ml: Panic rather than return error on tensor allocation failure

FromFloatSlice and FromIntSlice return an error if the shape doesn't
match the passed data or if memory can't be allocated. Since these
are inputs, the memory being allocated is system memory rather than VRAM.

In many cases, the caller can't really handle the error and panics.

Empty and Zeros directly panic if they can't allocate memory.

This makes things consistent by panicing for the first two cases,
removing a fair amount of error handling code. This is also consistent
with how Go typically handles these situations.

model/models/llama/model.go

@@ -142,10 +142,7 @@ func (l *Layer) Forward(ctx ml.Context, hiddenState, positions, outputs ml.Tenso
 }
 
 func (m *Model) Forward(ctx ml.Context, batch input.Batch) (ml.Tensor, error) {
-	positions, err := ctx.Input().FromIntSlice(batch.Positions, len(batch.Positions))
-	if err != nil {
-		return nil, err
-	}
+	positions := ctx.Input().FromIntSlice(batch.Positions, len(batch.Positions))
 
 	hiddenState := m.TokenEmbedding.Forward(ctx, batch.Inputs)
 
@@ -154,10 +151,7 @@ func (m *Model) Forward(ctx ml.Context, batch input.Batch) (ml.Tensor, error) {
 
 		var outputs ml.Tensor
 		if i == len(m.Layers)-1 {
-			outputs, err = ctx.Input().FromIntSlice(batch.Outputs, len(batch.Outputs))
-			if err != nil {
-				return nil, err
-			}
+			outputs = ctx.Input().FromIntSlice(batch.Outputs, len(batch.Outputs))
 		}
 
 		hiddenState = layer.Forward(ctx, hiddenState, positions, outputs, m.Cache, m.Options)