squash! llama : rename batch.logits to batch.output

Update examples/batched.swift/Sources/main.swift,
examples/llama.android/llama/src/main/cpp/llama-android.cpp,
examples/llama.swiftui/llama.cpp.swift/LibLlama.swift to use the new
batch.output field instead of batch.logits.

Author: danbev

examples/batched.swift/Sources/main.swift

@@ -99,11 +99,11 @@ for (i, token) in tokens.enumerated() {
     if let seq_id = batch.seq_id[i] {
         seq_id[0] = 0
     }
-    batch.logits[i] = 0
+    batch.output[i] = 0
 }
 
 // llama_decode will output logits only for the last token of the prompt
-batch.logits[Int(batch.n_tokens) - 1] = 1
+batch.output[Int(batch.n_tokens) - 1] = 1
 
 if llama_decode(context, batch) != 0 {
     print("llama_decode() failed")
@@ -166,7 +166,7 @@ while n_cur <= n_len {
         if let seq_id = batch.seq_id[Int(batch.n_tokens)] {
             seq_id[0] = Int32(i)
         }
-        batch.logits[Int(batch.n_tokens)] = 1
+        batch.output[Int(batch.n_tokens)] = 1
 
         i_batch[i] = batch.n_tokens
 

examples/llama.android/llama/src/main/cpp/llama-android.cpp

@@ -193,7 +193,7 @@ Java_android_llama_cpp_LLamaAndroid_bench_1model(
             common_batch_add(*batch, 0, i, { 0 }, false);
         }
 
-        batch->logits[batch->n_tokens - 1] = true;
+        batch->output[batch->n_tokens - 1] = true;
         llama_kv_cache_clear(context);
 
         const auto t_pp_start = ggml_time_us();
@@ -297,7 +297,7 @@ Java_android_llama_cpp_LLamaAndroid_new_1batch(JNIEnv *, jobject, jint n_tokens,
     for (int i = 0; i < n_tokens; ++i) {
         batch->seq_id[i] = (llama_seq_id *) malloc(sizeof(llama_seq_id) * n_seq_max);
     }
-    batch->logits   = (int8_t *)        malloc(sizeof(int8_t)         * n_tokens);
+    batch->output   = (int8_t *)        malloc(sizeof(int8_t)         * n_tokens);
 
     return reinterpret_cast<jlong>(batch);
 }
@@ -377,7 +377,7 @@ Java_android_llama_cpp_LLamaAndroid_completion_1init(
     }
 
     // llama_decode will output logits only for the last token of the prompt
-    batch->logits[batch->n_tokens - 1] = true;
+    batch->output[batch->n_tokens - 1] = true;
 
     if (llama_decode(context, *batch) != 0) {
         LOGe("llama_decode() failed");

examples/llama.swiftui/llama.cpp.swift/LibLlama.swift

@@ -137,7 +137,7 @@ actor LlamaContext {
             let i = Int(i1)
             llama_batch_add(&batch, tokens_list[i], Int32(i), [0], false)
         }
-        batch.logits[Int(batch.n_tokens) - 1] = 1 // true
+        batch.output[Int(batch.n_tokens) - 1] = 1 // true
 
         if llama_decode(context, batch) != 0 {
             print("llama_decode() failed")
@@ -206,7 +206,7 @@ actor LlamaContext {
             for i in 0..<n_tokens {
                 llama_batch_add(&batch, 0, Int32(i), [0], false)
             }
-            batch.logits[Int(batch.n_tokens) - 1] = 1 // true
+            batch.output[Int(batch.n_tokens) - 1] = 1 // true
 
             llama_kv_cache_clear(context)