com.rest.huggingface 1.0.0-preview.13 (#12)

RageAgainstThePixel · Dec 15, 2023 · cdb6a65 · cdb6a65
1 parent 2e340d3
commit cdb6a65
Show file tree

Hide file tree

Showing 13 changed files with 28 additions and 28 deletions.
diff --git a/Runtime/Inference/Audio/AudioToAudio/AudioToAudioResponse.cs b/Runtime/Inference/Audio/AudioToAudio/AudioToAudioResponse.cs
@@ -24,10 +24,10 @@ public AudioToAudioResponse(string content, JsonSerializerSettings settings)
         public IReadOnlyList<AudioToAudioResult> Results { get; }
 
         /// <inheritdoc />
-        public override async Task DecodeAsync(CancellationToken cancellationToken = default)
-            => await Task.WhenAll(Results.Select(result => DecodeAudioAsync(result, cancellationToken)).ToList());
+        public override async Task DecodeAsync(bool debug = false, CancellationToken cancellationToken = default)
+            => await Task.WhenAll(Results.Select(result => DecodeAudioAsync(result, debug, cancellationToken)).ToList());
 
-        private static async Task DecodeAudioAsync(AudioToAudioResult result, CancellationToken cancellationToken)
+        private static async Task DecodeAudioAsync(AudioToAudioResult result, bool debug = false, CancellationToken cancellationToken = default)
         {
             await Rest.ValidateCacheDirectoryAsync();
             Rest.TryGetDownloadCacheItem(result.Blob, out var guid);
@@ -57,7 +57,7 @@ private static async Task DecodeAudioAsync(AudioToAudioResult result, Cancellati
                 await fileStream.DisposeAsync();
             }
 
-            result.AudioClip = await Rest.DownloadAudioClipAsync($"file://{localFilePath}", AudioType.MPEG, parameters: null, cancellationToken: cancellationToken);
+            result.AudioClip = await Rest.DownloadAudioClipAsync($"file://{localFilePath}", AudioType.MPEG, debug: debug, cancellationToken: cancellationToken);
         }
     }
 }
diff --git a/Runtime/Inference/Audio/TextToSpeech/TextToSpeechResponse.cs b/Runtime/Inference/Audio/TextToSpeech/TextToSpeechResponse.cs
@@ -15,7 +15,7 @@ public sealed class TextToSpeechResponse : BinaryInferenceTaskResponse
 
         public AudioClip AudioClip { get; private set; }
 
-        public override async Task DecodeAsync(Stream stream, CancellationToken cancellationToken = default)
+        public override async Task DecodeAsync(Stream stream, bool debug = false, CancellationToken cancellationToken = default)
         {
             await Rest.ValidateCacheDirectoryAsync();
             var filePath = Path.Combine(Rest.DownloadCacheDirectory, $"{DateTime.UtcNow:yyyy-MM-ddTHH-mm-ssffff}.mp3");
@@ -46,7 +46,7 @@ public override async Task DecodeAsync(Stream stream, CancellationToken cancella
                 await fileStream.DisposeAsync();
             }
 
-            AudioClip = await Rest.DownloadAudioClipAsync($"file://{filePath}", AudioType.MPEG, parameters: null, cancellationToken: cancellationToken);
+            AudioClip = await Rest.DownloadAudioClipAsync($"file://{filePath}", AudioType.MPEG, debug: debug, cancellationToken: cancellationToken);
         }
     }
 }
diff --git a/Runtime/Inference/B64JsonInferenceTaskResponse.cs b/Runtime/Inference/B64JsonInferenceTaskResponse.cs
@@ -13,6 +13,6 @@ protected B64JsonInferenceTaskResponse(string content, JsonSerializerSettings se
         {
         }
 
-        public abstract Task DecodeAsync(CancellationToken cancellationToken = default);
+        public abstract Task DecodeAsync(bool debug = false, CancellationToken cancellationToken = default);
     }
 }
diff --git a/Runtime/Inference/BinaryInferenceTaskResponse.cs b/Runtime/Inference/BinaryInferenceTaskResponse.cs
@@ -8,6 +8,6 @@ namespace HuggingFace.Inference
 {
     public abstract class BinaryInferenceTaskResponse : InferenceTaskResponse
     {
-        public abstract Task DecodeAsync(Stream stream, CancellationToken cancellationToken = default);
+        public abstract Task DecodeAsync(Stream stream, bool debug = false, CancellationToken cancellationToken = default);
     }
 }
diff --git a/Runtime/Inference/ComputerVision/ImageSegmentation/ImageSegmentationResponse.cs b/Runtime/Inference/ComputerVision/ImageSegmentation/ImageSegmentationResponse.cs
@@ -21,10 +21,10 @@ public ImageSegmentationResponse(string content, JsonSerializerSettings settings
 
         public IReadOnlyList<ImageSegmentationResult> Results { get; }
 
-        public override async Task DecodeAsync(CancellationToken cancellationToken = default)
-            => await Task.WhenAll(Results.Select(result => DecodeImageAsync(result, cancellationToken)).ToList());
+        public override async Task DecodeAsync(bool debug = false, CancellationToken cancellationToken = default)
+            => await Task.WhenAll(Results.Select(result => DecodeImageAsync(result, debug, cancellationToken)).ToList());
 
-        private static async Task DecodeImageAsync(ImageSegmentationResult result, CancellationToken cancellationToken)
+        private static async Task DecodeImageAsync(ImageSegmentationResult result, bool debug = false, CancellationToken cancellationToken = default)
         {
             await Rest.ValidateCacheDirectoryAsync();
             Rest.TryGetDownloadCacheItem(result.Blob, out var guid);
@@ -54,7 +54,7 @@ private static async Task DecodeImageAsync(ImageSegmentationResult result, Cance
                 await fileStream.DisposeAsync();
             }
 
-            result.Mask = await Rest.DownloadTextureAsync($"file://{localFilePath}", parameters: null, cancellationToken: cancellationToken);
+            result.Mask = await Rest.DownloadTextureAsync($"file://{localFilePath}", debug: debug, cancellationToken: cancellationToken);
         }
     }
 }
diff --git a/Runtime/Inference/ComputerVision/ImageToImage/ImageToImageResponse.cs b/Runtime/Inference/ComputerVision/ImageToImage/ImageToImageResponse.cs
@@ -15,7 +15,7 @@ public sealed class ImageToImageResponse : BinaryInferenceTaskResponse
 
         public Texture2D Image { get; private set; }
 
-        public override async Task DecodeAsync(Stream stream, CancellationToken cancellationToken = default)
+        public override async Task DecodeAsync(Stream stream, bool debug = false, CancellationToken cancellationToken = default)
         {
             await Rest.ValidateCacheDirectoryAsync();
             var filePath = Path.Combine(Rest.DownloadCacheDirectory, $"{DateTime.UtcNow:yyyy-MM-ddTHH-mm-ssffff}.jpg");
@@ -46,7 +46,7 @@ public override async Task DecodeAsync(Stream stream, CancellationToken cancella
                 await fileStream.DisposeAsync();
             }
 
-            Image = await Rest.DownloadTextureAsync($"file://{filePath}", parameters: null, cancellationToken: cancellationToken);
+            Image = await Rest.DownloadTextureAsync($"file://{filePath}", debug: debug, cancellationToken: cancellationToken);
         }
     }
 }
diff --git a/Runtime/Inference/InferenceEndpoint.cs b/Runtime/Inference/InferenceEndpoint.cs
@@ -120,7 +120,7 @@ async Task<Response> CallEndpointAsync()
 
                 if (jsonResponse is B64JsonInferenceTaskResponse b64JsonInferenceTaskResponse)
                 {
-                    await b64JsonInferenceTaskResponse.DecodeAsync(cancellationToken);
+                    await b64JsonInferenceTaskResponse.DecodeAsync(EnableDebug, cancellationToken);
                 }
 
                 return jsonResponse;
@@ -141,7 +141,7 @@ async Task<Response> CallEndpointAsync()
 
                     try
                     {
-                        await taskResponse.DecodeAsync(contentStream, cancellationToken);
+                        await taskResponse.DecodeAsync(contentStream, EnableDebug, cancellationToken);
                     }
                     finally
                     {

diff --git a/Runtime/Inference/Multimodal/TextToImage/TextToImageB64Response.cs b/Runtime/Inference/Multimodal/TextToImage/TextToImageB64Response.cs
@@ -23,7 +23,7 @@ public TextToImageB64Response(string content, JsonSerializerSettings settings) :
 
         public Texture2D Image { get; private set; }
 
-        public override async Task DecodeAsync(CancellationToken cancellationToken = default)
+        public override async Task DecodeAsync(bool debug = false, CancellationToken cancellationToken = default)
         {
             await Rest.ValidateCacheDirectoryAsync();
             var localFilePath = Path.Combine(Rest.DownloadCacheDirectory, $"{DateTime.UtcNow:yyyy-MM-ddTHH-mm-ssffff}.jpg");
@@ -53,7 +53,7 @@ public override async Task DecodeAsync(CancellationToken cancellationToken = def
                 await fileStream.DisposeAsync();
             }
 
-            Image = await Rest.DownloadTextureAsync($"file://{localFilePath}", parameters: null, cancellationToken: cancellationToken);
+            Image = await Rest.DownloadTextureAsync($"file://{localFilePath}", debug: debug, cancellationToken: cancellationToken);
         }
     }
 }
diff --git a/Runtime/Inference/Multimodal/TextToImage/TextToImageBinaryResponse.cs b/Runtime/Inference/Multimodal/TextToImage/TextToImageBinaryResponse.cs
@@ -15,7 +15,7 @@ public sealed class TextToImageBinaryResponse : BinaryInferenceTaskResponse
 
         public Texture2D Image { get; private set; }
 
-        public override async Task DecodeAsync(Stream stream, CancellationToken cancellationToken = default)
+        public override async Task DecodeAsync(Stream stream, bool debug, CancellationToken cancellationToken = default)
         {
             await Rest.ValidateCacheDirectoryAsync();
             var localFilePath = Path.Combine(Rest.DownloadCacheDirectory, $"{DateTime.UtcNow:yyyy-MM-ddTHH-mm-ssffff}.jpg");
@@ -45,7 +45,7 @@ public override async Task DecodeAsync(Stream stream, CancellationToken cancella
                 await fileStream.DisposeAsync();
             }
 
-            Image = await Rest.DownloadTextureAsync($"file://{localFilePath}", parameters: null, cancellationToken: cancellationToken);
+            Image = await Rest.DownloadTextureAsync($"file://{localFilePath}", debug: debug, cancellationToken: cancellationToken);
         }
     }
 }
diff --git a/Runtime/Inference/NaturalLanguageProcessing/TextGeneration/TextGenerationParameters.cs b/Runtime/Inference/NaturalLanguageProcessing/TextGeneration/TextGenerationParameters.cs
@@ -5,8 +5,8 @@ namespace HuggingFace.Inference.NaturalLanguageProcessing.TextGeneration
     public sealed class TextGenerationParameters
     {
         public TextGenerationParameters(
-            [JsonProperty("top_k")] int? topK = null,
-            [JsonProperty("top_p")] int? topP = null,
+            [JsonProperty("top_k")] float? topK = null,
+            [JsonProperty("top_p")] float? topP = null,
             [JsonProperty("temperature")] float? temperature = null,
             [JsonProperty("repetition_penalty")] float? repetitionPenalty = null,
             [JsonProperty("max_new_tokens")] int? maxTokens = null,
@@ -30,14 +30,14 @@ public TextGenerationParameters(
         /// (Default: None). Integer to define the top tokens considered within the sample operation to create new text.
         /// </summary>
         [JsonProperty("top_k")]
-        public int? TopK { get; set; }
+        public float? TopK { get; set; }
 
         /// <summary>
         /// (Default: None). Float to define the tokens that are within the sample operation of text generation.
         /// Add tokens in the sample for more probable to least probable until the sum of the probabilities is greater than top_p.
         /// </summary>
         [JsonProperty("top_p")]
-        public int? TopP { get; set; }
+        public float? TopP { get; set; }
 
         /// <summary>
         /// (Default: 1.0). Float (0.0-100.0). The temperature of the sampling operation.

diff --git a/Tests/TestFixture_03_Inference_Audio.cs b/Tests/TestFixture_03_Inference_Audio.cs
@@ -26,7 +26,7 @@ public async Task Test_01_AutomaticSpeechRecognitionTask()
             var audioPath = AssetDatabase.GUIDToAssetPath("900a512d644c38c47846d9a6e41961f6");
             var audioClip = AssetDatabase.LoadAssetAtPath<AudioClip>(audioPath);
             using var input = new SingleSourceAudioInput(audioClip);
-            var task = new AutomaticSpeechRecognitionTask(input);
+            var task = new AutomaticSpeechRecognitionTask(input, "distil-whisper/distil-large-v2");
             var response = await HuggingFaceClient.InferenceEndpoint.RunInferenceTaskAsync<AutomaticSpeechRecognitionTask, AutomaticSpeechRecognitionResponse>(task);
             Assert.IsNotNull(response);
             Assert.IsFalse(string.IsNullOrWhiteSpace(response.Result));

diff --git a/Tests/TestFixture_05_Inference_ComputerVision.cs b/Tests/TestFixture_05_Inference_ComputerVision.cs
@@ -81,7 +81,7 @@ public async Task Test_04_ImageToImage()
             var texture = AssetDatabase.LoadAssetAtPath<Texture2D>(imagePath);
             using var input = new SingleSourceImageInput(texture);
             var param = new ImageToImageParameters("Girl with Pearl Earring");
-            var task = new ImageToImageTask(input, param);
+            var task = new ImageToImageTask(input, param, "stabilityai/stable-diffusion-xl-refiner-1.0");
             var response = await HuggingFaceClient.InferenceEndpoint.RunInferenceTaskAsync<ImageToImageTask, ImageToImageResponse>(task);
             Assert.IsNotNull(response);
             Assert.IsNotNull(response.Image);

diff --git a/package.json b/package.json
@@ -3,7 +3,7 @@
   "displayName": "HuggingFace",
   "description": "A Non-Official HuggingFace Rest Client for Unity (UPM)",
   "keywords": [],
-  "version": "1.0.0-preview.12",
+  "version": "1.0.0-preview.13",
   "unity": "2021.3",
   "documentationUrl": "https://github.com/RageAgainstThePixel/com.rest.huggingface#documentation",
   "changelogUrl": "https://github.com/RageAgainstThePixel/com.rest.huggingface/releases",
@@ -20,7 +20,7 @@
     "registry": "https://package.openupm.com"
   },
   "dependencies": {
-    "com.utilities.rest": "2.3.1",
+    "com.utilities.rest": "2.4.0",
     "com.utilities.encoder.wav": "1.0.8",
     "com.utilities.encoder.ogg": "3.0.12"
   },