Add some group parameter to convolutions. (#566)

* Add some group parameter to convolutions.

* Avoid some unnecessary groups checks.

* Move the tensor convolution bits.

* Properh handling of groups.

* Bump the crate version.

* And add a changelog.

candle-examples/Cargo.toml

@@ -11,11 +11,11 @@ readme = "README.md"
 
 [dependencies]
 accelerate-src = { workspace = true, optional = true }
-candle = { path = "../candle-core", version = "0.1.2", package = "candle-core" }
-candle-datasets = { path = "../candle-datasets", version = "0.1.2" }
-candle-nn = { path = "../candle-nn", version = "0.1.2" }
-candle-transformers = { path = "../candle-transformers", version = "0.1.2" }
-candle-flash-attn = { path = "../candle-flash-attn", version = "0.1.2", optional = true }
+candle = { path = "../candle-core", version = "0.1.3", package = "candle-core" }
+candle-datasets = { path = "../candle-datasets", version = "0.1.3" }
+candle-nn = { path = "../candle-nn", version = "0.1.3" }
+candle-transformers = { path = "../candle-transformers", version = "0.1.3" }
+candle-flash-attn = { path = "../candle-flash-attn", version = "0.1.3", optional = true }
 safetensors = { workspace = true }
 serde = { workspace = true }
 serde_json = { workspace = true }

candle-examples/examples/musicgen/encodec_model.rs

@@ -274,14 +274,22 @@ impl EncodecConv1d {
                 in_c,
                 out_c,
                 kernel_size,
-                Conv1dConfig { padding: 0, stride },
+                Conv1dConfig {
+                    padding: 0,
+                    stride,
+                    groups: 1,
+                },
                 vb.pp("conv"),
             )?,
             NormType::None => conv1d(
                 in_c,
                 out_c,
                 kernel_size,
-                Conv1dConfig { padding: 0, stride },
+                Conv1dConfig {
+                    padding: 0,
+                    stride,
+                    groups: 1,
+                },
                 vb.pp("conv"),
             )?,
         };

candle-examples/examples/stable-diffusion/resnet.rs

@@ -66,6 +66,7 @@ impl ResnetBlock2D {
         let conv_cfg = nn::Conv2dConfig {
             stride: 1,
             padding: 1,
+            groups: 1,
         };
         let norm1 = nn::group_norm(config.groups, in_channels, config.eps, vs.pp("norm1"))?;
         let conv1 = conv2d(in_channels, out_channels, 3, conv_cfg, vs.pp("conv1"))?;
@@ -79,6 +80,7 @@ impl ResnetBlock2D {
             let conv_cfg = nn::Conv2dConfig {
                 stride: 1,
                 padding: 0,
+                groups: 1,
             };
             Some(conv2d(
                 in_channels,

candle-examples/examples/stable-diffusion/unet_2d.rs

@@ -112,8 +112,8 @@ impl UNet2DConditionModel {
         let bl_attention_head_dim = config.blocks.last().unwrap().attention_head_dim;
         let time_embed_dim = b_channels * 4;
         let conv_cfg = nn::Conv2dConfig {
-            stride: 1,
             padding: 1,
+            ..Default::default()
         };
         let conv_in = conv2d(in_channels, b_channels, 3, conv_cfg, vs.pp("conv_in"))?;
 

candle-examples/examples/stable-diffusion/unet_2d_blocks.rs

@@ -24,7 +24,11 @@ impl Downsample2D {
         padding: usize,
     ) -> Result<Self> {
         let conv = if use_conv {
-            let config = nn::Conv2dConfig { stride: 2, padding };
+            let config = nn::Conv2dConfig {
+                stride: 2,
+                padding,
+                ..Default::default()
+            };
             let conv = conv2d(in_channels, out_channels, 3, config, vs.pp("conv"))?;
             Some(conv)
         } else {

candle-examples/examples/stable-diffusion/vae.rs

@@ -51,8 +51,8 @@ impl Encoder {
         config: EncoderConfig,
     ) -> Result<Self> {
         let conv_cfg = nn::Conv2dConfig {
-            stride: 1,
             padding: 1,
+            ..Default::default()
         };
         let conv_in = nn::conv2d(
             in_channels,
@@ -182,8 +182,8 @@ impl Decoder {
         let n_block_out_channels = config.block_out_channels.len();
         let last_block_out_channels = *config.block_out_channels.last().unwrap();
         let conv_cfg = nn::Conv2dConfig {
-            stride: 1,
             padding: 1,
+            ..Default::default()
         };
         let conv_in = nn::conv2d(
             in_channels,

candle-examples/examples/whisper/model.rs

@@ -308,10 +308,12 @@ impl AudioEncoder {
         let cfg1 = Conv1dConfig {
             padding: 1,
             stride: 1,
+            groups: 1,
         };
         let cfg2 = Conv1dConfig {
             padding: 1,
             stride: 2,
+            groups: 1,
         };
         let conv1 = conv1d(cfg.num_mel_bins, n_state, 3, cfg1, vb.pp("conv1"))?;
         let conv2 = conv1d(n_state, n_state, 3, cfg2, vb.pp("conv2"))?;

candle-examples/examples/yolo-v3/darknet.rs

@@ -128,7 +128,11 @@ fn conv(vb: VarBuilder, index: usize, p: usize, b: &Block) -> Result<(usize, Bl)
         }
         Some(_) | None => (None, true),
     };
-    let conv_cfg = candle_nn::Conv2dConfig { stride, padding };
+    let conv_cfg = candle_nn::Conv2dConfig {
+        stride,
+        padding,
+        groups: 1,
+    };
     let conv = if bias {
         conv2d(p, filters, size, conv_cfg, vb.pp(&format!("conv_{index}")))?
     } else {

candle-examples/examples/yolo-v8/main.rs

@@ -101,7 +101,11 @@ impl ConvBlock {
         padding: Option<usize>,
     ) -> Result<Self> {
         let padding = padding.unwrap_or(k / 2);
-        let cfg = Conv2dConfig { padding, stride };
+        let cfg = Conv2dConfig {
+            padding,
+            stride,
+            groups: 1,
+        };
         let conv = conv2d_no_bias(c1, c2, k, cfg, vb.pp("conv"))?;
         let bn = batch_norm(c2, 1e-3, vb.pp("bn"))?;
         Ok(Self { conv, bn })