Upgrade to candle 0.4.1 (#1382)

* Fix python main entrypoint in book example

* Remove candle windows safeguards (#1178)

* Bump candle-core from 0.3.3 to 0.4.1

* Remove windows current known issue

Cargo.lock

@@ -691,9 +691,9 @@ dependencies = [
 
 [[package]]
 name = "candle-core"
-version = "0.3.3"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6db8659ea87ee8197d2fc627348916cce0561330ee7ae3874e771691d3cecb2f"
+checksum = "6f1b20174c1707e20f4cb364a355b449803c03e9b0c9193324623cf9787a4e00"
 dependencies = [
  "accelerate-src",
  "byteorder",
@@ -718,18 +718,18 @@ dependencies = [
 
 [[package]]
 name = "candle-kernels"
-version = "0.3.3"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "d80cdd5f1cc60d30ba61353cdba5accd0fbc4d4ef2fe707fcb5179a9821adbea"
+checksum = "5845911a44164ebb73b56a0e23793ba1b583bad102af7400fe4768babc5815b2"
 dependencies = [
  "bindgen_cuda",
 ]
 
 [[package]]
 name = "candle-metal-kernels"
-version = "0.3.3"
+version = "0.4.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "52567e7a314ae0c59db5fbd4874ce461d99fa22adb22ddf7cf296b4d97035b40"
+checksum = "b20d6c0d49121e2709ed9faa958ba915ea59526036bcf27558817d1452a4ff09"
 dependencies = [
  "metal",
  "once_cell",
@@ -3521,7 +3521,6 @@ dependencies = [
  "burn",
  "burn-import",
  "burn-ndarray",
- "cfg-if",
  "float-cmp",
  "serde",
 ]

Cargo.toml

@@ -25,7 +25,7 @@ license = "MIT OR Apache-2.0"
 [workspace.dependencies]
 async-trait = "0.1.74"
 bytemuck = "1.14"
-candle-core = { version = "0.3.3" }
+candle-core = { version = "0.4.1" }
 clap = { version = "4.5.1", features = ["derive"] }
 console_error_panic_hook = "0.1.7"
 csv = "1.3.0"

burn-book/src/import/pytorch-model.md

@@ -30,7 +30,7 @@ class Net(nn.Module):
         x = self.conv2(x)
         return x
 
-def main():
+if __name__ == "__main__":
     torch.manual_seed(42)  # To make it reproducible
     model = Net().to(torch.device("cpu"))
     model_weights = model.state_dict()
@@ -254,5 +254,4 @@ defining the encoder in Burn, allowing the loading of its weights while excludin
 
 ## Current known issues
 
-1. [Candle's pickle library does not currently function on Windows due to a Candle bug](https://github.com/tracel-ai/burn/issues/1178).
-2. [Candle's pickle does not currently unpack boolean tensors](https://github.com/tracel-ai/burn/issues/1179).
+1. [Candle's pickle does not currently unpack boolean tensors](https://github.com/tracel-ai/burn/issues/1179).

crates/burn-candle/src/ops/module.rs

@@ -83,10 +83,6 @@ impl<F: FloatCandleElement, I: IntCandleElement> ModuleOps<Self> for Candle<F, I
         bias: Option<FloatTensor<Self, 1>>,
         options: ConvTransposeOptions<1>,
     ) -> FloatTensor<Self, 3> {
-        assert!(
-            options.groups == 1,
-            "Candle does not support groups in transposed convolutions"
-        );
         let conv_transpose = x
             .tensor
             .conv_transpose1d(
@@ -95,6 +91,7 @@ impl<F: FloatCandleElement, I: IntCandleElement> ModuleOps<Self> for Candle<F, I
                 options.padding_out[0],
                 options.stride[0],
                 options.dilation[0],
+                options.groups,
             )
             .unwrap();
         CandleTensor::new(match bias {

crates/burn-import/pytorch-tests/Cargo.toml

@@ -10,7 +10,6 @@ burn-ndarray = { path = "../../burn-ndarray" }
 serde = { workspace = true }
 float-cmp = { workspace = true }
 burn-import = { path = "../", features = ["pytorch"] }
-cfg-if = "1.0.0"
 
 
 [build-dependencies]

crates/burn-import/pytorch-tests/tests/mod.rs

@@ -1,23 +1,17 @@
-cfg_if::cfg_if! {
-    if #[cfg(not(target_os = "windows"))] {
-        // The crate is not supported on Windows because of Candle's pt bug on Windows
-        // (see https://github.com/huggingface/candle/issues/1454).
-        mod batch_norm;
-        mod boolean;
-        mod buffer;
-        mod complex_nested;
-        mod config;
-        mod conv1d;
-        mod conv2d;
-        mod conv_transpose1d;
-        mod conv_transpose2d;
-        mod embedding;
-        mod group_norm;
-        mod integer;
-        mod key_remap;
-        mod key_remap_chained;
-        mod layer_norm;
-        mod linear;
-        mod missing_module_field;
-    }
-}
+mod batch_norm;
+mod boolean;
+mod buffer;
+mod complex_nested;
+mod config;
+mod conv1d;
+mod conv2d;
+mod conv_transpose1d;
+mod conv_transpose2d;
+mod embedding;
+mod group_norm;
+mod integer;
+mod key_remap;
+mod key_remap_chained;
+mod layer_norm;
+mod linear;
+mod missing_module_field;

examples/pytorch-import/build.rs

@@ -13,16 +13,6 @@ use burn_import::pytorch::PyTorchFileRecorder;
 type B = NdArray<f32>;
 
 fn main() {
-    if cfg!(target_os = "windows") {
-        println!(
-            "{}",
-            "cargo:warning=The crate is not supported on Windows because of ".to_owned()
-                + "Candle's pt bug on Windows "
-                + "(see https://github.com/huggingface/candle/issues/1454)."
-        );
-        std::process::exit(0);
-    }
-
     let device = Default::default();
 
     // Load PyTorch weights into a model record.