causal_lm_cpp.yml

name: causal_lm_cpp
on:
  workflow_dispatch:
  pull_request:
  merge_group:
  push:
    branches:
      - master
      - 'releases/**'

permissions: read-all # Required by https://github.com/ossf/scorecard/blob/e23b8ad91fd6a64a0a971ca4fc0a4d1650725615/docs/checks.md#token-permissions

concurrency:
  # github.ref is not unique in post-commit
  group: ${{ github.event_name == 'push' && github.run_id || github.ref }}-causal-lm-cpp
  cancel-in-progress: true

env:
  l_ov_link: https://storage.openvinotoolkit.org/repositories/openvino/packages/nightly/2025.1.0-17911-83c047443de/l_openvino_toolkit_ubuntu20_2025.1.0.dev20250116_x86_64.tgz
  l_u22_ov_link: https://storage.openvinotoolkit.org/repositories/openvino/packages/nightly/2025.1.0-17911-83c047443de/l_openvino_toolkit_ubuntu22_2025.1.0.dev20250116_x86_64.tgz
  m_ov_link: https://storage.openvinotoolkit.org/repositories/openvino/packages/nightly/2025.1.0-17911-83c047443de/m_openvino_toolkit_macos_12_6_2025.1.0.dev20250116_x86_64.tgz
  w_ov_link: https://storage.openvinotoolkit.org/repositories/openvino/packages/nightly/2025.1.0-17911-83c047443de/w_openvino_toolkit_windows_2025.1.0.dev20250116_x86_64.zip

jobs:
  cpp-multinomial-greedy_causal_lm-ubuntu:
    runs-on: ubuntu-20.04-8-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.9
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model openlm-research/open_llama_3b_v2 open_llama_3b_v2
          optimum-cli export openvino -m TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
          wget https://huggingface.co/smangrul/tinyllama_lora_sql/resolve/main/adapter_model.safetensors?download=true -O adapter_model.safetensors
      - run: >
          . ./ov/setupvars.sh
          && timeout 35s ./build/samples/cpp/text_generation/multinomial_causal_lm ./open_llama_3b_v2/ a
        env:
          PYTHONPATH: "./build"
      - run: >
          . ./ov/setupvars.sh
          && timeout 35s ./samples/python/text_generation/multinomial_causal_lm.py ./open_llama_3b_v2/ b
        env:
          PYTHONPATH: "./build"
      - run: >
          . ./ov/setupvars.sh
          && timeout 35s ./build/samples/cpp/text_generation/greedy_causal_lm ./open_llama_3b_v2/ "return 0"
          | diff <(timeout 25s samples/python/text_generation/greedy_causal_lm.py ./open_llama_3b_v2/ "return 0") -
        env:
          PYTHONPATH: "./build"
      - run: >
          . ./ov/setupvars.sh
          && samples/python/text_generation/lora.py ./TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T/ adapter_model.safetensors "How to create a table with two columns, one of them has type float, another one has type int?"
        env:
          PYTHONPATH: "./build"

  cpp-beam_search_causal_lm-ubuntu:
    strategy:
      matrix:
        executable:
          [
            ./build/samples/cpp/text_generation/beam_search_causal_lm,
            python ./samples/python/text_generation/beam_search_causal_lm.py,
          ]
    runs-on: ubuntu-20.04
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: '3.10'
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model TinyLlama/TinyLlama-1.1B-Chat-v1.0 TinyLlama-1.1B-Chat-v1.0
      - name: Compare
        env:
          PYTHONPATH: "./build/" # C++ ignores that
        run: |
          source ./ov/setupvars.sh
          timeout 25s ${{ matrix.executable }} ./TinyLlama-1.1B-Chat-v1.0/ "Why is the Sun yellow?" > ./pred.txt
          python -c "
          import transformers
          with open('pred.txt', 'r') as file:
              predictions = file.read()
          tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
          prompt = 'Why is the Sun yellow?'
          if tokenizer.chat_template:
              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
          for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
              ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
              idx = predictions.find(ref)
              if -1 == idx:
                  raise RuntimeError(f'Missing "{ref=}" from predictions')
              predictions = predictions[:idx] + predictions[idx + len(ref):]
          "
          echo "Why is the Sun yellow?" passed

          timeout 25s ${{ matrix.executable }} ./TinyLlama-1.1B-Chat-v1.0/ 69 > ./pred.txt
          python -c "
          import transformers
          with open('pred.txt', 'r') as file:
              predictions = file.read()
          tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
          prompt = '69'
          if tokenizer.chat_template:
              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
          for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
              ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
              idx = predictions.find(ref)
              if -1 == idx:
                  raise RuntimeError(f'Missing "{ref=}" from predictions')
              predictions = predictions[:idx] + predictions[idx + len(ref):]
          "
          echo 69 passed

          timeout 25s ${{ matrix.executable }} ./TinyLlama-1.1B-Chat-v1.0/ Hi > ./pred.txt
          python -c "
          import transformers
          with open('pred.txt', 'r') as file:
              predictions = file.read()
          tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
          prompt = 'Hi'
          if tokenizer.chat_template:
            prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
          for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
              ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
              idx = predictions.find(ref)
              if -1 == idx:
                  raise RuntimeError(f'Missing "{ref=}" from predictions')
              predictions = predictions[:idx] + predictions[idx + len(ref):]
          "
          echo "Hi" passed

          timeout 25s ${{ matrix.executable }} ./TinyLlama-1.1B-Chat-v1.0/ "return 0" > ./pred.txt
          python -c "
          import transformers
          with open('pred.txt', 'r') as file:
              predictions = file.read()
          tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
          prompt = 'return 0'
          if tokenizer.chat_template:
              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
          for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
              ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
              idx = predictions.find(ref)
              if -1 == idx:
                  raise RuntimeError(f'Missing "{ref=}" from predictions')
              predictions = predictions[:idx] + predictions[idx + len(ref):]
          "
          echo "return 0" passed

          timeout 25s ${{ matrix.executable }} ./TinyLlama-1.1B-Chat-v1.0/ "你好！ 你好嗎？" > ./pred.txt
          python -c "
          import transformers
          with open('pred.txt', 'r', errors='ignore') as file:
              predictions = file.read()
          tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
          prompt = '你好！ 你好嗎？'
          if tokenizer.chat_template:
              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
          for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
              ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
              idx = predictions.find(ref.replace('�', ''))
              if -1 == idx:
                  raise RuntimeError(f'Missing "{ref=}" from predictions')
              predictions = predictions[:idx] + predictions[idx + len(ref):]
          "
          echo "你好！ 你好嗎？" passed

          timeout 1m ${{ matrix.executable }} ./TinyLlama-1.1B-Chat-v1.0/ "Why is the Sun yellow?" "return 0" "你好！ 你好嗎？" > ./pred.txt
          python -c "
          import transformers
          with open('pred.txt', 'r', errors='ignore') as file:
              predictions = file.read()
          tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0')
          prompts = [
            'Why is the Sun yellow?',
            'return 0',
            '你好！ 你好嗎？'
          ]
          for prompt in prompts:
            if tokenizer.chat_template:
                prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
            tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
            for beam in transformers.LlamaForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0').generate(**tokenized, num_beam_groups=3, num_beams=15, num_return_sequences=15, diversity_penalty=1.0, max_new_tokens=20, early_stopping=False, length_penalty=1.0, no_repeat_ngram_size=9**9, do_sample=False):
                ref = ': ' + tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True)
                idx = predictions.find(ref.replace('�', ''))
                if -1 == idx:
                    raise RuntimeError(f'Missing "{ref=}" from predictions')
                predictions = predictions[:idx] + predictions[idx + len(ref):]
          "
          echo "Multi prompt" passed

  cpp-greedy_causal_lm-windows:
    runs-on: windows-latest
    env:
      PYTHONIOENCODING: "utf8"
    defaults:
      run:
        shell: cmd
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.9
      - run: curl --output ov.zip ${{ env.w_ov_link }}
      - run: unzip -d ov ov.zip
      - run: dirs=(ov/*) && mv ov/*/* ov && rmdir "${dirs[@]}"
        shell: bash
      - name: Build app
        run: |
          call .\ov\setupvars.bat
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert model
        run: |
          call .\ov\setupvars.bat
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model TinyLlama/TinyLlama-1.1B-Chat-v1.0 TinyLlama-1.1B-Chat-v1.0
          optimum-cli export openvino -m TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T TinyLlama/TinyLlama-1.1B-intermediate-step-1431k-3T
          curl -o adapter_model.safetensors -s -L https://huggingface.co/smangrul/tinyllama_lora_sql/resolve/main/adapter_model.safetensors?download=true
      - run: >
          set PATH=.\build\openvino_genai\;%PATH%
          && call .\ov\setupvars.bat
          && .\build\samples\cpp\text_generation\Release\greedy_causal_lm.exe .\TinyLlama-1.1B-Chat-v1.0\ 69 > .\cpp.txt
      - run: |
          echo import transformers > ref.py
          echo predictions = open('cpp.txt', 'r').read() >> ref.py
          echo tokenizer = transformers.AutoTokenizer.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0', trust_remote_code=True) >> ref.py
          echo prompt = '69' >> ref.py
          echo if tokenizer.chat_template: >> ref.py
          echo     prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True) >> ref.py
          echo tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False) >> ref.py
          echo for beam in transformers.AutoModelForCausalLM.from_pretrained('TinyLlama/TinyLlama-1.1B-Chat-v1.0', trust_remote_code=True).generate(**tokenized, max_new_tokens=100, do_sample=False): >> ref.py
          echo     ref = tokenizer.decode(beam[tokenized['input_ids'].numel():], skip_special_tokens=True) >> ref.py
          echo     idx = predictions.find(ref) >> ref.py
          echo     if -1 == idx: >> ref.py
          echo         raise RuntimeError(f'Missing "{ref=}" from predictions') >> ref.py
          echo     predictions = predictions[:idx] + predictions[idx + len(ref):] >> ref.py
      - run: python ref.py
      - run: >
          set PATH=.\build\openvino_genai\;%PATH%
          && set "PYTHONPATH=./build/"
          && call .\ov\setupvars.bat
          && python samples\python\text_generation\greedy_causal_lm.py .\TinyLlama-1.1B-Chat-v1.0\ 69 > .\py.txt
      - run: fc .\cpp.txt .\py.txt
      - run: >
          set PATH=.\build\openvino_genai\;%PATH%
          && set "PYTHONPATH=./build/"
          && call .\ov\setupvars.bat
          && python samples\python\text_generation\lora.py .\TinyLlama\TinyLlama-1.1B-intermediate-step-1431k-3T\ adapter_model.safetensors "How to create a table with two columns, one of them has type float, another one has type int?"

  cpp-greedy_causal_lm-Qwen-7B-Chat:
    runs-on: ubuntu-20.04-16-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.11
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model Qwen/Qwen-7B-Chat Qwen-7B-Chat
      - run: >
          . ./ov/setupvars.sh
          && timeout 2m ./build/samples/cpp/text_generation/greedy_causal_lm ./Qwen-7B-Chat/ 69 | diff <(timeout 2m samples/python/text_generation/greedy_causal_lm.py ./Qwen-7B-Chat/ 69) -
        env:
          PYTHONPATH: "./build"

  cpp-beam_search_causal_lm-Qwen1_5-7B-Chat:
    runs-on: ubuntu-20.04-16-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.12
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model Qwen/Qwen1.5-7B-Chat Qwen1.5-7B-Chat
      - run: >
          . ./ov/setupvars.sh
          && timeout 50s ./build/samples/cpp/text_generation/beam_search_causal_lm ./Qwen1.5-7B-Chat/ "你好！"
          | diff <(timeout 50s ./samples/python/text_generation/beam_search_causal_lm.py ./Qwen1.5-7B-Chat/ "你好！") -
        env:
          PYTHONPATH: "./build"

  cpp-beam_search_causal_lm-Phi-2:
    runs-on: ubuntu-20.04-16-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.9
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model microsoft/phi-2 phi-2
      - run: >
          . ./ov/setupvars.sh
          && timeout 50s ./build/samples/cpp/text_generation/beam_search_causal_lm ./phi-2/ 69
          | diff <(timeout 50s ./samples/python/text_generation/beam_search_causal_lm.py ./phi-2/ 69) -
        env:
          PYTHONPATH: "./build"

  cpp-beam_search_causal_lm-notus-7b-v1:
    runs-on: ubuntu-20.04-16-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: '3.10'
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model argilla/notus-7b-v1 notus-7b-v1
      - run: >
          . ./ov/setupvars.sh
          && timeout 50s ./build/samples/cpp/text_generation/beam_search_causal_lm ./notus-7b-v1/ 69
          | diff <(timeout 50s ./samples/python/text_generation/beam_search_causal_lm.py ./notus-7b-v1/ 69) -
        env:
          PYTHONPATH: "./build"

  cpp-speculative_decoding_lm-ubuntu:
    runs-on: ubuntu-20.04-16-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.11
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model databricks/dolly-v2-3b dolly-v2-3b
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model databricks/dolly-v2-7b dolly-v2-7b
      - name: run and compare
        run: |
          source ./ov/setupvars.sh
          echo Running speculative_decoding_lm C++ sample...
          ./build/samples/cpp/text_generation/speculative_decoding_lm ./dolly-v2-7b/ ./dolly-v2-3b/ "Alan Turing was a" > predictions_speculative.txt
          echo Running greedy_causal_lm C++ sample...
          ./build/samples/cpp/text_generation/greedy_causal_lm ./dolly-v2-7b/ "Alan Turing was a" > predictions_greedy.txt
          echo Running speculative_decoding_lm Python sample...
          python ./samples/python/text_generation/speculative_decoding_lm.py ./dolly-v2-7b/ ./dolly-v2-3b/ "Alan Turing was a" > predictions_py.txt
          echo All samples executed, checking result correctness...
          python -c "
          with open('predictions_greedy.txt', 'r') as f:
              predicted_greedy = f.readline()
          with open('predictions_speculative.txt', 'r') as f:
              predicted_speculative = f.readline()
          with open('predictions_py.txt', 'r') as f:
              predicted_py = f.readline()
          print(f'Predicted greedy: {predicted_greedy}')
          print(f'Predicted speculative: {predicted_speculative}')
          assert predicted_greedy == predicted_speculative
          assert predicted_greedy == predicted_py
          assert predicted_speculative == predicted_py
          "
          echo "Alan Turing was a" passed
        env:
          PYTHONPATH: "./build/:$PYTHONPATH"
          LD_LIBRARY_PATH: "./build/openvino_genai/:$LD_LIBRARY_PATH"

  cpp-prompt_lookup_decoding_lm-ubuntu:
    runs-on: ubuntu-20.04-16-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.12
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model TinyLlama/TinyLlama-1.1B-Chat-v1.0 TinyLlama-1.1B-Chat-v1.0
      - name: run and compare
        run: |
          source ./ov/setupvars.sh

          echo 'Code:```python
          def add(a, b):
              return a + b
          ```
          Question: Can you please add 2 and 3
          A:' > ./prompt.txt
          echo Running prompt_lookup_decoding_lm C++ sample...
          ./build/samples/cpp/text_generation/prompt_lookup_decoding_lm ./TinyLlama-1.1B-Chat-v1.0/ "$(<prompt.txt)" > predictions_prompt_lookup.txt
          echo Running greedy_causal_lm C++ sample...
          ./build/samples/cpp/text_generation/greedy_causal_lm ./TinyLlama-1.1B-Chat-v1.0/ "$(<prompt.txt)" > predictions_greedy.txt
          echo Running prompt_lookup_decoding_lm Python sample...
          python ./samples/python/text_generation/prompt_lookup_decoding_lm.py ./TinyLlama-1.1B-Chat-v1.0/ "$(<prompt.txt)" > predictions_py.txt
          echo All samples executed, checking result correctness...
          python -c "
          with open('predictions_greedy.txt', 'r') as f:
              predicted_greedy = f.readline()
          with open('predictions_prompt_lookup.txt', 'r') as f:
              predicted_prompt_lookup = f.readline()
          with open('predictions_py.txt', 'r') as f:
              predicted_prompt_lookup_py = f.readline()

          print(f'Predicted greedy: {predicted_greedy}')
          print(f'Predicted prompt lookup: {predicted_prompt_lookup}')
          assert predicted_greedy == predicted_prompt_lookup
          assert predicted_greedy == predicted_prompt_lookup_py
          assert predicted_prompt_lookup == predicted_prompt_lookup_py
          "
          echo "Prompt lookup" passed
        env:
          PYTHONPATH: "./build/:$PYTHONPATH"
          LD_LIBRARY_PATH: "./build/openvino_genai/:$LD_LIBRARY_PATH"

  cpp-Phi-1_5:
    runs-on: ubuntu-20.04-16-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.9
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model microsoft/phi-1_5 phi-1_5
      - name: Run Generation
        run: |
          source ./ov/setupvars.sh
          timeout 50s ./build/samples/cpp/text_generation/greedy_causal_lm ./phi-1_5/ "Alan Turing was a" > ./pred_greedy.txt
      - name: Compare
        run: |
          python -c "
          import transformers
          with open('pred_greedy.txt', 'r') as file:
              predictions = file.read()
          tokenizer = transformers.AutoTokenizer.from_pretrained('microsoft/phi-1_5')
          prompt = 'Alan Turing was a'
          if tokenizer.chat_template:
              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
          for output in transformers.AutoModelForCausalLM.from_pretrained('microsoft/phi-1_5').generate(**tokenized, max_length=100, do_sample=False):
              ref = tokenizer.decode(output[tokenized['input_ids'].numel():], skip_special_tokens=True)
              idx = predictions.find(ref)
              if -1 == idx:
                  raise RuntimeError(f'Missing "{ref=}" from predictions')
              predictions = predictions[:idx] + predictions[idx + len(ref):]
          "
          echo Phi-1_5 passed
      - run: >
          . ./ov/setupvars.sh
          && timeout 50s samples/python/text_generation/greedy_causal_lm.py ./phi-1_5/ "Alan Turing was a"
          | diff ./pred_greedy.txt -
        env:
          PYTHONPATH: "./build"

  cpp-greedy_causal_lm-redpajama-3b-chat:
    runs-on: ubuntu-20.04-8-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: '3.10'
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model ikala/redpajama-3b-chat redpajama-3b-chat
      - name: Run Generation
        run: |
          source ./ov/setupvars.sh
          timeout 50s ./build/samples/cpp/text_generation/greedy_causal_lm ./redpajama-3b-chat/ "Alan Turing was a" > ./pred_greedy.txt
      - name: Compare
        run: |
          python -c "
          import transformers
          with open('pred_greedy.txt', 'r') as file:
              predictions = file.read()
          tokenizer = transformers.AutoTokenizer.from_pretrained('ikala/redpajama-3b-chat')
          prompt = 'Alan Turing was a'
          if tokenizer.chat_template:
              prompt = tokenizer.apply_chat_template([{'role': 'user', 'content': prompt}], tokenize=False, add_generation_prompt=True)
          tokenized = tokenizer(prompt, return_tensors='pt', add_special_tokens=False)
          for output in transformers.AutoModelForCausalLM.from_pretrained('ikala/redpajama-3b-chat').generate(**tokenized, max_length=100, do_sample=False):
              ref = tokenizer.decode(output[tokenized['input_ids'].numel():], skip_special_tokens=True)
              idx = predictions.find(ref)
              if -1 == idx:
                  raise RuntimeError(f'Missing "{ref}" from predictions')
              predictions = predictions[:idx] + predictions[idx + len(ref):]
          "
          echo "Alan Turing was a" passed
      - run: >
          . ./ov/setupvars.sh
          && timeout 50s samples/python/text_generation/greedy_causal_lm.py ./redpajama-3b-chat/ "Alan Turing was a"
          | diff ./pred_greedy.txt -
        env:
          PYTHONPATH: "./build"

  cpp-chat_sample-ubuntu:
    runs-on: ubuntu-24.04
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.11
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model TinyLlama/TinyLlama-1.1B-Chat-v1.0 TinyLlama-1.1B-Chat-v1.0
      - name: Compare
        env:
          PYTHONPATH: "./build"
        run: |
          source ./ov/setupvars.sh
          printf 'What is 2 + 2?\nWhat is the previous answer?\nAdd 1 to it.\nSubtract 5 from it.\nWhy is the sun yellow?\nWhat was my first question?\n' > ./input.txt
          timeout 30s ./build/samples/cpp/text_generation/chat_sample ./TinyLlama-1.1B-Chat-v1.0/ < input.txt > ./pred.txt
          python -c "
          from transformers import AutoTokenizer, AutoModelForCausalLM
          model_id = 'TinyLlama/TinyLlama-1.1B-Chat-v1.0'
          tokenizer = AutoTokenizer.from_pretrained(model_id)
          model = AutoModelForCausalLM.from_pretrained(model_id)
          prompts = ['What is 2 + 2?', 'What is the previous answer?', 'Add 1 to it.', 'Subtract 5 from it.', 'Why is the sun yellow?', 'What was my first question?']
          def gen_prompt(prompt):
              return {'role': 'user', 'content': prompt}
          def gen_answer(answer):
              return {'role': 'assistant', 'content': answer}
          chat_history = []
          chat_prompt = ''
          output = open('ref.txt', 'w')
          for prompt in prompts:
              output.write('question:\n')
              chat_history.append(gen_prompt(prompt))
              chat_prompt = tokenizer.apply_chat_template(chat_history, tokenize=False, add_generation_prompt=True)
              tokenized = tokenizer(chat_prompt, return_tensors='pt', add_special_tokens=False)
              answer = model.generate(**tokenized, max_length=1000, do_sample=False)
              answer_str = tokenizer.decode(answer[0, tokenized['input_ids'].numel():], skip_special_tokens=True)
              chat_history.append(gen_answer(answer_str))
              output.write(answer_str)
              output.write('\n----------\n')
          output.write('question:\n')
          output.close()
          "
          diff pred.txt ref.txt
          echo "Chat sample cpp" passed
          timeout 30s ./samples/python/text_generation/chat_sample.py ./TinyLlama-1.1B-Chat-v1.0/ < input.txt > ./pred2.txt
          diff pred2.txt ref.txt
          echo "Chat sample python" passed

  visual_language_chat_sample-ubuntu-minicpm_v2_6:
    runs-on: ubuntu-22.04-16-cores
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.11
      - uses: ./.github/actions/install_openvino
        with:
          ov_link: ${{ env.l_u22_ov_link }}
      - uses: ./.github/actions/build_app
        with:
          build_target: 'visual_language_chat benchmark_vlm py_openvino_genai'
      - uses: ./.github/actions/install_python_deps
      - name: Download and convert tiny-random-minicpmv-2_6 model and an image
        run: |
          source ./ov/setupvars.sh
          optimum-cli export openvino -m katuni4ka/tiny-random-minicpmv-2_6 tiny-random-minicpmv-2_6 --trust-remote-code --task image-text-to-text
          mkdir images
      - name: Generate images - tiny-random-minicpmv-2_6
        shell: python
        run: |
          from PIL import Image
          import numpy as np
          import requests
          res = 28, 28
          lines = np.arange(res[0] * res[1] * 3, dtype=np.uint8) % 255
          lines = lines.reshape([*res, 3])
          lines_image = Image.fromarray(lines)
          lines_image.save("images/lines.png")
          cat = Image.open(requests.get("https://github.com/openvinotoolkit/openvino_notebooks/assets/29454499/d5fbbd1a-d484-415c-88cb-9986625b7b11", stream=True).raw).convert('RGB')
          cat.save("images/cat.png")
      - name: Run visual_language_chat C++ sample - tiny-random-minicpmv-2_6
        run: >
          set -o pipefail
          && source ./ov/setupvars.sh
          && ./build/samples/cpp/visual_language_chat/visual_language_chat ./tiny-random-minicpmv-2_6/ ./images/
          <<< $'Describe the images?' | tee cpp.txt
        timeout-minutes: 2
      - name: Run benchmark_vlm C++ sample - tiny-random-minicpmv-2_6
        run: >
          set -o pipefail
          && source ./ov/setupvars.sh
          && ./build/samples/cpp/visual_language_chat/benchmark_vlm  -m ./tiny-random-minicpmv-2_6/ -i ./images/cat.png -n 3
        timeout-minutes: 2
      - name: Run visual_language_chat Python sample - tiny-random-minicpmv-2_6
        run: >
          set -o pipefail
          && source ./ov/setupvars.sh
          && ./samples/python/visual_language_chat/visual_language_chat.py ./tiny-random-minicpmv-2_6/ ./images/
          <<< $'Describe the images?' | tee py.txt
        env:
          PYTHONPATH: "./build/"
      - name: Run benchmark_vlm Python sample - tiny-random-minicpmv-2_6
        run: >
          set -o pipefail
          && source ./ov/setupvars.sh
          && ./samples/python/visual_language_chat/benchmark_vlm.py -m ./tiny-random-minicpmv-2_6/ -i ./images/cat.png -n 3
        env:
          PYTHONPATH: "./build/"
      - name: Encode cpp.txt with Python encoding instead of terminal one
        shell: python
        run: |
          with open("cpp.txt", "rb") as f:
            content = f.read().decode("utf-8", "replace")
          with open("cpp.txt", "wb") as f:
            f.write(content.encode("utf-8"))
      - run: diff cpp.txt py.txt
      - name: Run visual_language_chat C++ sample with 2 prompts - tiny-random-minicpmv-2_6
        run: >
          set -o pipefail
          && source ./ov/setupvars.sh
          && ./build/samples/cpp/visual_language_chat/visual_language_chat ./tiny-random-minicpmv-2_6/ ./images/cat.png
          <<< $'What is unusual on this image?\nGo on.' | tee cpp2.txt
        timeout-minutes: 2
      - name: Run visual_language_chat Python sample with 2 prompts - tiny-random-minicpmv-2_6
        run: >
          set -o pipefail
          && source ./ov/setupvars.sh
          && ./samples/python/visual_language_chat/visual_language_chat.py ./tiny-random-minicpmv-2_6/ ./images/cat.png
          <<< $'What is unusual on this image?\nGo on.' | tee py2.txt
        env:
          PYTHONPATH: "./build/"
      - name: Encode cpp2.txt with Python encoding instead of terminal one
        shell: python
        run: |
          with open("cpp2.txt", "rb") as f:
            content = f.read().decode("utf-8", "replace")
          with open("cpp2.txt", "wb") as f:
            f.write(content.encode("utf-8"))
      - run: diff cpp2.txt py2.txt

  visual_language_chat_sample-ubuntu-llava_1_5:
    uses: ./.github/workflows/job_vlm_sample_llava.yml
    with:
      model_id: llava-hf/llava-1.5-7b-hf
      model_dir: llava_1_5_7b_ov

  visual_language_chat_sample-ubuntu-llava_next:
    uses: ./.github/workflows/job_vlm_sample_llava.yml
    with:
      model_id: llava-hf/llava-v1.6-mistral-7b-hf
      model_dir: llava_v1_6_mistral_7b_ov

  visual_language_chat_sample-ubuntu-internvl2:
    runs-on: ubuntu-22.04-16-cores
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.11
      - uses: ./.github/actions/install_openvino
        with:
          ov_link: ${{ env.l_u22_ov_link }}
      - uses: ./.github/actions/build_app
        with:
          build_target: 'visual_language_chat py_openvino_genai'
      - uses: ./.github/actions/install_python_deps
      - name: Download and convert InternVL2 model
        run: |
          # Lowering transformers version, workaround for https://huggingface.co/OpenGVLab/InternVL2-1B/discussions/7
          python -m pip install -U "transformers<4.45.0"
          source ./ov/setupvars.sh
          optimum-cli export openvino --model OpenGVLab/InternVL2-4B ./internvl2_4b_ov/ --trust-remote-code
      - name: Download images
        run: |
          wget https://llava-vl.github.io/static/images/monalisa.jpg
      - name: Run visual_language_chat C++ sample - InternVL2
        run: >
          source ./ov/setupvars.sh
          && ./build/samples/cpp/visual_language_chat/visual_language_chat ./internvl2_4b_ov/ monalisa.jpg
          <<< $'Who drew this painting?\nWhen did the painter live?'
        timeout-minutes: 4

  visual_language_chat_sample-ubuntu-qwen2vl:
    runs-on: ubuntu-22.04-16-cores
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.11
      - uses: ./.github/actions/install_openvino
        with:
          ov_link: ${{ env.l_u22_ov_link }}
      - uses: ./.github/actions/build_app
        with:
          build_target: 'visual_language_chat py_openvino_genai'
      - uses: ./.github/actions/install_python_deps
      - name: Download and convert Qwen2VL model
        run: |
          source ./ov/setupvars.sh
          optimum-cli export openvino --model Qwen/Qwen2-VL-2B-Instruct ./qwen2_vl_2b_ov/ --trust-remote-code
      - name: Download images
        run: |
          wget https://llava-vl.github.io/static/images/monalisa.jpg
      - name: Run visual_language_chat C++ sample - Qwen2VL
        run: >
          source ./ov/setupvars.sh
          && ./build/samples/cpp/visual_language_chat/visual_language_chat ./qwen2_vl_2b_ov/ monalisa.jpg
          <<< $'Who drew this painting?\nWhen did the painter live?'
        timeout-minutes: 4

  cpp-continuous-batching-ubuntu:
    runs-on: ubuntu-20.04-8-cores
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.12
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.l_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          sudo ./ov/install_dependencies/install_openvino_dependencies.sh
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model TinyLlama/TinyLlama-1.1B-Chat-v1.0 TinyLlama-1.1B-Chat-v1.0
      - name: Run gtests
        run: |
          source ./ov/setupvars.sh
          ./build/tests/cpp/tests_continuous_batching
      - name: Run accuracy_sample
        run: |
          source ./ov/setupvars.sh
          timeout --verbose 50s ./build/tools/continuous_batching/accuracy/continuous_batching_accuracy -m ./TinyLlama-1.1B-Chat-v1.0/ -n 5
      - name: Run throughput_benchmark
        run: |
          wget -q https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
          source ./ov/setupvars.sh
          timeout --verbose 200s ./build/tools/continuous_batching/benchmark/continuous_batching_benchmark -n 10 -m ./TinyLlama-1.1B-Chat-v1.0/ --dataset ./ShareGPT_V3_unfiltered_cleaned_split.json --cache_size 1
          timeout --verbose 200s ./build/tools/continuous_batching/benchmark/continuous_batching_benchmark -n 10 --dynamic_split_fuse --max_batch_size 256 --max_input_len 256 -m ./TinyLlama-1.1B-Chat-v1.0/ --dataset ./ShareGPT_V3_unfiltered_cleaned_split.json --cache_size 1

  cpp-continuous-batching-windows:
    runs-on: windows-latest
    env:
      PYTHONIOENCODING: "utf8"
    defaults:
      run:
        shell: cmd
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.9
      - name: Install OpenVINO
        run: |
          curl --output ov.zip ${{ env.w_ov_link }}
          unzip -d ov ov.zip
          dirs=(ov/*) && mv ov/*/* ov && rmdir "${dirs[@]}"
        shell: bash
      - name: Build app
        run: |
          call .\ov\setupvars.bat
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          call .\ov\setupvars.bat
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model TinyLlama/TinyLlama-1.1B-Chat-v1.0 TinyLlama-1.1B-Chat-v1.0
      - name: Run gtests
        run: |
          set PATH=.\build\openvino_genai\;%PATH%
          call .\ov\setupvars.bat
          .\build\tests\cpp\Release\tests_continuous_batching.exe
      - name: Run accuracy_sample
        run: |
          set PATH=.\build\openvino_genai\;%PATH%
          call .\ov\setupvars.bat
          .\build\tools\continuous_batching\accuracy\Release\continuous_batching_accuracy.exe -m .\TinyLlama-1.1B-Chat-v1.0\ -n 5
      - name: Run throughput_benchmark
        run: |
          curl -o .\ShareGPT_V3_unfiltered_cleaned_split.json -s -L "https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json"
          set PATH=.\build\openvino_genai\;%PATH%
          call .\ov\setupvars.bat
          .\build\tools\continuous_batching\benchmark\Release\continuous_batching_benchmark.exe -n 2 -m .\TinyLlama-1.1B-Chat-v1.0\ --dataset .\ShareGPT_V3_unfiltered_cleaned_split.json --cache_size 1

  cpp-continuous-batching-macos:
    runs-on: macos-13
    defaults:
      run:
        shell: bash
    steps:
      - uses: actions/checkout@v4
        with:
          submodules: recursive
      - uses: actions/setup-python@v4
        with:
          python-version: 3.9
      - name: Install OpenVINO
        run: |
          mkdir ./ov/
          curl ${{ env.m_ov_link }} | tar --directory ./ov/ --strip-components 1 -xz
          brew install coreutils scons
      - name: Build app
        run: |
          source ./ov/setupvars.sh
          cmake -DCMAKE_BUILD_TYPE=Release -S ./ -B ./build/
          cmake --build ./build/ --config Release -j
      - name: Download and convert and model
        run: |
          source ./ov/setupvars.sh
          python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/nightly
          python -m pip install -r ./samples/requirements.txt
          optimum-cli export openvino --trust-remote-code --weight-format fp16 --model TinyLlama/TinyLlama-1.1B-Chat-v1.0 TinyLlama-1.1B-Chat-v1.0
      - name: Run gtests
        run: |
          source ./ov/setupvars.sh
          ./build/tests/cpp/tests_continuous_batching
      - name: Run accuracy_sample
        run: |
          source ./ov/setupvars.sh
          timeout --verbose 120s ./build/tools/continuous_batching/accuracy/continuous_batching_accuracy -m ./TinyLlama-1.1B-Chat-v1.0/ -n 5
      - name: Run throughput_benchmark
        run: |
          wget -q https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
          source ./ov/setupvars.sh
          ./build/tools/continuous_batching/benchmark/continuous_batching_benchmark -n 5 -m ./TinyLlama-1.1B-Chat-v1.0/ --dataset ./ShareGPT_V3_unfiltered_cleaned_split.json --cache_size 1

  Overall_Status:
    name: ci/gha_overall_status_causal_lm
    needs: [cpp-multinomial-greedy_causal_lm-ubuntu, cpp-beam_search_causal_lm-ubuntu, cpp-greedy_causal_lm-windows,
            cpp-greedy_causal_lm-Qwen-7B-Chat, cpp-beam_search_causal_lm-Qwen1_5-7B-Chat, cpp-beam_search_causal_lm-Phi-2,
            cpp-beam_search_causal_lm-notus-7b-v1, cpp-speculative_decoding_lm-ubuntu, cpp-prompt_lookup_decoding_lm-ubuntu,
            cpp-Phi-1_5, cpp-greedy_causal_lm-redpajama-3b-chat, cpp-chat_sample-ubuntu, cpp-continuous-batching-ubuntu,
            visual_language_chat_sample-ubuntu-minicpm_v2_6, visual_language_chat_sample-ubuntu-llava_1_5, visual_language_chat_sample-ubuntu-llava_next, visual_language_chat_sample-ubuntu-internvl2,
            cpp-continuous-batching-windows, cpp-continuous-batching-macos]
    if: ${{ always() }}
    runs-on: ubuntu-latest
    steps:
      - name: Check status of all jobs
        if: >-
          ${{
            contains(needs.*.result, 'failure') ||
            contains(needs.*.result, 'cancelled')
          }}
        run: exit 1