Fix .tensor() method; add tests

Author: vvmnnnkv

package.json

@@ -1,6 +1,6 @@
 {
   "name": "libtorchjs",
-  "version": "1.0.0-alpha.3",
+  "version": "1.0.0-alpha.4",
   "description": "Node.js N-API wrapper for Libtorch",
   "main": "lib/index.js",
   "author": "Vova Manannikov <vova@promail.spb.ru>",

src/libtorchjs.cc

@@ -34,7 +34,8 @@ namespace libtorchjs {
         Napi::Float32Array arr = info[0].As<Napi::Float32Array>();
         size_t elements = arr.ElementLength();
         // make torch tensor
-        at::Tensor tensor = torch::tensor(at::ArrayRef<float>(arr.Data(), elements), torch::requires_grad(false));
+        torch::TensorOptions options;
+        at::Tensor tensor = torch::tensor(at::ArrayRef<float>(arr.Data(), elements), options);
         // napi tensor
         auto napiTensor = Tensor::NewInstance();
         Napi::ObjectWrap<Tensor>::Unwrap(napiTensor)->setTensor(tensor);

src/tensor.cc

@@ -56,7 +56,7 @@ namespace libtorchjs {
         uint64_t size = this->tensor.numel();
         // make unit8 type tensor
         auto byteTensor = this->tensor.clamp(0, 255).to(at::ScalarType::Byte);
-        auto byteData = byteTensor.contiguous().data<uint8_t>();
+        auto byteData = byteTensor.contiguous().data_ptr<uint8_t>();
         // wrap in napi unit8 array
         auto arr = Napi::Uint8Array::New(env, size);
         for (uint64_t i = 0; i < size; i++) {

tests/data/make_jit.py

@@ -0,0 +1,6 @@
+import torch
+
+def mul2(x):
+  return x * 2
+  
+torch.jit.trace(mul2, torch.randn(3, 3)).save("mul2.pt")  

tests/data/mul2.pt

@@ -0,0 +1,29 @@
+const expect = require('chai').expect;
+const torch = require('../lib/');
+const path = require('path');
+
+describe('LibtorchJS', function() {
+
+    it('ones', function() {
+        const ones = torch.ones([2, 2]);
+        expect([...ones.toUint8Array()]).to.deep.equal([1, 1, 1, 1]);
+    });
+
+    it('tensor', function() {
+        const arr = new Float32Array([1, 2, 3.3, 4]);
+        const tensor = torch.tensor(arr);
+        expect([...tensor.toUint8Array()]).to.deep.equal([1, 2, 3, 4]);
+    });
+
+    it('load', function(done) {
+        const input = torch.tensor(new Float32Array([2.5, 3.5]));
+        torch.load(path.join(__dirname, 'data', 'mul2.pt'), function(err, model) {
+          model.forward(input, function(err, result) {
+            const output = result.toUint8Array();
+            expect([...output]).to.deep.equal([5, 7]);
+            done();
+          });
+        });
+    });
+
+});