initial commit

app.py

@@ -0,0 +1,230 @@
+def transcribe_youtube_video(url, force_transcribe=False):
+
+    text = ''
+    try:
+        from youtube_transcript_api import YouTubeTranscriptApi
+        import pytube
+        from pytube import YouTube
+        
+        vid_id = pytube.extract.video_id(url)
+        temp = YouTubeTranscriptApi.get_transcript(vid_id)
+        for t in temp:
+            text+=t['text']+' '
+        yt = YouTube(str(url))
+        
+    except:
+        pass
+
+    if text == '' or force_transcribe:
+        from pytube import YouTube
+        import torch
+        
+        save_dir="./docs/youtube/"
+        yt = YouTube(str(url))
+        audio = yt.streams.filter(only_audio = True).first()
+        out_file = audio.download(filename="audio.mp3",output_path = save_dir)
+        
+        import transformers
+
+        whisper_asr = transformers.pipeline(
+            "automatic-speech-recognition", model="openai/whisper-large", device_map= 'auto',
+        )
+        
+        whisper_asr.model.config.forced_decoder_ids = (
+            whisper_asr.tokenizer.get_decoder_prompt_ids(
+                language="en", 
+                task="transcribe"
+            )
+        )
+        temp = whisper_asr(out_file,chunk_length_s=20)
+        text = temp['text']
+
+        del(whisper_asr)
+        torch.cuda.empty_cache()
+    
+    return yt.title, text
+
+def summarize_text(title,text):
+    
+    from langchain.chains.llm import LLMChain
+    from langchain.prompts import PromptTemplate
+    from langchain.chains import ReduceDocumentsChain, MapReduceDocumentsChain
+    from langchain.chains.combine_documents.stuff import StuffDocumentsChain
+    import torch
+    import transformers
+    from transformers import BitsAndBytesConfig
+    from transformers import AutoTokenizer, AutoModelForCausalLM
+
+    quantization_config = BitsAndBytesConfig(
+        load_in_4bit=True,
+        bnb_4bit_compute_dtype=torch.float16,
+        bnb_4bit_quant_type="nf4",
+        bnb_4bit_use_double_quant=True,
+    )
+
+    # model = "nomic-ai/gpt4all-falcon"
+    model = "tiiuae/falcon-7b-instruct"
+
+    tokenizer = AutoTokenizer.from_pretrained(model,trust_remote_code=True,)
+    model = AutoModelForCausalLM.from_pretrained(model,
+                                                # trust_remote_code=True,
+                                                quantization_config=quantization_config,
+                                                )
+
+    from langchain import HuggingFacePipeline
+    import torch
+
+    pipeline = transformers.pipeline(
+        "text-generation",
+        model=model,
+        tokenizer=tokenizer,
+        torch_dtype=torch.bfloat16,
+        device_map="auto",
+        max_new_tokens = 150,
+        pad_token_id=tokenizer.eos_token_id,
+        # device=-1,
+    )
+
+    llm = HuggingFacePipeline(pipeline=pipeline)
+
+    pipeline2 = transformers.pipeline(
+        "text-generation",
+        model=model,
+        tokenizer=tokenizer,
+        torch_dtype=torch.bfloat16,
+        device_map="auto",
+        max_new_tokens = 250,
+        pad_token_id=tokenizer.eos_token_id,
+        repetition_penalty= 2.0,
+        # device=-1,
+    )
+
+    llm2 = HuggingFacePipeline(pipeline=pipeline2)
+
+    # Map
+    map_template = """
+    Summarize the following text in a clear and concise way:
+    TITLE: `{title}`
+    TEXT:`{docs}`
+    Brief Summary:
+    """
+    map_prompt = PromptTemplate(template = map_template, 
+                                input_variables = ['title','docs'])
+    map_chain = LLMChain(llm=llm, prompt=map_prompt)
+
+    # Reduce - Collapse
+    reduce_template = """
+    The following is set of partial summaries of a video titled {title}:
+    partial summaries: {doc_summaries}
+    Take these and distill them into a consolidated summary. 
+    Summary:
+    """
+
+    reduce_prompt = PromptTemplate(template = reduce_template, 
+                                    input_variables = ['title','doc_summaries'])
+    reduce_chain = LLMChain(llm=llm, prompt=reduce_prompt)
+
+    # Takes a list of documents, combines them into a single string, and passes this to an LLMChain
+    collapse_documents_chain = StuffDocumentsChain(
+        llm_chain=reduce_chain, document_variable_name="doc_summaries"
+    )
+
+    # Final Reduce - Combine
+    final_reduce_template = """
+    The following is set of partial summaries of a video titled '{title}':
+    partial summaries: 
+
+    {doc_summaries}
+
+    Generate a summary of the whole text that includes `Video Subject`, and the `Key Highlights` as maximum 10 pullet points listing the main facts, arguments, or points:
+    """
+    final_reduce_prompt = PromptTemplate(template = final_reduce_template, 
+                                        input_variables = ['title','doc_summaries'])
+    final_reduce_chain = LLMChain(llm=llm2, prompt=final_reduce_prompt)
+
+    # Takes a list of documents, combines them into a single string, and passes this to an LLMChain
+    combine_documents_chain = StuffDocumentsChain(
+        llm_chain=final_reduce_chain, document_variable_name="doc_summaries"
+    )
+
+    # Combines and iteravely reduces the mapped documents
+    reduce_documents_chain = ReduceDocumentsChain(
+        # This is final chain that is called.
+        combine_documents_chain=combine_documents_chain,
+        # If documents exceed context for `StuffDocumentsChain`
+        collapse_documents_chain=collapse_documents_chain,
+        # The maximum number of tokens to group documents into.
+        token_max=500,
+    )
+
+    # Combining documents by mapping a chain over them, then combining results
+    map_reduce_chain = MapReduceDocumentsChain(
+        # Map chain
+        llm_chain=map_chain,
+        # Reduce chain
+        reduce_documents_chain=reduce_documents_chain,
+        # The variable name in the llm_chain to put the documents in
+        document_variable_name="docs",
+        # Return the results of the map steps in the output
+        return_intermediate_steps=False,
+    )
+
+    from langchain.document_loaders import TextLoader
+    from langchain.text_splitter import TokenTextSplitter
+    
+    with open('./transcript.txt','w') as f:
+        f.write(text)
+    loader = TextLoader("./transcript.txt")
+    doc = loader.load()
+    text_splitter = TokenTextSplitter(chunk_size=500, chunk_overlap=0)
+    docs = text_splitter.split_documents(doc)
+
+    summary = map_reduce_chain.run({'input_documents':docs, 'title':title})
+
+    del(llm)
+    del(llm2)
+    del(model)
+    del(tokenizer)
+    torch.cuda.empty_cache()
+
+    return summary
+
+import gradio as gr 
+import pytube
+from pytube import YouTube
+
+def get_youtube_title(url):
+    yt = YouTube(str(url))
+    return yt.title
+
+def get_video(url):
+    vid_id = pytube.extract.video_id(url)
+    embed_html = '<iframe width="100%" height="315" src="https://www.youtube.com/embed/{}" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>'.format(vid_id)
+    return embed_html
+
+def summarize_youtube_video(url,force_transcribe):
+    title,text = transcribe_youtube_video(url,force_transcribe)
+    Summary = summarize_text(title,text)
+    return Summary
+
+html = '<iframe width="100%" height="315" src="https://www.youtube.com/embed/" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>'
+
+with gr.Blocks() as demo:
+    # gr.Markdown("Transribe a YouTube video using this demo.")
+    with gr.Row():
+        with gr.Column(scale=3):
+            url = gr.Textbox(label="Enter YouTube video URL here:",placeholder="https://www.youtube.com/watch?v=")
+            force_transcribe = gr.Checkbox(label="Transcribe even if transcription is available.")
+        with gr.Column(scale=1):
+            gr.Markdown("# Summarize a YouTube video using this demo!",scale=2)
+            sum_btn = gr.Button("Summarize!",scale=1)
+    title = gr.Textbox(label="Video Title",placeholder="title...")
+    with gr.Row():
+        video = gr.HTML(html)
+        output = gr.Textbox(label="Summary",placeholder="summary...")
+    sum_btn.click(fn=get_youtube_title, inputs=url, outputs=title, api_name="get_youtube_title")
+    sum_btn.click(fn=summarize_youtube_video, inputs=[url,force_transcribe], outputs=output, api_name="summarize_youtube_video", queue=True)
+    sum_btn.click(fn=get_video, inputs=url, outputs=video, api_name="get_youtube_video",queue=False)
+
+demo.queue()    
+demo.launch(share=True)

requirements.txt

@@ -0,0 +1,282 @@
+# This file may be used to create an environment using:
+# $ conda create --name <env> --file <this file>
+# platform: win-64
+abseil-cpp=20211102.0=h0e60522_0
+accelerate=0.22.0=pypi_0
+aiofiles=23.2.1=pypi_0
+aiohttp=3.8.3=py38h2bbff1b_0
+aiosignal=1.3.1=pyhd8ed1ab_0
+altair=5.1.1=pypi_0
+anyio=3.7.1=pypi_0
+arrow-cpp=11.0.0=ha81ea56_2
+astroid=2.15.8=pypi_0
+asttokens=2.2.1=pyhd8ed1ab_0
+async-timeout=4.0.3=pyhd8ed1ab_0
+attrs=23.1.0=pyh71513ae_1
+audioread=3.0.1=pypi_0
+aws-c-common=0.6.8=h2bbff1b_1
+aws-c-event-stream=0.1.6=hd77b12b_6
+aws-checksums=0.1.11=h2bbff1b_2
+aws-sdk-cpp=1.8.185=hd77b12b_1
+backcall=0.2.0=pyh9f0ad1d_0
+backports=1.0=pyhd8ed1ab_3
+backports.functools_lru_cache=1.6.5=pyhd8ed1ab_0
+bitsandbytes=0.41.1=pypi_0
+blas=1.0=mkl
+boost-cpp=1.78.0=h5b4e17d_0
+bottleneck=1.3.5=py38h080aedc_0
+brotlipy=0.7.0=py38h2bbff1b_1003
+bzip2=1.0.8=h8ffe710_4
+c-ares=1.19.0=h2bbff1b_0
+ca-certificates=2023.7.22=h56e8100_0
+cachetools=5.3.1=pyhd8ed1ab_0
+certifi=2023.7.22=pyhd8ed1ab_0
+cffi=1.15.1=py38h2bbff1b_3
+charset-normalizer=2.0.4=pyhd3eb1b0_0
+click=8.1.7=win_pyh7428d3b_0
+colorama=0.4.6=pyhd8ed1ab_0
+comm=0.1.4=pyhd8ed1ab_0
+contourpy=1.1.1=pypi_0
+cryptography=41.0.2=py38hac1b9e3_0
+cuda-cccl=12.2.140=0
+cuda-cudart=11.8.89=0
+cuda-cudart-dev=11.8.89=0
+cuda-cupti=11.8.87=0
+cuda-libraries=11.8.0=0
+cuda-libraries-dev=11.8.0=0
+cuda-nvrtc=11.8.89=0
+cuda-nvrtc-dev=11.8.89=0
+cuda-nvtx=11.8.86=0
+cuda-profiler-api=12.2.140=0
+cuda-runtime=11.8.0=0
+cycler=0.12.0=pypi_0
+dataclasses=0.8=pyhc8e2a94_3
+dataclasses-json=0.5.7=pyhd8ed1ab_0
+datasets=2.14.4=pyhd8ed1ab_0
+debugpy=1.6.7=py38hd77b12b_0
+decorator=5.1.1=pyhd8ed1ab_0
+dill=0.3.7=pyhd8ed1ab_0
+einops=0.6.1=pypi_0
+exceptiongroup=1.1.3=pypi_0
+executing=1.2.0=pyhd8ed1ab_0
+faiss=1.7.4=py38h7a05997_0_cpu
+fastapi=0.103.2=pypi_0
+ffmpeg=4.3.1=ha925a31_0
+ffmpeg-python=0.2.0=pypi_0
+ffmpy=0.3.1=pypi_0
+filelock=3.9.0=py38haa95532_0
+fonttools=4.43.0=pypi_0
+freetype=2.12.1=ha860e81_0
+frozenlist=1.3.3=py38h2bbff1b_0
+fsspec=2023.6.0=pyh1a96a4e_0
+future=0.18.3=pypi_0
+gflags=2.2.2=ha925a31_1004
+giflib=5.2.1=h8cc25b3_3
+glog=0.5.0=h4797de2_0
+google-api-core=2.11.1=pyhd8ed1ab_0
+google-auth=2.22.0=pyh1a96a4e_0
+googleapis-common-protos=1.60.0=pyhd8ed1ab_0
+gpt4all=1.0.9=pypi_0
+gradio=3.45.2=pypi_0
+gradio-client=0.5.3=pypi_0
+greenlet=2.0.1=py38hd77b12b_0
+grpc-cpp=1.48.2=hfe90ff0_1
+h11=0.14.0=pypi_0
+httpcore=0.18.0=pypi_0
+httpx=0.25.0=pypi_0
+huggingface_hub=0.16.4=pyhd8ed1ab_0
+idna=3.4=py38haa95532_0
+importlib-metadata=6.8.0=pyha770c72_0
+importlib-resources=6.1.0=pypi_0
+importlib_metadata=6.8.0=hd8ed1ab_0
+intel-openmp=2023.1.0=h59b6b97_46319
+ipykernel=6.25.1=pyh6817e22_0
+ipython=8.12.0=pyh08f2357_0
+isort=5.12.0=pypi_0
+jedi=0.19.0=pyhd8ed1ab_0
+jinja2=3.1.2=py38haa95532_0
+joblib=1.3.2=pyhd8ed1ab_0
+jpeg=9e=h2bbff1b_1
+jsonschema=4.19.1=pypi_0
+jsonschema-specifications=2023.7.1=pypi_0
+jupyter_client=8.3.1=pyhd8ed1ab_0
+jupyter_core=5.3.1=py38haa244fe_0
+kiwisolver=1.4.5=pypi_0
+langchain=0.0.277=pyhd8ed1ab_0
+langsmith=0.0.27=pyhd8ed1ab_0
+lazy-loader=0.3=pypi_0
+lazy-object-proxy=1.9.0=pypi_0
+lerc=3.0=hd77b12b_0
+libblas=3.9.0=1_h8933c1f_netlib
+libbrotlicommon=1.0.9=h8ffe710_7
+libbrotlidec=1.0.9=h8ffe710_7
+libbrotlienc=1.0.9=h8ffe710_7
+libcublas=11.11.3.6=0
+libcublas-dev=11.11.3.6=0
+libcufft=10.9.0.58=0
+libcufft-dev=10.9.0.58=0
+libcurand=10.3.3.141=0
+libcurand-dev=10.3.3.141=0
+libcurl=8.1.1=h86230a5_0
+libcusolver=11.4.1.48=0
+libcusolver-dev=11.4.1.48=0
+libcusparse=11.7.5.86=0
+libcusparse-dev=11.7.5.86=0
+libdeflate=1.17=h2bbff1b_0
+libevent=2.1.10=h91dae50_4
+libfaiss=1.7.4=hba6d9cf_0_cpu
+libfaiss-avx2=1.7.4=h1234567_0_cpu
+libffi=3.4.4=hd77b12b_0
+liblapack=3.9.0=5_hd5c7e75_netlib
+libnpp=11.8.0.86=0
+libnpp-dev=11.8.0.86=0
+libnvjpeg=11.9.0.86=0
+libnvjpeg-dev=11.9.0.86=0
+libpng=1.6.39=h8cc25b3_0
+libprotobuf=3.20.3=h23ce68f_0
+librosa=0.10.1=pypi_0
+libsodium=1.0.18=h8d14728_1
+libssh2=1.10.0=he2ea4bf_2
+libthrift=0.15.0=h4364b78_2
+libtiff=4.5.0=h6c2663c_2
+libuv=1.44.2=h2bbff1b_0
+libwebp=1.2.4=hbc33d0d_1
+libwebp-base=1.2.4=h2bbff1b_1
+llvmlite=0.41.0=pypi_0
+lz4-c=1.9.4=h2bbff1b_0
+m2w64-gcc-libgfortran=5.3.0=6
+m2w64-gcc-libs=5.3.0=7
+m2w64-gcc-libs-core=5.3.0=7
+m2w64-gmp=6.1.0=2
+m2w64-libwinpthread-git=5.0.0.4634.697f757=2
+markupsafe=2.1.1=py38h2bbff1b_0
+marshmallow=3.20.1=pyhd8ed1ab_0
+marshmallow-enum=1.5.1=pyh9f0ad1d_3
+matplotlib=3.7.3=pypi_0
+matplotlib-inline=0.1.6=pyhd8ed1ab_0
+mccabe=0.7.0=pypi_0
+mkl=2023.1.0=h6b88ed4_46357
+mkl-service=2.4.0=py38h2bbff1b_1
+mkl_fft=1.3.6=py38hf11a4ad_1
+mkl_random=1.2.2=py38hf11a4ad_1
+mpmath=1.3.0=py38haa95532_0
+msgpack=1.0.7=pypi_0
+msys2-conda-epoch=20160418=1
+multidict=6.0.2=py38h2bbff1b_0
+multiprocess=0.70.15=py38haa95532_0
+mypy_extensions=1.0.0=pyha770c72_0
+nest-asyncio=1.5.6=pyhd8ed1ab_0
+networkx=2.8.8=pyhd8ed1ab_0
+nltk=3.8.1=pypi_0
+numba=0.58.0=pypi_0
+numexpr=2.8.4=py38h7b80656_1
+numpy=1.24.3=py38h79a8e48_1
+numpy-base=1.24.3=py38h8a87ada_1
+openapi-schema-pydantic=1.2.4=pyhd8ed1ab_0
+openssl=3.1.2=hcfcfb64_0
+orc=1.7.4=h623e30f_1
+orjson=3.9.7=pypi_0
+packaging=23.1=pyhd8ed1ab_0
+pandas=2.0.3=py38h4ed8f06_0
+parso=0.8.3=pyhd8ed1ab_0
+pickleshare=0.7.5=py_1003
+pillow=9.4.0=py38hd77b12b_0
+pip=23.2.1=py38haa95532_0
+pkgutil-resolve-name=1.3.10=pypi_0
+platformdirs=3.10.0=py38haa95532_0
+pooch=1.7.0=pypi_0
+prompt-toolkit=3.0.39=pyha770c72_0
+prompt_toolkit=3.0.39=hd8ed1ab_0
+protobuf=3.20.3=py38haa244fe_1
+psutil=5.9.0=py38h2bbff1b_0
+pure_eval=0.2.2=pyhd8ed1ab_0
+pyarrow=11.0.0=py38h790e06d_1
+pyasn1=0.4.8=py_0
+pyasn1-modules=0.2.7=py_0
+pycparser=2.21=pyhd3eb1b0_0
+pydantic=1.10.8=py38h2bbff1b_0
+pydub=0.25.1=pypi_0
+pygments=2.16.1=pyhd8ed1ab_0
+pylint=2.17.6=pypi_0
+pyopenssl=23.2.0=py38haa95532_0
+pyparsing=3.1.1=pypi_0
+pysocks=1.7.1=py38haa95532_0
+python=3.8.17=h1aa4202_0
+python-dateutil=2.8.2=pyhd8ed1ab_0
+python-multipart=0.0.6=pypi_0
+python-tzdata=2023.3=pyhd8ed1ab_0
+python-xxhash=2.0.2=py38h2bbff1b_1
+python_abi=3.8=2_cp38
+pytorch=2.0.1=py3.8_cuda11.8_cudnn8_0
+pytorch-cuda=11.8=h24eeafa_5
+pytorch-mutex=1.0=cuda
+pytube=15.0.0=pypi_0
+pytz=2023.3=pyhd8ed1ab_0
+pyu2f=0.1.5=pyhd8ed1ab_0
+pywin32=305=py38h2bbff1b_0
+pyyaml=6.0=py38h294d835_4
+pyzmq=23.2.1=py38h09162b1_0
+re2=2022.04.01=h0e60522_0
+referencing=0.30.2=pypi_0
+regex=2022.7.9=py38h2bbff1b_0
+requests=2.31.0=py38haa95532_0
+rpds-py=0.10.3=pypi_0
+rsa=4.9=pyhd8ed1ab_0
+sacremoses=0.0.53=pyhd8ed1ab_0
+safetensors=0.3.2=py38h062c2fa_0
+scikit-learn=1.3.0=pypi_0
+scipy=1.10.1=pypi_0
+semantic-version=2.10.0=pypi_0
+sentence-transformers=2.2.2=pypi_0
+sentencepiece=0.1.99=pypi_0
+setuptools=68.0.0=py38haa95532_0
+six=1.16.0=pyh6c4a22f_0
+snappy=1.1.9=h6c2663c_0
+sniffio=1.3.0=pypi_0
+soundfile=0.12.1=pypi_0
+soxr=0.3.6=pypi_0
+sqlalchemy=1.4.39=py38h2bbff1b_0
+sqlite=3.41.2=h2bbff1b_0
+stack_data=0.6.2=pyhd8ed1ab_0
+starlette=0.27.0=pypi_0
+stringcase=1.2.0=py_0
+sympy=1.12=pyh04b8f61_3
+tbb=2021.8.0=h59b6b97_0
+tenacity=8.2.3=pyhd8ed1ab_0
+threadpoolctl=3.2.0=pypi_0
+tiktoken=0.4.0=pypi_0
+tk=8.6.12=h2bbff1b_0
+tokenizers=0.13.2=py38h49fca51_1
+tomli=2.0.1=pypi_0
+tomlkit=0.12.1=pypi_0
+toolz=0.12.0=pypi_0
+torchaudio=2.0.2=pypi_0
+torchvision=0.15.2=pypi_0
+tornado=6.2=py38h294d835_0
+tqdm=4.66.1=pyhd8ed1ab_0
+traitlets=5.9.0=pyhd8ed1ab_0
+transformers=4.33.3=pypi_0
+typing-extensions=4.7.1=py38haa95532_0
+typing_extensions=4.7.1=py38haa95532_0
+typing_inspect=0.9.0=pyhd8ed1ab_0
+ucrt=10.0.22621.0=h57928b3_0
+urllib3=1.26.16=py38haa95532_0
+utf8proc=2.6.1=h2bbff1b_0
+uvicorn=0.23.2=pypi_0
+vc=14.2=h21ff451_1
+vc14_runtime=14.36.32532=hfdfe4a8_17
+vs2015_runtime=14.36.32532=h05e6639_17
+wcwidth=0.2.6=pyhd8ed1ab_0
+websockets=11.0.3=pypi_0
+wheel=0.38.4=py38haa95532_0
+win_inet_pton=1.1.0=py38haa95532_0
+wrapt=1.15.0=pypi_0
+xxhash=0.8.0=h2bbff1b_3
+xz=5.4.2=h8cc25b3_0
+yaml=0.2.5=h8ffe710_2
+yarl=1.7.2=py38h294d835_2
+youtube-transcript-api=0.6.1=pypi_0
+zeromq=4.3.4=h0e60522_1
+zipp=3.16.2=pyhd8ed1ab_0
+zlib=1.2.13=h8cc25b3_0
+zstd=1.5.5=hd43e919_0