Switch two unwrap to context.

candle-core/src/quantized/mod.rs

@@ -1,5 +1,5 @@
 //! Code for GGML and GGUF files
-use crate::{CpuStorage, DType, Device, Result, Shape, Storage, Tensor};
+use crate::{Context, CpuStorage, DType, Device, Result, Shape, Storage, Tensor};
 use k_quants::*;
 use std::borrow::Cow;
 
@@ -481,7 +481,7 @@ impl crate::CustomOp1 for QTensor {
             crate::bail!("input tensor has only one dimension {layout:?}")
         }
         let mut dst_shape = src_shape.dims().to_vec();
-        let last_k = dst_shape.pop().unwrap();
+        let last_k = dst_shape.pop().context("empty dst_shape")?;
         if last_k != k {
             crate::bail!("input tensor {layout:?} incompatible with {:?}", self.shape)
         }

candle-transformers/src/generation/mod.rs

@@ -3,7 +3,7 @@
 //! Functionality for modeling sampling strategies and logits processing in text generation
 //! with support for temperature-based sampling, top-k filtering, nucleus sampling (top-p),
 //! and combinations thereof.
-use candle::{DType, Error, Result, Tensor};
+use candle::{Context, DType, Error, Result, Tensor};
 use rand::{distributions::Distribution, SeedableRng};
 
 #[derive(Clone, PartialEq, Debug)]
@@ -45,7 +45,7 @@ impl LogitsProcessor {
             .enumerate()
             .max_by(|(_, u), (_, v)| u.total_cmp(v))
             .map(|(i, _)| i as u32)
-            .unwrap();
+            .context("empty logits")?;
         Ok(next_token)
     }