Python bindings for https://github.com/openvinotoolkit/openvino.genai

These details have not been verified by PyPI

Project description

OpenVINO™ GenAI Library

OpenVINO™ GenAI is a flavor of OpenVINO™, aiming to simplify running inference of generative AI models. It hides the complexity of the generation process and minimizes the amount of code required.

Install OpenVINO™ GenAI

The OpenVINO™ GenAI flavor is available for installation via Archive and PyPI distributions. To install OpenVINO™ GenAI, refer to the Install Guide.

To build OpenVINO™ GenAI library from source, refer to the Build Instructions.

Usage

Prerequisites

Installed OpenVINO™ GenAI

If OpenVINO GenAI is installed via archive distribution or built from source, you will need to install additional python dependencies (e.g. optimum-cli for simplified model downloading and exporting):

# (Optional) Clone OpenVINO GenAI repository if it does not exist
git clone --recursive https://github.com/openvinotoolkit/openvino.genai.git
cd openvino.genai
# Install python dependencies
python -m pip install ./thirdparty/openvino_tokenizers/[transformers] --extra-index-url https://storage.openvinotoolkit.org/simple/wheels/pre-release
python -m pip install --upgrade-strategy eager -r ./samples/cpp/requirements.txt

A model in OpenVINO IR format

Download and convert a model with optimum-cli:

optimum-cli export openvino --model "TinyLlama/TinyLlama-1.1B-Chat-v1.0" --trust-remote-code "TinyLlama-1.1B-Chat-v1.0"

LLMPipeline is the main object used for decoding. You can construct it straight away from the folder with the converted model. It will automatically load the main model, tokenizer, detokenizer and default generation configuration.

Python

A simple example:

import openvino_genai as ov_genai
pipe = ov_genai.LLMPipeline(model_path, "CPU")
print(pipe.generate("The Sun is yellow because"))

Calling generate with custom generation config parameters, e.g. config for grouped beam search:

import openvino_genai as ov_genai
pipe = ov_genai.LLMPipeline(model_path, "CPU")

result = pipe.generate("The Sun is yellow because", max_new_tokens=30, num_groups=3, group_size=5, diversity_penalty=1.5)
print(result)

output:

'it is made up of carbon atoms. The carbon atoms are arranged in a linear pattern, which gives the yellow color. The arrangement of carbon atoms in'

A simple chat in Python:

import openvino_genai as ov_genai
pipe = ov_genai.LLMPipeline(model_path)

config = {'max_new_tokens': 100, 'num_groups': 3, 'group_size': 5, 'diversity_penalty': 1.5}
pipe.set_generation_config(config)

pipe.start_chat()
while True:
    print('question:')
    prompt = input()
    if prompt == 'Stop!':
        break
    print(pipe(prompt))
pipe.finish_chat()

Test to compare with Huggingface outputs

C++

A simple example:

#include "openvino/genai/llm_pipeline.hpp"
#include <iostream>

int main(int argc, char* argv[]) {
    std::string model_path = argv[1];
    ov::genai::LLMPipeline pipe(model_path, "CPU");
    std::cout << pipe.generate("The Sun is yellow because");
}

Using group beam search decoding:

#include "openvino/genai/llm_pipeline.hpp"
#include <iostream>

int main(int argc, char* argv[]) {
    std::string model_path = argv[1];
    ov::genai::LLMPipeline pipe(model_path, "CPU");

    ov::genai::GenerationConfig config;
    config.max_new_tokens = 256;
    config.num_groups = 3;
    config.group_size = 5;
    config.diversity_penalty = 1.0f;

    std::cout << pipe.generate("The Sun is yellow because", config);
}

A simple chat in C++ using grouped beam search decoding:

#include "openvino/genai/llm_pipeline.hpp"
#include <iostream>

int main(int argc, char* argv[]) {
    std::string prompt;

    std::string model_path = argv[1];
    ov::genai::LLMPipeline pipe(model_path, "CPU");
    
    ov::genai::GenerationConfig config;
    config.max_new_tokens = 100;
    config.num_groups = 3;
    config.group_size = 5;
    config.diversity_penalty = 1.0f;
    
    pipe.start_chat();
    for (;;;) {
        std::cout << "question:\n";
        std::getline(std::cin, prompt);
        if (prompt == "Stop!")
            break;

        std::cout << "answer:\n";
        auto answer = pipe(prompt, config);
        std::cout << answer << std::endl;
    }
    pipe.finish_chat();
}

Streaming example with lambda function:

#include "openvino/genai/llm_pipeline.hpp"
#include <iostream>

int main(int argc, char* argv[]) {
    std::string model_path = argv[1];
    ov::genai::LLMPipeline pipe(model_path, "CPU");
        
    auto streamer = [](std::string word) { 
        std::cout << word << std::flush; 
        // Return flag correspods whether generation should be stopped.
        // false means continue generation.
        return false;
    };
    std::cout << pipe.generate("The Sun is yellow bacause", streamer);
}

Streaming with a custom class:

#include "openvino/genai/streamer_base.hpp"
#include "openvino/genai/llm_pipeline.hpp"
#include <iostream>

class CustomStreamer: public ov::genai::StreamerBase {
public:
    bool put(int64_t token) {
        bool stop_flag = false; 
        /* 
        custom decoding/tokens processing code
        tokens_cache.push_back(token);
        std::string text = m_tokenizer.decode(tokens_cache);
        ...
        */
        return stop_flag;  // flag whether generation should be stoped, if true generation stops.
    };

    void end() {
        /* custom finalization */
    };
};

int main(int argc, char* argv[]) {
    CustomStreamer custom_streamer;

    std::string model_path = argv[1];
    ov::genai::LLMPipeline pipe(model_path, "CPU");
    std::cout << pipe.generate("The Sun is yellow because", custom_streamer);
}

How It Works

For information on how OpenVINO™ GenAI works, refer to the How It Works Section.

Supported Models

For a list of supported models, refer to the Supported Models Section.

Project details

These details have not been verified by PyPI

Release history Release notifications | RSS feed

2024.5.0.0

Nov 20, 2024

2024.4.1.0.dev20240926 pre-release

Oct 1, 2024

2024.4.0.0

Sep 19, 2024

2024.3.0.0

Jul 31, 2024

This version

2024.2.0.0

Jun 17, 2024

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distributions

No source distribution files available for this release.See tutorial on generating distribution archives.

Built Distributions

openvino_genai-2024.2.0.0-305-cp312-cp312-win_amd64.whl (794.4 kB view details)

Uploaded Jun 17, 2024 CPython 3.12 Windows x86-64

openvino_genai-2024.2.0.0-305-cp312-cp312-manylinux_2_31_aarch64.whl (984.9 kB view details)

Uploaded Jun 17, 2024 CPython 3.12 manylinux: glibc 2.31+ ARM64

openvino_genai-2024.2.0.0-305-cp312-cp312-manylinux_2_17_x86_64.whl (1.1 MB view details)

Uploaded Jun 17, 2024 CPython 3.12 manylinux: glibc 2.17+ x86-64

openvino_genai-2024.2.0.0-305-cp311-cp311-win_amd64.whl (794.1 kB view details)

Uploaded Jun 17, 2024 CPython 3.11 Windows x86-64

openvino_genai-2024.2.0.0-305-cp311-cp311-manylinux_2_31_aarch64.whl (985.9 kB view details)

Uploaded Jun 17, 2024 CPython 3.11 manylinux: glibc 2.31+ ARM64

openvino_genai-2024.2.0.0-305-cp311-cp311-manylinux_2_17_x86_64.whl (1.1 MB view details)

Uploaded Jun 17, 2024 CPython 3.11 manylinux: glibc 2.17+ x86-64

openvino_genai-2024.2.0.0-305-cp311-cp311-macosx_11_0_arm64.whl (2.5 MB view details)

Uploaded Jun 17, 2024 CPython 3.11 macOS 11.0+ ARM64

openvino_genai-2024.2.0.0-305-cp311-cp311-macosx_10_15_x86_64.whl (2.9 MB view details)

Uploaded Jun 17, 2024 CPython 3.11 macOS 10.15+ x86-64

openvino_genai-2024.2.0.0-305-cp310-cp310-win_amd64.whl (792.9 kB view details)

Uploaded Jun 17, 2024 CPython 3.10 Windows x86-64

openvino_genai-2024.2.0.0-305-cp310-cp310-manylinux_2_31_aarch64.whl (984.5 kB view details)

Uploaded Jun 17, 2024 CPython 3.10 manylinux: glibc 2.31+ ARM64

openvino_genai-2024.2.0.0-305-cp310-cp310-manylinux_2_17_x86_64.whl (1.1 MB view details)

Uploaded Jun 17, 2024 CPython 3.10 manylinux: glibc 2.17+ x86-64

openvino_genai-2024.2.0.0-305-cp310-cp310-macosx_11_0_arm64.whl (2.5 MB view details)

Uploaded Jun 17, 2024 CPython 3.10 macOS 11.0+ ARM64

openvino_genai-2024.2.0.0-305-cp310-cp310-macosx_10_15_x86_64.whl (2.9 MB view details)

Uploaded Jun 17, 2024 CPython 3.10 macOS 10.15+ x86-64

openvino_genai-2024.2.0.0-305-cp39-cp39-win_amd64.whl (793.2 kB view details)

Uploaded Jun 17, 2024 CPython 3.9 Windows x86-64

openvino_genai-2024.2.0.0-305-cp39-cp39-manylinux_2_31_aarch64.whl (984.9 kB view details)

Uploaded Jun 17, 2024 CPython 3.9 manylinux: glibc 2.31+ ARM64

openvino_genai-2024.2.0.0-305-cp39-cp39-manylinux_2_17_x86_64.whl (1.1 MB view details)

Uploaded Jun 17, 2024 CPython 3.9 manylinux: glibc 2.17+ x86-64

openvino_genai-2024.2.0.0-305-cp39-cp39-macosx_11_0_arm64.whl (2.5 MB view details)

Uploaded Jun 17, 2024 CPython 3.9 macOS 11.0+ ARM64

openvino_genai-2024.2.0.0-305-cp39-cp39-macosx_10_15_x86_64.whl (2.9 MB view details)

Uploaded Jun 17, 2024 CPython 3.9 macOS 10.15+ x86-64

openvino_genai-2024.2.0.0-305-cp38-cp38-win_amd64.whl (793.1 kB view details)

Uploaded Jun 17, 2024 CPython 3.8 Windows x86-64

openvino_genai-2024.2.0.0-305-cp38-cp38-manylinux_2_31_aarch64.whl (984.3 kB view details)

Uploaded Jun 17, 2024 CPython 3.8 manylinux: glibc 2.31+ ARM64

openvino_genai-2024.2.0.0-305-cp38-cp38-manylinux_2_17_x86_64.whl (1.1 MB view details)

Uploaded Jun 17, 2024 CPython 3.8 manylinux: glibc 2.17+ x86-64

openvino_genai-2024.2.0.0-305-cp38-cp38-macosx_11_0_arm64.whl (2.5 MB view details)

Uploaded Jun 17, 2024 CPython 3.8 macOS 11.0+ ARM64

openvino_genai-2024.2.0.0-305-cp38-cp38-macosx_10_15_x86_64.whl (2.9 MB view details)

Uploaded Jun 17, 2024 CPython 3.8 macOS 10.15+ x86-64

File details

Details for the file openvino_genai-2024.2.0.0-305-cp312-cp312-win_amd64.whl.

File metadata

Download URL: openvino_genai-2024.2.0.0-305-cp312-cp312-win_amd64.whl
Upload date: Jun 17, 2024
Size: 794.4 kB
Tags: CPython 3.12, Windows x86-64
Uploaded using Trusted Publishing? No
Uploaded via: twine/5.1.0 CPython/3.8.10

File hashes

Hashes for openvino_genai-2024.2.0.0-305-cp312-cp312-win_amd64.whl
Algorithm	Hash digest
SHA256	`02172dd8e6e6638623c323208a0eba2147cd9f43d588b836848870ad3bf8a471`
MD5	`54d025a8dd9f3215104c35ab31e227ff`
BLAKE2b-256	`bc5ad3d168d738676f7c4e2d6c61d6c5ceb614331495130a1c972e40a94cc856`

Algorithm	Hash digest
SHA256	`d9ad88273c4b705030a240aee595b7f00802160bb2231ca2a3ea8b1066909f33`
MD5	`aadbc5db0189ed9b22fa6e99a9c67477`
BLAKE2b-256	`adff1b38d48f9fe1b5c2c012b38048007b05892a6e256d83a0a80892ae4cae73`

Algorithm	Hash digest
SHA256	`a1fa65e380526ebc0ea462c1c4b98c72f6b6d7282515f77eff0d3b988c100af9`
MD5	`fc08f3a01731c39f9b61f6714e7a193a`
BLAKE2b-256	`64cfe04cc16cf7cd2fcf945cda7e87de19819b73154e84815ab9234e32736216`

Algorithm	Hash digest
SHA256	`7ec48980cb61aa3308c15586d09cb86ea204cf0ed165678b3f594f6cfec4b827`
MD5	`e385fe4bb30c134142e1386a34289932`
BLAKE2b-256	`277386a8ca834d5b5522226320a981f70ace208e8aefa2434f0a1b99f1596f37`

Algorithm	Hash digest
SHA256	`87512f8f1a62b1addd287ecd2c9a97ba13409528682ae688b69d4c841ee44ffc`
MD5	`403720413c8e490a803ab9c5ae2e0434`
BLAKE2b-256	`7a934c1db9be826f609f4023b8521b0f7eef5b43035dbc3adb998a4d762dfc20`

Algorithm	Hash digest
SHA256	`1d763dec72245e7340208ec11300d0f2f8681877bd99273f05f8e752c06ab24d`
MD5	`f0ff2c545759ca17b942d96d30162146`
BLAKE2b-256	`f0842897f91d948b282578505e42183b58ca1c09a270d8aba1737a64ca9555c2`

Algorithm	Hash digest
SHA256	`d0b399167b1a27acf267ca8d35fb7b38180645a3f266cd86267433a6347caf27`
MD5	`755e5b8e32469a707bb26b66919db9db`
BLAKE2b-256	`6798dcc8633372bb138b303e0d88e8d4cab4dd27f71412f11952a1eb2d9ea976`

Algorithm	Hash digest
SHA256	`83eefc5aeee58da29a279262c1da7f34419a666517cbf4699a7be3301f7a7a41`
MD5	`51d481f9b87ad715e001f910020e0608`
BLAKE2b-256	`72f814efb0c9640c654bd230118b8cc82aacb8be82d43e6b52304a43d538b74a`

Algorithm	Hash digest
SHA256	`58ad2c5f7141bd78d9dbbadc658384349534fbb71b19d7e22831f40e0095db78`
MD5	`1be6423e57f5e14ab2fd01e5516046e2`
BLAKE2b-256	`87f57e150c5249b44a1129e71ca0740ede2d9cc897f098610d50bda2964b5cf1`

Algorithm	Hash digest
SHA256	`3879e241335e5821787a596ef1aabc63f870d4aa823c65c520bd5979ae125c5c`
MD5	`a5dce21c05d674d171738a5e54634467`
BLAKE2b-256	`c61f512a8bc8d4435dcda973b3ce2646a14131998f18ca2d3b6d3c1b4cb597e0`

Algorithm	Hash digest
SHA256	`5dd2a08ba5d23d86d7c3f25b1446f5d91fb3e07de8a569ff73d78692a3d61c08`
MD5	`eecd0966af1c2e74c6b8f2dca2c57b7a`
BLAKE2b-256	`405963caa88e9eac036fb247afff7352f7a3abb8390356cde77df061746bc643`

Algorithm	Hash digest
SHA256	`7886d7ceca24d3cbc7f470977744924c79fa2a98c8de0e82e750bc012eaa75e1`
MD5	`c511182c947b1d07598555c3e1d54e55`
BLAKE2b-256	`e8ae60df172c22c70ea18d6ad38a94a0247b2709399815704c1bc427e29d3e29`

Algorithm	Hash digest
SHA256	`a77516b11115a3c25d71c3193c4998c5aab4770f0120bd71ecd4ed80dba9432b`
MD5	`f93a8624354f9e3c78d622d7c1edf529`
BLAKE2b-256	`8234409cea1db43ceedb41a5febc9339ac45555e8c02380285d175922317e30c`

Algorithm	Hash digest
SHA256	`5a5030149bcd4c3b8dad61b475808cfd816bb1b44c649ec23e28bc3731998c4c`
MD5	`a98798e3a13f4f8c275873dafb0b67b1`
BLAKE2b-256	`70e204e90661d8fd349f86118715caecf15f44889398cb0f5328cd798a87a4cc`

Algorithm	Hash digest
SHA256	`29bb8b198edcb9739c661fdedf9ad7ad57c75a2b7e379900e8195f6edf456b3f`
MD5	`7750739fe3342b82bcb067b8ed5465db`
BLAKE2b-256	`87610925f3dbde7dd42a30e143dc4d1dc4c1746488631de715e19def99d5d4b1`

Algorithm	Hash digest
SHA256	`2fe301d9f20cbe434cca3e88960ad038170c6a626be50fb7c86c520ce5ab906e`
MD5	`21f67fd14ec7ec9f38b9ce8a729b56bd`
BLAKE2b-256	`33b2f5a3fd0ccd97f4fa14f0866cc8f8a0cec646cc4b8bd2af23dcc73e20d7e2`

Algorithm	Hash digest
SHA256	`5a3d508677f946e2d3bbe0ec9e27c855448f7692a0d935c917ad0f945e6a8415`
MD5	`90e4c9a2992fda6f70c3fea928716c5c`
BLAKE2b-256	`0b90fa66309a24907187b72f362314e9269bb7d21d926040dfc69b64c880fd96`

Algorithm	Hash digest
SHA256	`886861d3432f3e3c608944773597f3c87bf4e72246446f2289da618e63da869b`
MD5	`ad7fbd41aa6bfee8f00c1b4a307a916c`
BLAKE2b-256	`54a93ffd09579a106524b15671beef58844091ef638cd7e8565e61489eeaff47`

Algorithm	Hash digest
SHA256	`2f314ee20d2d8b81c3dc4eb9534d6dbc572d387c723489e3b3950b692cfe9db6`
MD5	`bd0f984bffe3c2722a15efdb10759d7a`
BLAKE2b-256	`3b5df8bf6c1de32a172f6829b5b43de1d30626c87e766a0d4ff18d34508fdc04`

Algorithm	Hash digest
SHA256	`92376694db73e86a850c8aaaedbfa8ac042418226a86d3629fcaa5df8f23ec24`
MD5	`4b307e08d35ac4b90aac342bea519f19`
BLAKE2b-256	`d5e6e950aa70be1b71893119a77e4a1d3723e7437bc4704fc11ae5c4610024ba`

Algorithm	Hash digest
SHA256	`482a79d65235d04fefa320e03a8628467e9ed117488313fcbfa8c8a91f8b35de`
MD5	`b14bd737dd58af522aa4824fdf902888`
BLAKE2b-256	`f17638f12e58c3bd7f749c80bd899724bdab02908fe49572648b5ae8a3cf278b`

openvino-genai 2024.2.0.0

Navigation

Verified details

Maintainers

Unverified details

Meta

Classifiers

Project description

OpenVINO™ GenAI Library

Install OpenVINO™ GenAI

Usage

Prerequisites

Python

C++

How It Works

Supported Models

Project details

Verified details

Maintainers

Unverified details

Meta

Classifiers

Release history Release notifications | RSS feed

Download files

Source Distributions

Built Distributions

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

File details

File metadata

File hashes

Algorithm	Hash digest
SHA256	`a2722180832d9ba21268c851319d5b38742dcac354d92683fad00a8ec386df41`
MD5	`d16cc2855dfee648eb1fa8295f93980c`
BLAKE2b-256	`d6b1832b164adb1c3961711a1593ee30ca754bcc9692531ee8c996c97dd0458b`

Algorithm	Hash digest
SHA256	`dd5d8cc883ae968123359dbb91d7a0e0152c8936ef91327c005fe563fb2de818`
MD5	`d66fb36699fec8db10f624f226c0bed7`
BLAKE2b-256	`72ca3c5ba483b52654aa7184186643a9b4099faa815eb89dc22e2131344f33c3`