[js/webgpu] fixes for fp16 attention (#20440)

Author: guschmue

js/web/lib/wasm/jsep/webgpu/ops/attention.ts

@@ -264,7 +264,7 @@ const createInPlaceSoftmaxProgramInfo = (_context: ComputeContext, input: Tensor
     let local_offset = local_idx * uniforms.elements_per_thread;
     let offset = workgroup_id.x * uniforms.d_comp + local_offset;
 
-    var thread_max_vector = ${inputHelper.type.value}(-3.402823e+38f);
+    var thread_max_vector = ${f32Type}(-3.402823e+38f);
     for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < uniforms.d_comp; i++) {
       thread_max_vector = max(${f32Type}(x[offset + i]), thread_max_vector);
     }
@@ -282,12 +282,12 @@ const createInPlaceSoftmaxProgramInfo = (_context: ComputeContext, input: Tensor
     })()};
     workgroupBarrier();
 
-    var max_value: f32 = -3.402823e+38f;
+    var max_value = -3.402823e+38f;
     for (var i = 0u; i < ${WG}; i++) {
       max_value = max(thread_max[i], max_value);
     }
 
-    var sum_vector = ${inputHelper.type.value}(${0});
+    var sum_vector = ${f32Type}(${0});
     for (var i: u32 = 0; i < uniforms.elements_per_thread && i + local_offset < uniforms.d_comp; i++) {
       sum_vector += exp(${f32Type}(x[offset + i]) - max_value);
     }

js/web/lib/wasm/jsep/webgpu/ops/common.ts

@@ -313,7 +313,7 @@ export const castToF32 = (dataType: string, components: number, value: string) =
     return `f32(${value})`;
   }
 
-  return `vec${components}f32(${value})`;
+  return `vec${components}<f32>(${value})`;
 };
 
 /**