refactor: Move AutoConfig and PretrainedConfig to appropriate directories

Author: CodeWithKyrian

src/Utils/AutoConfig.php src/Configs/AutoConfig.phpsrc/Utils/AutoConfig.php renamed to src/Configs/AutoConfig.php

@@ -2,9 +2,7 @@
 
 declare(strict_types=1);
 
-namespace Codewithkyrian\Transformers\Utils;
-
-use Codewithkyrian\Transformers\Configs\PretrainedConfig;
+namespace Codewithkyrian\Transformers\Configs;
 
 /**
  * Helper class which is used to instantiate pretrained configs with the `fromPretrained` function.
@@ -18,8 +16,7 @@ public static function fromPretrained(
         ?string   $cacheDir = null,
         string    $revision = 'main',
         ?callable $onProgress = null
-    ): PretrainedConfig
-    {
+    ): PretrainedConfig {
         return PretrainedConfig::fromPretrained($modelNameOrPath, $config, $cacheDir, $revision, $onProgress);
     }
 }

src/Utils/GenerationConfig.php src/Configs/GenerationConfig.phpsrc/Utils/GenerationConfig.php renamed to src/Configs/GenerationConfig.php

@@ -3,7 +3,7 @@
 declare(strict_types=1);
 
 
-namespace Codewithkyrian\Transformers\Utils;
+namespace Codewithkyrian\Transformers\Configs;
 
 /**
  * Class representing a configuration for a generation task.

src/Generation/LogitsProcessors/WhisperTimeStampLogitsProcessor.php

@@ -5,8 +5,8 @@
 
 namespace Codewithkyrian\Transformers\Generation\LogitsProcessors;
 
+use Codewithkyrian\Transformers\Configs\GenerationConfig;
 use Codewithkyrian\Transformers\Tensor\Tensor;
-use Codewithkyrian\Transformers\Utils\GenerationConfig;
 use function Codewithkyrian\Transformers\Utils\timeUsage;
 
 class WhisperTimeStampLogitsProcessor extends LogitsProcessor
@@ -111,4 +111,4 @@ public function __invoke(array $inputIds, Tensor $logits): Tensor
 
         return $logits;
     }
-}
+}

src/Generation/Samplers/Sampler.php

@@ -5,17 +5,15 @@
 
 namespace Codewithkyrian\Transformers\Generation\Samplers;
 
+use Codewithkyrian\Transformers\Configs\GenerationConfig;
 use Codewithkyrian\Transformers\Tensor\Tensor;
-use Codewithkyrian\Transformers\Utils\GenerationConfig;
 
 /**
  * Sampler is a base class for all sampling methods used for text generation.
  */
 abstract class Sampler
 {
-    public function __construct(protected GenerationConfig $generationConfig)
-    {
-    }
+    public function __construct(protected GenerationConfig $generationConfig) {}
 
     /**
      * Executes the sampler, using the specified logits.

src/Models/Auto/PretrainedMixin.php

@@ -5,10 +5,10 @@
 
 namespace Codewithkyrian\Transformers\Models\Auto;
 
+use Codewithkyrian\Transformers\Configs\AutoConfig;
 use Codewithkyrian\Transformers\Exceptions\UnsupportedModelTypeException;
 use Codewithkyrian\Transformers\Models\ModelArchitecture;
 use Codewithkyrian\Transformers\Models\Pretrained\PretrainedModel;
-use Codewithkyrian\Transformers\Utils\AutoConfig;
 
 /**
  * Base class of all AutoModels. Contains the `from_pretrained` function
@@ -47,8 +47,7 @@ public static function fromPretrained(
         string           $revision = 'main',
         ?string          $modelFilename = null,
         ?callable $onProgress = null
-    ): PretrainedModel
-    {
+    ): PretrainedModel {
         $config = AutoConfig::fromPretrained($modelNameOrPath, $config, $cacheDir, $revision, $onProgress);
 
         foreach (static::MODEL_CLASS_MAPPINGS as $modelClassMapping) {
@@ -72,7 +71,7 @@ public static function fromPretrained(
         }
 
         if (static::BASE_IF_FAIL) {
-//            echo "Unknown model class for model type {$config->modelType}. Using base class PreTrainedModel.";
+            //            echo "Unknown model class for model type {$config->modelType}. Using base class PreTrainedModel.";
 
             return PretrainedModel::fromPretrained(
                 modelNameOrPath: $modelNameOrPath,
@@ -109,4 +108,4 @@ protected static function getModelArchitecture($modelClass): ModelArchitecture
             default => ModelArchitecture::EncoderOnly,
         };
     }
-}
+}

src/Models/ModelArchitecture.php

@@ -4,11 +4,9 @@
 
 namespace Codewithkyrian\Transformers\Models;
 
-use Codewithkyrian\Transformers\Exceptions\MissingModelInputException;
-use Codewithkyrian\Transformers\Exceptions\ModelExecutionException;
 use Codewithkyrian\Transformers\Models\Pretrained\PretrainedModel;
 use Codewithkyrian\Transformers\Tensor\Tensor;
-use Codewithkyrian\Transformers\Utils\GenerationConfig;
+
 use function Codewithkyrian\Transformers\Utils\array_pick;
 use function Codewithkyrian\Transformers\Utils\array_pop_key;
 
@@ -102,8 +100,8 @@ function decoderPrepareInputsForGeneration(PretrainedModel $model, $inputIds, ar
             } // Case 3: Past length >= Input IDs
             else {
                 if (
-                    isset($model->config->image_token_index) &&
-                    in_array($model->config->image_token_index, $inputIds->toArray())
+                    isset($model->config['image_token_index']) &&
+                    in_array($model->config['image_token_index'], $inputIds->toArray())
                 ) {
                     // Support for multiple image tokens
                     $numImageTokens = $model->config['num_image_tokens'] ?? null;
@@ -181,7 +179,7 @@ protected function seq2seqForward(PretrainedModel $model, array $modelInputs): a
         return $this->decoderForward($model, $decoderFeeds, true);
     }
 
-    protected function createPositionIds(array $modelInputs, array $pastKeyValues = null): Tensor
+    protected function createPositionIds(array $modelInputs, ?array $pastKeyValues = null): Tensor
     {
         $inputIds = $modelInputs['input_ids'] ?? null;
         $inputsEmbeds = $modelInputs['inputs_embeds'] ?? null;
@@ -211,7 +209,7 @@ protected function createPositionIds(array $modelInputs, array $pastKeyValues =
         $positionIds = new Tensor($data, Tensor::int64, $attentionMask->shape());
 
         if ($pastKeyValues) {
-            $offset = -(($inputIds ?? $inputsEmbeds)->shape()[1]);
+            $offset = - (($inputIds ?? $inputsEmbeds)->shape()[1]);
             $positionIds = $positionIds->slice(null, [$offset, null]); //  position_ids[:, -input_ids.shape[1] :]
         }
 

src/Models/Pretrained/BartForConditionalGeneration.php

@@ -2,13 +2,11 @@
 
 declare(strict_types=1);
 
-
 namespace Codewithkyrian\Transformers\Models\Pretrained;
 
+use Codewithkyrian\Transformers\Configs\GenerationConfig;
 use Codewithkyrian\Transformers\Configs\PretrainedConfig;
 use Codewithkyrian\Transformers\Models\ModelArchitecture;
-use Codewithkyrian\Transformers\Utils\AutoConfig;
-use Codewithkyrian\Transformers\Utils\GenerationConfig;
 use Codewithkyrian\Transformers\Utils\InferenceSession;
 
 /**
@@ -17,13 +15,12 @@
 class BartForConditionalGeneration extends BartPretrainedModel
 {
     public function __construct(
-        PretrainedConfig               $config,
+        PretrainedConfig         $config,
         InferenceSession         $session,
         public InferenceSession  $decoderMergedSession,
         public ModelArchitecture $modelArchitecture,
         public GenerationConfig  $generationConfig
-    )
-    {
+    ) {
         parent::__construct($config, $session, $modelArchitecture);
     }
 }

src/Models/Pretrained/CLIPVisionModelWithProjection.php

@@ -2,11 +2,10 @@
 
 declare(strict_types=1);
 
-
 namespace Codewithkyrian\Transformers\Models\Pretrained;
 
+use Codewithkyrian\Transformers\Configs\PretrainedConfig;
 use Codewithkyrian\Transformers\Models\ModelArchitecture;
-use Codewithkyrian\Transformers\Utils\AutoConfig;
 
 /**
  * CLIP Vision Model with a projection layer on top (a linear layer on top of the pooled output)
@@ -18,17 +17,15 @@ class CLIPVisionModelWithProjection extends CLIPPretrainedModel
     public static function fromPretrained(
         string $modelNameOrPath,
         bool $quantized = true,
-        AutoConfig|array $config = null,
+        array|PretrainedConfig|null $config = null,
         ?string $cacheDir = null,
         ?string $token = null,
         string $revision = 'main',
         ?string $modelFilename = null,
         ModelArchitecture $modelArchitecture = ModelArchitecture::EncoderOnly,
         ?callable $onProgress = null
-    ): PretrainedModel
-    {
-        // Update default model file name if not provided
+    ): PretrainedModel {
         $modelFilename ??= 'vision_model';
         return parent::fromPretrained($modelNameOrPath, $quantized, $config, $cacheDir, $token, $revision, $modelFilename, $modelArchitecture, $onProgress);
     }
-}
+}

src/Models/Pretrained/GPT2PretrainedModel.php

@@ -2,15 +2,6 @@
 
 declare(strict_types=1);
 
-
 namespace Codewithkyrian\Transformers\Models\Pretrained;
 
-use Codewithkyrian\Transformers\Configs\PretrainedConfig;
-use Codewithkyrian\Transformers\Models\ModelArchitecture;
-use Codewithkyrian\Transformers\Utils\AutoConfig;
-use Codewithkyrian\Transformers\Utils\GenerationConfig;
-use Codewithkyrian\Transformers\Utils\InferenceSession;
-
-class GPT2PretrainedModel extends PretrainedModel
-{
-}
+class GPT2PretrainedModel extends PretrainedModel {}

src/Models/Pretrained/GPTBigCodePretrainedModel.php

@@ -2,15 +2,6 @@
 
 declare(strict_types=1);
 
-
 namespace Codewithkyrian\Transformers\Models\Pretrained;
 
-use Codewithkyrian\Transformers\Configs\PretrainedConfig;
-use Codewithkyrian\Transformers\Models\ModelArchitecture;
-use Codewithkyrian\Transformers\Utils\AutoConfig;
-use Codewithkyrian\Transformers\Utils\GenerationConfig;
-use Codewithkyrian\Transformers\Utils\InferenceSession;
-
-class GPTBigCodePretrainedModel extends PretrainedModel
-{
-}
+class GPTBigCodePretrainedModel extends PretrainedModel {}