feat: fix log style,etc.

Author: magicheng0816

xllm/core/framework/sampling/constrained_decoding.h

@@ -20,16 +20,16 @@ limitations under the License.
 
 namespace xllm {
 
-// constrained decoding is used to ensure that the generated content
+// Constrained decoding is used to ensure that the generated content
 // conforms to specific formats or rules.
 class ConstrainedDecoding {
  public:
   virtual ~ConstrainedDecoding();
 
   virtual bool build_mask_cache();
 
-  // input generated_token_list: [sequence_num][generated_token_ids]
-  // output: mask tensor[sequence_num,vocab_size]
+  // Input generated_token_list: [sequence_num][generated_token_ids]
+  // Output: mask tensor[sequence_num,vocab_size]
   virtual torch::Tensor generate_mask(
       const std::vector<std::vector<int32_t>>& generated_token_list);
 };

xllm/core/framework/sampling/rec_constrained_decoding.cpp

@@ -59,7 +59,7 @@ bool RecConstrainedDecoding::build_mask_cache() {
 
   build_mask_cache_ = true;
 
-  LOG(INFO) << "build mask cache, first token ids size:"
+  LOG(INFO) << "Build mask cache, first token ids size:"
             << first_token_ids.size();
 
   return true;
@@ -73,14 +73,14 @@ torch::Tensor RecConstrainedDecoding::generate_mask(
 
   size_t token_size = generated_token_list[0].size();
 
-  // generate mask for first token
+  // Generate mask for first token
   if (0 == token_size) {
     size_t sequence_num = generated_token_list.size();
     auto mask = first_token_mask_.unsqueeze(0);
     return mask.repeat({sequence_num, 1});
   }
 
-  // generate mask for non-first token
+  // Generate mask for non-first token
   return generate_decode_mask(generated_token_list);
 }
 
@@ -121,12 +121,12 @@ torch::Tensor RecConstrainedDecoding::generate_decode_mask(
           local_vocab_indices.push_back(static_cast<int64_t>(vocab_idx));
         }
       } else {
-        LOG(ERROR) << "fail to generate mask for tokens:"
+        LOG(ERROR) << "Fail to generate mask for tokens:"
                    << generated_token_list[token_idx];
       }
     }
 
-    // merge local results to global batch (thread-safe)
+    // Merge local results to global batch (thread-safe)
     if (!local_token_indices.empty()) {
       std::lock_guard<std::mutex> lock(global_batch_mutex);
       global_batch_token_indices.insert(global_batch_token_indices.end(),