initial conversion to new format, utf8 errors?

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

@@ -9,13 +9,18 @@ enum LlamaError: Error {
 actor LlamaContext {
     private var model: OpaquePointer
     private var context: OpaquePointer
-    
+    private var batch: llama_batch
     private var tokens_list: [llama_token]
     
+    var n_len: Int32 = 32
+    var n_cur: Int32 = 0
+    var n_decode: Int32 = 0
+    
     init(model: OpaquePointer, context: OpaquePointer) {
         self.model = model
         self.context = context
         self.tokens_list = []
+        self.batch = llama_batch_init(512, 0)
     }
     
     deinit {
@@ -26,14 +31,15 @@ actor LlamaContext {
     
     static func createContext(path: String) throws -> LlamaContext {
         llama_backend_init(false)
-        let params = llama_context_default_params()
-        let model = llama_load_model_from_file(path, params)
+        let model_params = llama_model_default_params()
+        
+        let model = llama_load_model_from_file(path, model_params)
         guard let model else {
             print("Could not load model at \(path)")
             throw LlamaError.couldNotInitializeContext
         }
-        
-        let context = llama_new_context_with_model(model, params)
+        let ctx_params = llama_context_default_params()
+        let context = llama_new_context_with_model(model, ctx_params)
         guard let context else {
             print("Could not load context!")
             throw LlamaError.couldNotInitializeContext
@@ -42,50 +48,52 @@ actor LlamaContext {
         return LlamaContext(model: model, context: context)
     }
     
-    func get_kv_cache() -> Int32 {
-        return llama_get_kv_cache_token_count(context)
+    func get_n_tokens() -> Int32 {
+        return batch.n_tokens;
     }
     
-    func completion_init(text: String) -> Int32 {
+    func completion_init(text: String) {
         print("attempting to complete \(text)...")
         
         tokens_list = tokenize(text: text, add_bos: true)
         
-        let max_context_size = llama_n_ctx(context)
-        let max_tokens_list_size = max_context_size - 4
+        let n_ctx = llama_n_ctx(context)
+        let n_kv_req = tokens_list.count + (Int(n_len) - tokens_list.count)
         
-        if tokens_list.count > max_tokens_list_size {
-            print("error: prompt too long (\(tokens_list.count) tokens, max \(max_tokens_list_size)")
+        print("\n n_len = \(n_len), n_ctx = \(n_ctx), n_kv_req = \(n_kv_req)")
+
+        if n_kv_req > n_ctx {
+            print("error: n_kv_req > n_ctx, the required KV cache size is not big enough")
         }
         
         for id in tokens_list {
             print(token_to_piece(token: id))
         }
         
-        let n_gen = min(32, max_context_size)
-        return n_gen
+        // batch = llama_batch_init(512, 0) // done in init()
+        batch.n_tokens = Int32(tokens_list.count)
+        
+        for i1 in 0...batch.n_tokens {
+            let i = Int(i1)
+            batch.token[i] = tokens_list[i]
+            batch.pos[i] = i1
+            batch.seq_id[i] = 0
+            batch.logits[i] = 0 // false
+        }
+        batch.logits[Int(batch.n_tokens) - 1] = 1 // true
+        
+        if llama_decode(context, batch) != 0 {
+            print("llama_decode() failed")
+        }
+        
+        n_cur = batch.n_tokens
     }
     
     func completion_loop() -> String {
-        var done = false
-        tokens_list.withUnsafeBufferPointer() { cArray in
-            let res = llama_eval(context, cArray.baseAddress, Int32(tokens_list.count), llama_get_kv_cache_token_count(context), 8)
-            if res != 0 {
-                print("error evaluating llama!")
-                done = true
-                return
-            }
-        }
-        if done {
-            return ""
-        }
-        
-        tokens_list.removeAll()
-        
         var new_token_id: llama_token = 0
         
-        let logits = llama_get_logits(context)
         let n_vocab = llama_n_vocab(context)
+        let logits = llama_get_logits(context)
         
         var candidates = Array<llama_token_data>()
         candidates.reserveCapacity(Int(n_vocab))
@@ -98,9 +106,30 @@ actor LlamaContext {
             
             new_token_id = llama_sample_token_greedy(context, &candidates_p)
         }
+        
+        if new_token_id == llama_token_eos(context) || n_cur == n_len {
+            print("\n")
+            return ""
+        }
+        
         let new_token_str = token_to_piece(token: new_token_id)
         print(new_token_str)
         tokens_list.append(new_token_id)
+        
+        batch.token[Int(batch.n_tokens)] = new_token_id
+        batch.pos[Int(batch.n_tokens)] = n_cur
+        batch.seq_id[Int(batch.n_tokens)] = 0
+        batch.logits[Int(batch.n_tokens)] = 1 // true
+        
+        batch.n_tokens += 1
+        n_decode += 1
+        
+        n_cur += 1
+        
+        if llama_decode(context, batch) != 0 {
+            print("failed to evaluate llama!")
+        }
+        
         return new_token_str
     }
     

examples/llama.swiftui/llama.swiftui/Models/LlamaState.swift

@@ -33,12 +33,10 @@ class LlamaState: ObservableObject {
             return
         }
         messageLog += "Attempting to complete text...\n"
-        let n_ctx = await llamaContext.completion_init(text: text)
-        messageLog += "context size: \(n_ctx)\n"
+        await llamaContext.completion_init(text: text)
         messageLog += "\(text)"
         
-        if n_ctx > 0 {
-            while await llamaContext.get_kv_cache() < n_ctx {
+        while await llamaContext.n_cur <= llamaContext.n_len {
             let result = await llamaContext.completion_loop()
             messageLog += "\(result)"
         }
@@ -46,4 +44,3 @@ class LlamaState: ObservableObject {
         messageLog += "\n\ndone\n"
     }
 }
-}