Mark the backend as WIP

tracel-ai · Apr 29, 2024 · cc1a719 · cc1a719
1 parent afb20b1
commit cc1a719
Show file tree

Hide file tree

Showing 5 changed files with 8 additions and 18 deletions.
diff --git a/backend-comparison/Cargo.toml b/backend-comparison/Cargo.toml
@@ -24,8 +24,6 @@ tch-cpu = ["burn/tch"]
 tch-gpu = ["burn/tch"]
 wgpu = ["burn/wgpu", "burn/autotune"]
 wgpu-fusion = ["wgpu", "burn/fusion"]
-cuda = ["burn/cuda", "burn/autotune"]
-cuda-fusion = ["cuda", "burn/fusion"]
 
 [dependencies]
 arboard = { workspace = true }

diff --git a/backend-comparison/src/lib.rs b/backend-comparison/src/lib.rs
@@ -70,13 +70,6 @@ macro_rules! bench_on_backend {
             );
         }
 
-        #[cfg(feature = "cuda")]
-        {
-            use burn::backend::Cuda;
-
-            bench::<Cuda>(&Default::default(), url, token);
-        }
-
         #[cfg(feature = "tch-gpu")]
         {
             use burn::backend::{libtorch::LibTorchDevice, LibTorch};

diff --git a/crates/burn-core/Cargo.toml b/crates/burn-core/Cargo.toml
@@ -20,7 +20,6 @@ default = [
     "burn-tch?/default",
     "burn-tensor/default",
     "burn-wgpu?/default",
-    "burn-cuda?/default",
     "burn-autodiff?/default",
 ]
 std = [
@@ -31,7 +30,6 @@ std = [
     "burn-ndarray?/std",
     "burn-tensor/std",
     "burn-wgpu?/std",
-    "burn-cuda?/std",
     "flate2",
     "half/std",
     "log",
@@ -60,7 +58,6 @@ doc = [
     "burn-tch/doc",
     "burn-tensor/doc",
     "burn-wgpu/doc",
-    "burn-cuda/doc",
 ]
 dataset = ["burn-dataset"]
 network = ["burn-common/network"]
@@ -72,23 +69,21 @@ wasm-sync = ["burn-tensor/wasm-sync", "burn-common/wasm-sync"]
 
 # Backend
 autodiff = ["burn-autodiff"]
-fusion = ["burn-wgpu?/fusion", "burn-cuda?/fusion"]
+fusion = ["burn-wgpu?/fusion"]
 
 ## Backend features
-candle-cuda = ["burn-candle?/cuda"]
 metal = ["burn-candle?/metal"]
 accelerate = ["burn-candle?/accelerate", "burn-ndarray?/blas-accelerate"]
 openblas = ["burn-ndarray?/blas-openblas"]
 openblas-system = ["burn-ndarray?/blas-openblas-system"]
 blas-netlib = ["burn-ndarray?/blas-netlib"]
-autotune = ["burn-wgpu?/autotune", "burn-cuda?/autotune"]
+autotune = ["burn-wgpu?/autotune"]
 template = ["burn-wgpu?/template"]
 
 ndarray = ["burn-ndarray"]
 tch = ["burn-tch"]
 candle = ["burn-candle"]
 wgpu = ["burn-wgpu"]
-cuda = ["burn-cuda"]
 
 # Custom deserializer for Record that is helpful for importing data, such as PyTorch pt files.
 record-item-custom-serde = ["thiserror", "regex"]
@@ -109,7 +104,6 @@ burn-derive = { path = "../burn-derive", version = "0.14.0" }
 burn-tensor = { path = "../burn-tensor", version = "0.14.0", default-features = false }
 
 # Backends
-burn-cuda = { path = "../burn-cuda", version = "0.14.0", optional = true, default-features = false }
 burn-ndarray = { path = "../burn-ndarray", version = "0.14.0", optional = true, default-features = false }
 burn-wgpu = { path = "../burn-wgpu", version = "0.14.0", optional = true, default-features = false }
 burn-autodiff = { path = "../burn-autodiff", version = "0.14.0", optional = true }

diff --git a/crates/burn-cuda/README.md b/crates/burn-cuda/README.md
@@ -0,0 +1,5 @@
+# Burn-Cuda
+
+This backend is still a work in progress and not ready to be used.
+
+See #1525
diff --git a/crates/burn-jit/src/compute/kernel.rs b/crates/burn-jit/src/compute/kernel.rs
@@ -55,7 +55,7 @@ pub struct CompiledKernel {
     pub source: String,
     /// Size of a workgroup for the compiled kernel
     pub workgroup_size: WorkgroupSize,
-    /// TODO:
+    /// The number of bytes used by the share memory
     pub shared_mem_bytes: usize,
 }