Refactor FFI libraries into distinct wrapper classes for more structure and clarity

Author: CodeWithKyrian

examples/pipelines/asr.php

@@ -11,26 +11,25 @@
 
 ini_set('memory_limit', '-1');
 
-//$transcriber = pipeline('automatic-speech-recognition', 'Xenova/whisper-tiny.en');
-$transcriber = pipeline('automatic-speech-recognition', 'Xenova/whisper-tiny');
+$transcriber = pipeline('automatic-speech-recognition', 'Xenova/whisper-tiny.en');
+//$transcriber = pipeline('automatic-speech-recognition', 'Xenova/whisper-tiny');
 //$transcriber = pipeline('automatic-speech-recognition', 'Xenova/whisper-base');
 //$transcriber = pipeline('automatic-speech-recognition', 'Xenova/wav2vec2-large-xlsr-53-english');
 
 $audioUrl = __DIR__ . '/../sounds/kyrian-dev.wav';
-//$audioUrl = __DIR__ . '/../sounds/jfk.wav';
+$audioUrl = __DIR__ . '/../sounds/jfk.wav';
 //$audioUrl = __DIR__ . '/../sounds/preamble.wav';
 //$audioUrl = __DIR__ . '/../sounds/taunt.wav';
-//$audioUrl = __DIR__ . '/../sounds/gettysburg.wav';
+$audioUrl = __DIR__ . '/../sounds/gettysburg.wav';
 //$audioUrl = __DIR__ . '/../sounds/kyrian-speaking.wav';
 //$audioUrl = __DIR__ . '/../sounds/ted_60.wav';
-$audioUrl = __DIR__ . '/../sounds/french-audio.wav';
+//$audioUrl = __DIR__ . '/../sounds/french-audio.wav';
 
 $streamer = StdOutStreamer::make();
 $output = $transcriber($audioUrl,
     maxNewTokens: 256,
     chunkLengthSecs: 24,
-    task: 'translate'
-//    returnTimestamps: true,
+//    returnTimestamps: 'word',
 //    streamer: $streamer
 );
 

libs/VERSIONS

@@ -1,4 +1,7 @@
 DOWNLOADER=1.0.0
 OPENBLAS=0.3.27
 RINDOW_MATLIB=1.0.1
-ONNXRUNTIME=1.17.0
+ONNXRUNTIME=1.17.0
+SNDFILE=1.2.2
+SAMPLERATE=0.2.2
+FAST_TRANSFORMERS_UTILS=1.0.0

src/FFI/FastTransformersUtils.php

@@ -0,0 +1,62 @@
+<?php
+
+declare(strict_types=1);
+
+
+namespace Codewithkyrian\Transformers\FFI;
+
+use Codewithkyrian\Transformers\Transformers;
+use Codewithkyrian\TransformersLibrariesDownloader\Libraries;
+use FFI;
+use FFI\CData;
+
+class FastTransformersUtils
+{
+    private static FFI $ffi;
+
+    private static function ffi(): FFI
+    {
+        if (!isset(self::$ffi)) {
+            $headerCode = file_get_contents(Libraries::FastTransformersUtils->headerFile(Transformers::$libsDir));
+            self::$ffi = FFI::cdef($headerCode, Libraries::FastTransformersUtils->libFile(Transformers::$libsDir));
+        }
+
+        return self::$ffi;
+    }
+
+    public static function new($type, bool $owned = true, bool $persistent = false): ?CData
+    {
+        return self::ffi()->new($type, $owned, $persistent);
+    }
+
+    public static function enum(string $name)
+    {
+        return self::ffi()->{$name};
+    }
+
+    public static function padReflect($input, int $length, int $paddedLength): CData
+    {
+        $padded = FFI::new("float[$paddedLength]");
+        self::ffi()->pad_reflect($input, $length, $padded, $paddedLength);
+
+        return $padded;
+    }
+
+    public static function spectrogram(
+        $waveform, int $waveformLength, int $spectrogramLength, int $hopLength, int $fftLength,
+        $window, int $windowLength, int $d1, int $d1Max, float $power, bool $center, float $preemphasis,
+        $melFilters, int $nMelFilters, $nFreqBins, float $melFloor, int $logMel, ?bool $removeDcOffset,
+        bool $doPad, bool $transpose
+    ): CData
+    {
+        $spectrogram = FFI::new("float[$spectrogramLength]");
+
+        self::ffi()->spectrogram(
+            $waveform, $waveformLength, $spectrogram, $spectrogramLength, $hopLength, $fftLength, $window,
+            $windowLength, $d1, $d1Max, $power, $center, $preemphasis, $melFilters, $nMelFilters, $nFreqBins,
+            $melFloor, $logMel, $removeDcOffset, $doPad, $transpose,
+        );
+
+        return $spectrogram;
+    }
+}

src/FFI/Libc.php

@@ -0,0 +1,58 @@
+<?php
+
+declare(strict_types=1);
+
+
+namespace Codewithkyrian\Transformers\FFI;
+
+use FFI;
+use FFI\CData;
+use http\Exception\RuntimeException;
+
+class Libc
+{
+    private static FFI $ffi;
+
+    public static function ffi(): FFI
+    {
+        if (!isset(self::$ffi)) {
+            if (PHP_OS_FAMILY == 'Windows') {
+                self::$ffi = FFI::cdef(
+                    'size_t mbstowcs(void *wcstr, const char *mbstr, size_t count);',
+                    'msvcrt.dll'
+                );
+            }
+            else{
+                self::$ffi = FFI::cdef();
+            }
+        }
+
+        return self::$ffi;
+    }
+
+    public static function new($type, bool $owned = true, bool $persistent = false): ?CData
+    {
+        return self::ffi()->new($type, $owned, $persistent);
+    }
+
+    public static function mbStringToWcString(CData $wcStr, string $mbStr, int $count): CData
+    {
+        $length = self::ffi()->mbstowcs($wcStr, $mbStr, $count);
+        if ($length != strlen($mbStr)) {
+            throw new RuntimeException('Expected mbstowcs to return ' . strlen($mbStr) . ", got $length");
+        }
+
+        return $wcStr;
+    }
+
+    public static function cstring($str): CData
+    {
+        $bytes = strlen($str) + 1;
+        // TODO fix?
+        $ptr = self::new("char[$bytes]", owned: false);
+        FFI::memcpy($ptr, $str, $bytes - 1);
+        $ptr[$bytes - 1] = "\0";
+
+        return $ptr;
+    }
+}