OpenVINO support #1037

RyanMetcalfeInt8 · 2023-06-22T18:41:31Z

Running Whisper inference using OpenVINO

This PR extends whisper.cpp to run the Whisper Encoder on OpenVINO supported devices such as CPU, and Intel GPUs (integrated & discrete).

I've tested this on number of platforms, including

NUC Intel(R) Core(TM) i7-6770HQ ('Skylake' Skull Canyon NUC) running Ubuntu 22.04
Core(TM) i7-1185G7 ('Tiger Lake' laptop) running Windows 11 Pro
Core(TM) i7-12700 ('Alder Lake' Beast Canyon NUC) with installed Intel(R) ARC(TM) A770 discrete graphics card, running Windows 11 Pro

For each platform, the performance of using OpenVINO-based encoder gives a great boost in performance over the default encoder -- even for CPU -- and the ability to easily offload to another OpenVINO-supported device by simply specifying a different string at runtime (e.g. "CPU" --> "GPU") is very convenient.

High-level description of changes

This introduction of OpenVINO Encode support is modeled very closely to how whisper.cpp uses CoreML (this should be pretty obvious in the change-set). If the project is built with OpenVINO support, an OpenVINO-specific encoder is pulled into the build and instantiated at application startup time.

Also similar to CoreML, the models required to be present to take advantage of the OpenVINO encoder can be generated using a new python script in 'models' directory.

Just to point out -- something that does differ between CoreML and the new OpenVINO integration is how/when support is enabled at runtime. CoreML is enabled within the call to whisper_init_*. For OpenVINO, because we want the ability to specify a device string ("CPU", "GPU", etc.), I exposed a new API that is dedicated to initializing OpenVINO, given a ctx:

(in whisper.h):

    // Given a context, enable use of OpenVINO for encode inference. 
    // openvino_model_path: Optional path to OpenVINO encoder IR model. If set to nullptr,
    //                      the path will be generated from the ggml model path that was passed
    //                      in to whisper_init_from_file. For example, if 'path_model' was
    //                      "/path/to/ggml-base.en.bin", then OpenVINO IR model path will be 
    //                      assumed to be "/path/to/ggml-base.en-encoder-openvino.xml".
    // openvino_device: OpenVINO device to run inference on ("CPU", "GPU", etc.)
    // openvino_cache_dir: Optional cache directory that can speed up init time, especially for 
    //                     GPU, by caching compiled 'blobs' there.
    //                     Set to nullptr if not used.
    // Returns 1 on success. If OpenVINO is not enabled in build, this
    // simply returns 0.
    WHISPER_API int whisper_ctx_init_openvino_encoder(struct whisper_context* ctx,
        const char* openvino_model_path,
        const char* openvino_device,
        const char* openvino_cache_dir);

I'm happy to rework this if anyone has a better idea of how to enable OpenVINO support at init time.

main.cpp exposes a new parameter for user to set OpenVINO encode inference device (default is "CPU"):

...
else if (arg == "-oved" || arg == "--ov-e-device")    { params.openvino_encode_device = argv[++i]; }
...

And the new whisper_ctx_init_openvino_encoder API is called right after ctx creation:

   // whisper init

    struct whisper_context * ctx = whisper_init_from_file(params.model.c_str());

    if (ctx == nullptr) {
        fprintf(stderr, "error: failed to initialize whisper context\n");
        return 3;
    }

    // initialize openvino encoder. This has no effect on whisper.cpp builds that don't have OpenVINO configured.
    whisper_ctx_init_openvino_encoder(ctx, nullptr, params.openvino_encode_device.c_str(), nullptr);

How to generate models and enable OpenVINO for whisper.cpp builds

Here are the instructions for generating the OpenVINO models for use with OpenVINO-enabled builds of whisper.cpp:

First, setup python virtual env. and install python dependencies. Python 3.10 is recommended.

Windows:

cd models
python -m venv openvino_conv_env
openvino_conv_env\Scripts\activate
python -m pip install --upgrade pip
pip install -r openvino-conversion-requirements.txt

Linux and macOS:

cd models
python3 -m venv openvino_conv_env
source openvino_conv_env/bin/activate
python -m pip install --upgrade pip
pip install -r openvino-conversion-requirements.txt

Generate an OpenVINO encoder model. For example, to generate a base.en model, use:
```
python convert-whisper-to-openvino.py --model base.en
```
This will produce ggml-base.en-encoder-openvino.xml/.bin IR model files. It's recommended to relocate these to the same folder as ggml models, as that is the default location that the OpenVINO extension will search at runtime.
Build whisper.cpp with OpenVINO support:

Download OpenVINO package from release page. The recommended version to use is 2023.0.0.

After downloading & extracting package onto your development system, set up required environment by sourcing setupvars script. For example:

Linux:
```
source /path/to/l_openvino_toolkit_ubuntu22_2023.0.0.10926.b4452d56304_x86_64/setupvars.sh
```
Windows (cmd):
```
C:\Path\To\w_openvino_toolkit_windows_2023.0.0.10926.b4452d56304_x86_64\setupvars.bat
```
And then build the project using cmake:
```
cd build
cmake -DWHISPER_OPENVINO=1 ..
```

Run the examples as usual. For example:

./main -m models/ggml-base.en.bin -f samples/jfk.wav

...

whisper_ctx_init_openvino_encoder: loading OpenVINO model from 'models/ggml-base.en-encoder-openvino.xml'
whisper_ctx_init_openvino_encoder: first run on a device may take a while ...
whisper_openvino_init: path_model = models/ggml-base.en-encoder-openvino.xml, device = CPU, cache_dir = models/ggml-base.en-encoder-openvino-cache
whisper_ctx_init_openvino_encoder: OpenVINO model loaded

system_info: n_threads = 4 / 8 | AVX = 1 | AVX2 = 1 | AVX512 = 0 | FMA = 1 | NEON = 0 | ARM_FMA = 0 | F16C = 1 | FP16_VA = 0 | WASM_SIMD = 0 | BLAS = 0 | SSE3 = 1 | VSX = 0 | COREML = 0 | OPENVINO = 1 |

...

The first time run on an OpenVINO device is slow, since the OpenVINO framework will compile the IR (Intermediate Representation) model to a device-specific 'blob'. This device-specific blob will get
cached for the next run.

You can use -oved [DEVICE] argument to main to specify OpenVINO device to offload encoder inference to. For example:

main -m ggml-base.bin -f gb1.wav -oved GPU

ggerganov · 2023-06-25T12:19:25Z

Wow - this is quite interesting. First time I hear about OpenVINO - will try to get familiar.
I'll need to look into more details, but overall the implementation looks very nice. Very good PR description

Just to point out -- something that does differ between CoreML and the new OpenVINO integration is how/when support is enabled at runtime. CoreML is enabled within the call to whisper_init_*. For OpenVINO, because we want the ability to specify a device string ("CPU", "GPU", etc.), I exposed a new API that is dedicated to initializing OpenVINO, given a ctx:

Do you think it makes sense to do the same for Core ML so that the implementations follow similar pattern?

RyanMetcalfeInt8 · 2023-06-25T19:22:24Z

@ggerganov, thanks for taking a look!

Do you think it makes sense to do the same for Core ML so that the implementations follow similar pattern?

I think that makes sense, especially if CoreML exposes parameters to control how inference is performed -- but to be honest I know very little about CoreML.

ggerganov

Minor changes - should be good to merge after that

whisper.h

whisper.cpp

ggerganov · 2023-06-28T19:05:16Z

CMakeLists.txt

@@ -310,6 +321,7 @@ add_library(${TARGET}
    ${GGML_OPENCL_SOURCES}
    whisper.h
    whisper.cpp
+    ${OpenVINO_SOURCES}


Use OPENVINO_SOURCES

However, why not make a separate target whisper.openvino similar to how whisper.coreml works?

Let me try it again. I had originally tried to add it as a separate target and had some weird issues (something like the corresponding .Lib wasn't being generated in Windows build)-- I intended to circle back though, so thanks for the reminder.

okay, see latest commit (76c4186)

I added openvino-encoder to dedicated OBJECT target:

add_library(${TARGET} OBJECT openvino/whisper-openvino-encoder.h openvino/whisper-openvino-encoder.cpp )

And this target is linked to whisper just like coreml:

if (WHISPER_OPENVINO) target_link_libraries(${TARGET} PRIVATE whisper.openvino) endif()

I was thinking of making it SHARED, but I think it'd be more of a hassle to have to carry around a separate .dll / .so..

This builds fine, and did some minimal testing on Windows 11 & Ubuntu.

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

whisper.cpp

…ino_path_cache to non-const func signatures

ggerganov

Great stuff 👍

Hopefully I didn't break something - haven't tested

Nabaralathep · 2023-07-28T21:07:36Z

Hi!
in the OpenVino instructions there is the next sentence
cd build
cmake -DWHISPER_OPENVINO=1 ..
where is that "build" dir?

And when I run:
./main -m models/ggml-base.en.bin -f samples/jfk.wav
I don't see the "OPENVINO = 1" or any other info about loading openvino

All the other instructions was executed with success
What is missing?

Distro info
I am running on parrot OS 5.3

Amazing work, and thanks for sharing.

RyanMetcalfeInt8 · 2023-07-29T00:35:03Z

Hi @Nabaralathep,

Looks like I forgot the mkdir build, so it should be:

mkdir build
cd build
cmake -DWHISPER_OPENVINO=1 ..
make

Let me know how it goes.

Nabaralathep · 2023-07-30T03:49:48Z

Hi @RyanMetcalfeInt8,
Thank you very much for your help! You pulled me out of a hole, but... I had some issues that I would like to share to help someone else get out of the hole as well.

1.When I run cmake -DWHISPER_OPENVINO=1 .. the build files are created in the back folder and not in build, maybe because of my cmake version (3.18.4), I solved this with cmake -DWHISPER_OPENVINO=1 .. -B build, all being in the "openvino_conv_env" folder.

2.When I ran make I received an error, it turns out that I had the debian arm version and my computer is x86_64, but when I went to the repository to download the appropriate one, I discovered that all the packages for debian are arm so what?

So this is a dead end, and I'm going to install the pink windows called ubuntu.

In any case, I really appreciate (you don't know how much) your answer, since at least it made me understand the problem, thank you very much and your work is incredible.

tazz4843 · 2023-08-29T18:28:55Z

Does this implementation of OpenVINO support the GNA in 10th to 14th generation Intel CPUs? Intel advertises it as follows:

Intel® Gaussian & Neural Accelerator is a low-power neural coprocessor for continuous inference at the edge.

When power and performance are critical, the Intel® Gaussian & Neural Accelerator (Intel® GNA) provides power-efficient, always-on support. Intel® GNA is designed to deliver AI speech and audio applications such as neural noise cancellation, while simultaneously freeing up CPU resources for overall system performance and responsiveness.

They also later mention it could be used for tasks such as speech-to-text, and I'm curious if/how well whisper would perform on it.

Setting the OpenVINO device to "gna" just throws an error with assertion failed

whisper_ctx_init_openvino_encoder: loading OpenVINO model from '../../models/ggml-base-encoder-openvino.xml'
whisper_ctx_init_openvino_encoder: first run on a device may take a while ...
whisper_openvino_init: path_model = ../../models/ggml-base-encoder-openvino.xml, device = GNA, cache_dir = ../../models/ggml-base-encoder-openvino-cache
in openvino encoder compile routine: exception: Check 'false' failed at src/inference/src/core.cpp:114:
[ GENERAL_ERROR ]  AssertionFailed: split_sizes.size() > 1

whisper_ctx_init_openvino_encoder: failed to init OpenVINO encoder from '../../models/ggml-base-encoder-openvino.xml'

ilya-lavrenov · 2023-10-24T01:46:05Z

2.When I ran make I received an error, it turns out that I had the debian arm version and my computer is x86_64, but when I went to the repository to download the appropriate one, I discovered that all the packages for debian are arm so what? !

OpenVINO Ubuntu packages are compatible with Debian OS. You can use OpenVINO archives as well as install via apt and Debian packages.

* openvino: use OpenVINO encoder inference * openvino: add python script for OpenVINO model generation * whisper: Fix 'unused' warnings when OpenVINO isn't enabled in build * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * whisper: Fix compilation error * whisper: revert whisper_get_openvino_path_encoder & whisper_get_openvino_path_cache to non-const func signatures * cmake: Add openvino-encoder as separate object target * whisper : minor style fixes * minor : indentation fixes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

Hopefully I didn't break something - haven't tested

* openvino: use OpenVINO encoder inference * openvino: add python script for OpenVINO model generation * whisper: Fix 'unused' warnings when OpenVINO isn't enabled in build * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * whisper: Fix compilation error * whisper: revert whisper_get_openvino_path_encoder & whisper_get_openvino_path_cache to non-const func signatures * cmake: Add openvino-encoder as separate object target * whisper : minor style fixes * minor : indentation fixes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

Hopefully I didn't break something - haven't tested

ilya-lavrenov · 2023-10-26T03:53:23Z

models/convert-whisper-to-openvino.py

+        onnx_path,
+        input_names=["mel"],
+        output_names=["output_features"]
+    )


it's not required to export to ONNX before usage in OpenVINO.
You can use convert_model with PyTorch in-memory object https://docs.openvino.ai/2023.1/openvino_docs_OV_Converter_UG_prepare_model_convert_model_Convert_Model_From_PyTorch.html

ilya-lavrenov · 2023-10-26T03:54:42Z

models/openvino-conversion-requirements.txt

@@ -0,0 +1,2 @@
+openvino-dev[pytorch,onnx]


we can use openvino>=2023.1.0 which contains update version of convert_model directly in main openvino pip package, while openvino-dev is actually deprecated.

ilya-lavrenov · 2023-10-26T03:56:15Z

openvino/whisper-openvino-encoder.cpp

+        std::shared_ptr<ov::Model> model = core.read_model(path_model);
+
+        // Produce a compiled-model object, given the device ("CPU", "GPU", etc.)
+        auto compiledModel = core.compile_model(model, device);


you can pass path_model directly to compile_model, which can speed-up loading with ov::cache_dir enabled. See https://docs.openvino.ai/2023.1/openvino_docs_OV_UG_Model_caching_overview.html#make-it-even-faster-use-compile-model-modelpath

Any practical speedup from this change?

I'm on OpenVINO 2022.3.1 for device which is EOL'ed. I can compile master and run it with cache:

whisper_openvino_init: path_model = models/ggml-base.en-encoder-openvino.xml, device = MYRIAD, cache_dir = models/ggml-base.en-encoder-openvino-cache

The speed is on par with CPU/GPU OpenVINO. And it helps RPi to inference on base model.

Probably some yes, but the speedup will be during initialization (i.e. the time it takes to pull the model / cached blob from disk and prep the execution device).

RyanMetcalfeInt8 · 2023-10-30T14:18:50Z

@ilya-lavrenov -- good suggestions, looks like OpenVINO made some nice improvements for 2023.1+. Did you want to submit a PR with the updates / fixes?

* openvino: use OpenVINO encoder inference * openvino: add python script for OpenVINO model generation * whisper: Fix 'unused' warnings when OpenVINO isn't enabled in build * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * whisper: Fix compilation error * whisper: revert whisper_get_openvino_path_encoder & whisper_get_openvino_path_cache to non-const func signatures * cmake: Add openvino-encoder as separate object target * whisper : minor style fixes * minor : indentation fixes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

Hopefully I didn't break something - haven't tested

* openvino: use OpenVINO encoder inference * openvino: add python script for OpenVINO model generation * whisper: Fix 'unused' warnings when OpenVINO isn't enabled in build * Apply suggestions from code review Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> * whisper: Fix compilation error * whisper: revert whisper_get_openvino_path_encoder & whisper_get_openvino_path_cache to non-const func signatures * cmake: Add openvino-encoder as separate object target * whisper : minor style fixes * minor : indentation fixes --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

Hopefully I didn't break something - haven't tested

RyanMetcalfeInt8 added 3 commits June 21, 2023 16:21

openvino: use OpenVINO encoder inference

c352893

openvino: add python script for OpenVINO model generation

93b8be4

whisper: Fix 'unused' warnings when OpenVINO isn't enabled in build

58eae32

ggerganov requested changes Jun 28, 2023

View reviewed changes

RyanMetcalfeInt8 and others added 2 commits June 28, 2023 15:18

Apply suggestions from code review

4bc1ebc

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

whisper: Fix compilation error

6bfa371

ggerganov reviewed Jun 28, 2023

View reviewed changes

whisper.cpp Outdated Show resolved Hide resolved

RyanMetcalfeInt8 added 2 commits June 28, 2023 16:13

whisper: revert whisper_get_openvino_path_encoder & whisper_get_openv…

df77368

…ino_path_cache to non-const func signatures

cmake: Add openvino-encoder as separate object target

76c4186

RyanMetcalfeInt8 requested a review from ggerganov June 29, 2023 15:37

ggerganov added 3 commits July 4, 2023 15:37

whisper : minor style fixes

bc5746e

Merge branch 'master' into openvino_integration

0ed471c

minor : indentation fixes

df98287

ggerganov approved these changes Jul 4, 2023

View reviewed changes

ggerganov merged commit 62b8127 into ggerganov:master Jul 4, 2023

ggerganov added a commit that referenced this pull request Jul 4, 2023

whisper : minor OpenVINO refactoring (#1037)

4774d2f

Hopefully I didn't break something - haven't tested

RyanMetcalfeInt8 mentioned this pull request Jul 16, 2023

README.md: Add OpenVINO support details #1112

Merged

jacobwu-b pushed a commit to jacobwu-b/Transcriptify-by-whisper.cpp that referenced this pull request Oct 24, 2023

whisper : minor OpenVINO refactoring (ggerganov#1037)

4611479

Hopefully I didn't break something - haven't tested

jacobwu-b pushed a commit to jacobwu-b/Transcriptify-by-whisper.cpp that referenced this pull request Oct 24, 2023

whisper : minor OpenVINO refactoring (ggerganov#1037)

e266996

Hopefully I didn't break something - haven't tested

ilya-lavrenov reviewed Oct 26, 2023

View reviewed changes

landtanin pushed a commit to landtanin/whisper.cpp that referenced this pull request Dec 16, 2023

whisper : minor OpenVINO refactoring (ggerganov#1037)

7fc7ac6

Hopefully I didn't break something - haven't tested

iThalay pushed a commit to iThalay/whisper.cpp that referenced this pull request Sep 23, 2024

whisper : minor OpenVINO refactoring (ggerganov#1037)

f5956e6

Hopefully I didn't break something - haven't tested

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

OpenVINO support #1037

OpenVINO support #1037

RyanMetcalfeInt8 commented Jun 22, 2023 •

edited

Loading

ggerganov commented Jun 25, 2023

RyanMetcalfeInt8 commented Jun 25, 2023

ggerganov left a comment

ggerganov Jun 28, 2023

RyanMetcalfeInt8 Jun 28, 2023 •

edited

Loading

RyanMetcalfeInt8 Jun 28, 2023

ggerganov left a comment

Nabaralathep commented Jul 28, 2023

RyanMetcalfeInt8 commented Jul 29, 2023 •

edited

Loading

Nabaralathep commented Jul 30, 2023

tazz4843 commented Aug 29, 2023 •

edited

Loading

ilya-lavrenov commented Oct 24, 2023 •

edited

Loading

ilya-lavrenov Oct 26, 2023

ilya-lavrenov Oct 26, 2023

ilya-lavrenov Oct 26, 2023

pukerpanda Dec 20, 2023

RyanMetcalfeInt8 Dec 20, 2023 •

edited

Loading

RyanMetcalfeInt8 commented Oct 30, 2023

OpenVINO support #1037

OpenVINO support #1037

Conversation

RyanMetcalfeInt8 commented Jun 22, 2023 • edited Loading

Running Whisper inference using OpenVINO

High-level description of changes

How to generate models and enable OpenVINO for whisper.cpp builds

ggerganov commented Jun 25, 2023

RyanMetcalfeInt8 commented Jun 25, 2023

ggerganov left a comment

Choose a reason for hiding this comment

ggerganov Jun 28, 2023

Choose a reason for hiding this comment

RyanMetcalfeInt8 Jun 28, 2023 • edited Loading

Choose a reason for hiding this comment

RyanMetcalfeInt8 Jun 28, 2023

Choose a reason for hiding this comment

ggerganov left a comment

Choose a reason for hiding this comment

Nabaralathep commented Jul 28, 2023

RyanMetcalfeInt8 commented Jul 29, 2023 • edited Loading

Nabaralathep commented Jul 30, 2023

tazz4843 commented Aug 29, 2023 • edited Loading

ilya-lavrenov commented Oct 24, 2023 • edited Loading

ilya-lavrenov Oct 26, 2023

Choose a reason for hiding this comment

ilya-lavrenov Oct 26, 2023

Choose a reason for hiding this comment

ilya-lavrenov Oct 26, 2023

Choose a reason for hiding this comment

pukerpanda Dec 20, 2023

Choose a reason for hiding this comment

RyanMetcalfeInt8 Dec 20, 2023 • edited Loading

Choose a reason for hiding this comment

RyanMetcalfeInt8 commented Oct 30, 2023

RyanMetcalfeInt8 commented Jun 22, 2023 •

edited

Loading

RyanMetcalfeInt8 Jun 28, 2023 •

edited

Loading

RyanMetcalfeInt8 commented Jul 29, 2023 •

edited

Loading

tazz4843 commented Aug 29, 2023 •

edited

Loading

ilya-lavrenov commented Oct 24, 2023 •

edited

Loading

RyanMetcalfeInt8 Dec 20, 2023 •

edited

Loading