
+    def build_inference(
+            self,
+            sample_func: typing.Callable,
+    ) -> gr.Blocks:
+        """
+        The function "build_inference" returns a gr.Blocks object that model
+        interface components.
+        :return: a gr.Blocks object.
+        """
+        with gr.Blocks() as block:
+            self.chat_interface_components(
+                sample_func=sample_func,
+            )
+        return block
+
+
+class AssistantRole:
+    def __init__(self, name, seed_urls, poison_files_pattern):
+        self.name = name
+        self.seed_urls = seed_urls
+        self.poison_files_pattern = poison_files_pattern
+        
+
+
+class OutputParsingException(Exception):
+    pass
+
+class RemoteLLM(object):
+    def __init__(self, base_url):
+        """
+        Initialize instance.
+
+        Parameters:
+        - base_url (str): Base URL of the API.
+        """
+        self._base_url = base_url
+        self._client = GradioClient(base_url, verbose=False)
+
+    def generate(self, prompt: str):
+        """
+        Generate text using the model.
+
+        Parameters:
+        - prompt (str): Input prompt to be generated.
+
+        Returns:
+        - str: Reponse 
+        
+        """
+        result = self._client.predict(
+            prompt,	# str  in 'parameter_24' Textbox component
+            [],	# Tuple[str | Dict(file: filepath, alt_text: str | None) | None, str | Dict(file: filepath, alt_text: str | None) | None]  in Chatbot component
+            "",	# str  in 'System Prompt' Textbox component
+            "Chat",	# Literal['Chat', 'Instruct']  in 'Mode' Dropdown component
+            2048,	# float (numeric value between 1 and 10000) in 'Max Tokens' Slider component
+            360,	# float (numeric value between 256 and 10000) in 'Max New Tokens' Slider component
+            256,	# float (numeric value between 256 and 256) in 'Max Compile Tokens' Slider component
+            True,	# Literal[]  in 'Do Sample or Greedy Generation' Radio component
+            1,	# float (numeric value between 0.1 and 1) in 'Temperature' Slider component
+            1,	# float (numeric value between 0.1 and 1) in 'Top P' Slider component
+            50,	# float (numeric value between 1 and 100) in 'Top K' Slider component
+            5,	# float (numeric value between 0.1 and 5) in 'Repetition Penalty' Slider component
+            api_name="/sample_gradio"
+        )
+
+        try:
+            out = result[1][0][1]
+            return out
+        except Exception as ex:
+            raise OutputParsingException(ex, result)
+
+
+class RAGApp(GradioUserInference):
+    def __init__(self, url):
+        self._llm = RemoteLLM(url)
+        self._gradio_app_handle=None
+
+    def _generate(self, prompt, input):
+        if input:
+            combined_prompt = f"{prompt} INPUT {input}"
+        else:
+            combined_prompt = prompt
+        return self._llm.generate(combined_prompt)
+
+    def _handle_gradio_input(self,
+            prompt: str,
+            input: str,
+            history: List[List[str]]):
+        response = self._generate(prompt, input)
+        if input:
+            combined_prompt = f"{prompt}\n\n\"{input}\""
+        else:
+            combined_prompt = prompt
+        history.append([combined_prompt, ""])
+        history[-1][-1] = response
+        yield "", "", history
+        # Initial update of documents and launch interface
+
+    def run(self):
+        self._gradio_app_handle = self.build_inference(self._handle_gradio_input)
+        self._gradio_app_handle.launch()
+
+
+if __name__ == "__main__":
+    url = os.environ['INDIC_GEMMA_HOSTED_URL']
+    app = RAGApp(url)
+    app.run()

requirements.txt

@@ -0,0 +1,3 @@
+retry
+gradio-client
+gradio