Merge branch 'main' into remove_wrapper

candle-examples/examples/bert/main.rs

@@ -4,9 +4,9 @@ mod model;
 
 use anyhow::{anyhow, Error as E, Result};
 use candle::Tensor;
-use candle_hub::{api::sync::Api, Cache, Repo, RepoType};
 use candle_nn::VarBuilder;
 use clap::Parser;
+use hf_hub::{api::sync::Api, Cache, Repo, RepoType};
 use model::{BertModel, Config, DTYPE};
 use tokenizers::{PaddingParams, Tokenizer};
 

candle-examples/examples/falcon/main.rs

@@ -5,10 +5,10 @@ extern crate intel_mkl_src;
 
 use anyhow::{Error as E, Result};
 use candle::{DType, Device, Tensor};
-use candle_hub::{api::sync::Api, Repo, RepoType};
 use candle_nn::VarBuilder;
 use candle_transformers::generation::LogitsProcessor;
 use clap::Parser;
+use hf_hub::{api::sync::Api, Repo, RepoType};
 use tokenizers::Tokenizer;
 
 mod model;

candle-examples/examples/llama/main.rs

@@ -16,9 +16,9 @@ use anyhow::{Error as E, Result};
 use clap::Parser;
 
 use candle::{DType, Device, Tensor, D};
-use candle_hub::{api::sync::Api, Repo, RepoType};
 use candle_nn::VarBuilder;
 use candle_transformers::generation::LogitsProcessor;
+use hf_hub::{api::sync::Api, Repo, RepoType};
 
 mod model;
 use model::{Config, Llama};

candle-examples/examples/simple-training/main.rs

@@ -0,0 +1,44 @@
+// This should rearch 91.5% accuracy.
+#[cfg(feature = "mkl")]
+extern crate intel_mkl_src;
+
+use anyhow::Result;
+use candle::{DType, Var, D};
+
+const IMAGE_DIM: usize = 784;
+const LABELS: usize = 10;
+
+pub fn main() -> Result<()> {
+    let dev = candle::Device::cuda_if_available(0)?;
+    let m = candle_nn::vision::mnist::load_dir("data")?;
+    println!("train-images: {:?}", m.train_images.shape());
+    println!("train-labels: {:?}", m.train_labels.shape());
+    println!("test-images: {:?}", m.test_images.shape());
+    println!("test-labels: {:?}", m.test_labels.shape());
+    let ws = Var::zeros((IMAGE_DIM, LABELS), DType::F32, &dev)?;
+    let bs = Var::zeros(LABELS, DType::F32, &dev)?;
+    let sgd = candle_nn::SGD::new(&[&ws, &bs], 0.1);
+    for epoch in 1..200 {
+        let logits = m.train_images.matmul(&ws)?.broadcast_add(&bs)?;
+        let loss = logits.softmax(D::Minus1)?;
+        // TODO: log_softmax + let loss = loss.nll_loss(&m.train_labels);
+        sgd.backward_step(&loss)?;
+
+        let _test_logits = m.test_images.matmul(&ws)?.broadcast_add(&bs)?;
+        /* TODO
+        let test_accuracy = test_logits
+            .argmax(Some(-1), false)
+            .eq_tensor(&m.test_labels)
+            .to_kind(Kind::Float)
+            .mean(Kind::Float)
+            .double_value(&[]);
+        */
+        let test_accuracy = 0.;
+        println!(
+            "{epoch:4} train loss: {:8.5} test acc: {:5.2}%",
+            loss.to_scalar::<f32>()?,
+            100. * test_accuracy
+        )
+    }
+    Ok(())
+}

candle-examples/examples/whisper/main.rs

@@ -11,9 +11,9 @@ extern crate intel_mkl_src;
 
 use anyhow::{Error as E, Result};
 use candle::{safetensors::Load, DType, Device, Tensor};
-use candle_hub::{api::sync::Api, Repo, RepoType};
 use candle_nn::VarBuilder;
 use clap::Parser;
+use hf_hub::{api::sync::Api, Repo, RepoType};
 use rand::{distributions::Distribution, SeedableRng};
 use tokenizers::Tokenizer;