Update to Scala 3.4.0

Due to multiple new warnings:

- Added infix modifier for commonly used infix operators and types
- Replaced `: _*` for `*`

Due to new compilation errors:

- Added Promotion `(T, DType) => T` -- match types are behaving slightly
different now.
  - Although I'm still unsure what are the implications of this change
  everything runs properly

build.sbt

@@ -28,7 +28,7 @@ val pytorchVersion = "2.1.2"
 val cudaVersion = "12.3-8.9"
 val openblasVersion = "0.3.26"
 val mklVersion = "2024.0"
-ThisBuild / scalaVersion := "3.3.1"
+ThisBuild / scalaVersion := "3.4.0"
 ThisBuild / javaCppVersion := "1.5.10"
 ThisBuild / resolvers ++= Resolver.sonatypeOssRepos("snapshots")
 

core/src/main/scala/torch/DType.scala

@@ -413,6 +413,7 @@ type DTypeOrDeriveArange[
   *   rules](https://github.com/pytorch/pytorch/blob/fb6749d977e33b5f463c2d0a1b56a939428105e5/c10/core/ScalarType.h#L423-L444)
   */
 type Promoted[T <: DType, U <: DType] <: DType = (T, U) match
+  case (T, DType)                                => T
   case (T, T)                                    => T
   case (U, U)                                    => U
   case (Undefined, U) | (T, Undefined)           => Undefined

core/src/main/scala/torch/Tensor.scala

@@ -281,7 +281,7 @@ sealed abstract class Tensor[D <: DType]( /* private[torch]  */ val native: pyto
       case _                                        => false
 
   /** True if `other` has the same size and elements as this tensor, false otherwise. */
-  def equal(other: Tensor[D]): Boolean = native.equal(other.native)
+  infix def equal(other: Tensor[D]): Boolean = native.equal(other.native)
 
   /** Returns the tensor with elements exponentiated. */
   def exp: Tensor[D] = fromNative(native.exp())
@@ -415,7 +415,7 @@ sealed abstract class Tensor[D <: DType]( /* private[torch]  */ val native: pyto
 
   def <(other: ScalaType): Tensor[Bool] = lt(other)
 
-  def matmul[D2 <: DType](u: Tensor[D2]): Tensor[Promoted[D, D2]] =
+  infix def matmul[D2 <: DType](u: Tensor[D2]): Tensor[Promoted[D, D2]] =
     fromNative(native.matmul(u.native))
 
   def `@`[D2 <: DType](u: Tensor[D2]): Tensor[Promoted[D, D2]] = matmul(u)

core/src/main/scala/torch/Types.scala

@@ -56,4 +56,4 @@ type AtLeastOneFloatOrComplex[A <: DType, B <: DType] = A <:< (FloatNN | Complex
   B <:< (FloatNN | ComplexNN)
 
 /* Evidence that two dtypes are not the same */
-type NotEqual[D <: DType, D2 <: DType] = NotGiven[D =:= D2]
+infix type NotEqual[D <: DType, D2 <: DType] = NotGiven[D =:= D2]

core/src/main/scala/torch/nn/functional/Convolution.scala

@@ -123,7 +123,7 @@ private[torch] trait Convolution {
         toArray(padding),
         toArray(outputPadding),
         groups,
-        toArray(dilation): _*
+        toArray(dilation)*
       )
     )
 
@@ -151,7 +151,7 @@ private[torch] trait Convolution {
         toArray(padding),
         toArray(outputPadding),
         groups,
-        toArray(dilation): _*
+        toArray(dilation)*
       )
     )
 
@@ -179,7 +179,7 @@ private[torch] trait Convolution {
         toArray(padding),
         toArray(outputPadding),
         groups,
-        toArray(dilation): _*
+        toArray(dilation)*
       )
     )
 

core/src/main/scala/torch/nn/modules/container/ModuleList.scala

@@ -77,7 +77,7 @@ final class ModuleList[D <: DType](override val modules: TensorModule[D]*)
     // TODO: not in Python code. Note other modules retain index, so we have repeats
     this.register(module)(using Name(index.toString()))
     // TODO: make modules list mutable?
-    ModuleList(all: _*)
+    ModuleList(all*)
 
   /** Appends a given module to the end of the list.
     *
@@ -94,7 +94,7 @@ final class ModuleList[D <: DType](override val modules: TensorModule[D]*)
     this.register(module)(using Name(index.toString()))
     val all = modules.appended(module)
     // TODO: make modules list mutable?
-    ModuleList(all: _*)
+    ModuleList(all*)
 
   /** Appends modules from a Python iterable to the end of the list.
     *
@@ -115,7 +115,7 @@ final class ModuleList[D <: DType](override val modules: TensorModule[D]*)
       this.register(module)(using Name(index.toString()))
     )
     // TODO: make modules list mutable?
-    ModuleList(all: _*)
+    ModuleList(all*)
 
   override def hasBias(): Boolean = modules.exists(_.hasBias())
 

core/src/main/scala/torch/nn/modules/normalization/LayerNorm.scala

@@ -92,7 +92,7 @@ final class LayerNorm[ParamType <: FloatNN | ComplexNN: Default](
 ) extends HasWeight[ParamType]
     with TensorModule[ParamType]:
 
-  private val shape: LongVector = LongVector(normalizedShape.map(_.toLong): _*)
+  private val shape: LongVector = LongVector(normalizedShape.map(_.toLong)*)
   private val options: LayerNormOptions = LayerNormOptions(shape)
   options.eps().put(eps)
   options.elementwise_affine().put(elementWiseAffine)

core/src/main/scala/torch/ops/ReductionOps.scala

@@ -916,6 +916,6 @@ private[torch] trait ReductionOps {
     val nativeDim = dim.toArray
     fromNative(
       if nativeDim.isEmpty then torchNative.count_nonzero(input.native)
-      else torchNative.count_nonzero(input.native, nativeDim: _*)
+      else torchNative.count_nonzero(input.native, nativeDim*)
     )
 }

core/src/test/scala/TrainingSuite.scala

@@ -17,7 +17,7 @@
 package torch
 import torch.data.*
 
-class TraininSuite extends munit.FunSuite {
+class TrainingSuite extends munit.FunSuite {
   test("training") {
 
     val xTrain = torch.arange(end = 10, dtype = float32) // .reshape(10, 1)

core/src/test/scala/torch/ops/RandomSamplingOpsSuite.scala

@@ -37,13 +37,13 @@ class RandomSamplingOpsSuite extends TensorCheckSuite {
 
     val g1 = torch.Generator()
     g1.manualSeed(0)
-    val t1 = torch.randint(high = 100, Seq(2, 2), generator = g1)
-    val t2 = torch.randint(high = 100, Seq(2, 2), generator = g1)
+    val t1 = torch.randint(high = 100, size = Seq(2, 2), generator = g1)
+    val t2 = torch.randint(high = 100, size = Seq(2, 2), generator = g1)
     assertNotEquals(t1, t2)
 
     val g2 = torch.Generator()
     g2.manualSeed(0)
-    val t3 = torch.randint(high = 100, Seq(2, 2), generator = g2)
+    val t3 = torch.randint(high = 100, size = Seq(2, 2), generator = g2)
     assertEquals(t1, t3)
 
   }

examples/src/main/scala/gpt/V2.scala

@@ -151,7 +151,7 @@ object V2:
    */
 
   // here are all the unique characters that occur in this text
-  val chars = SortedSet(text: _*)
+  val chars = SortedSet(text*)
   println(s"chars = ${chars.mkString(", ")}")
   val vocab_size = chars.size
   println(s"vocab_size = $vocab_size")
@@ -413,7 +413,7 @@ object V2:
       Utils.register_i(this, Head_2(nEmbed, headSize, blockSize), i)
     }
     // val hs = 0 until numHeads map{ i => Utils.register_i(this, Head(nEmbed, headSize, blockSize, dropout), i) }
-    val heads = register(nn.ModuleList(hs: _*))
+    val heads = register(nn.ModuleList(hs*))
     // TODO: BUG - self.proj = nn.Linear(head_size * num_heads, n_embd)
     val proj = register(nn.Linear(headSize * numHeads, nEmbed))
     // val proj = register( nn.Linear(nEmbed, nEmbed) )
@@ -603,7 +603,7 @@ object V2:
     val token_embedding_table = register(nn.Embedding(vocabSize, nEmbed))
     val position_embedding_table = register(nn.Embedding(blockSize, nEmbed))
     val blocks_i = 0 until nBlocks map { i => Block(nEmbed, nHead, blockSize, vocabSize, dropout) }
-    val blocks = register(nn.Sequential(blocks_i: _*))
+    val blocks = register(nn.Sequential(blocks_i*))
     val ln_f = register(nn.LayerNorm(Seq(nEmbed)))
     val lm_head = register(nn.Linear(nEmbed, vocabSize))
 

vision/src/main/scala/torchvision/transforms/presets.scala

@@ -19,8 +19,7 @@ package transforms
 
 import com.sksamuel.scrimage.ImmutableImage
 import com.sksamuel.scrimage.ScaleMethod
-import torch.Tensor
-import torch.Float32
+import torch.{DType, Float32, Promoted, Tensor}
 import torchvision.transforms.functional.toTensor
 
 object Presets: