Add ingestion progress callback

include/bridge.hpp

@@ -61,7 +61,7 @@ namespace fastllama {
         FastLlama& operator=(FastLlama &&) noexcept = default;
         ~FastLlama() { m_model.unload(); }
 
-        bool ingest(std::string prompt, bool is_system_prompt = false);
+        bool ingest(std::string prompt, std::function<void(size_t const&, size_t const&)> fn, bool is_system_prompt = false);
         bool generate(
             std::function<void(std::string const&)> fn,
             std::size_t num_tokens,

interfaces/c/fastllama.h

@@ -98,21 +98,23 @@ bool llama_set_stop_words(struct llama_model_context* model_context, char const*
  * 
  * @param model_context is the context that is constructed using `llama_create_context`.
  * @param prompt is user string that will be processed and produce output.
+ * @param progress_fn callback used to keep track of the advancement of the ingestion
  * @return true if it successfully ingests the prompt.
  * @return false if it is unable to ingest the prompt.
  */
-bool llama_ingest_system_prompt(struct llama_model_context* model_context, char const* prompt);
+bool llama_ingest_system_prompt(struct llama_model_context* model_context, char const* prompt, LLAMA_LOGGER_PROGRESS_FUNC progress_fn);
 
 /**
  * @brief Ingests the prompt that will not be preserved across memory reset to save memory.
  *        It is used for having conversation with the model.
  * 
  * @param model_context is the context that is constructed using `llama_create_context`.
  * @param prompt is user string that will be processed and produce output.
+ * @param progress_fn callback used to keep track of the advancement of the ingestion
  * @return true if it successfully ingests the prompt.
  * @return false if it is unable to ingest the prompt.
  */
-bool llama_ingest(struct llama_model_context* model_context, char const* prompt);
+bool llama_ingest(struct llama_model_context* model_context, char const* prompt, LLAMA_LOGGER_PROGRESS_FUNC progress_fn);
 
 /**
  * @brief Generate the model output from pervious ingested prompt or past conversation.

interfaces/c/main.cpp

@@ -149,16 +149,16 @@ extern "C" {
         return true;
     }
 
-    bool llama_ingest(struct llama_model_context* model_context, char const* prompt) {
+    bool llama_ingest(struct llama_model_context* model_context, char const* prompt, LLAMA_LOGGER_PROGRESS_FUNC progress_fn) {
         if (!is_model_valid(model_context)) return false;
 
-        return model_context->inner->ingest(std::string(prompt), false);
+        return model_context->inner->ingest(std::string(prompt),progress_fn, false);
     }
 
-    bool llama_ingest_system_prompt(struct llama_model_context* model_context, char const* prompt) {
+    bool llama_ingest_system_prompt(struct llama_model_context* model_context, char const* prompt, LLAMA_LOGGER_PROGRESS_FUNC progress_fn) {
         if (!is_model_valid(model_context)) return false;
 
-        return model_context->inner->ingest(std::string(prompt), true);
+        return model_context->inner->ingest(std::string(prompt), progress_fn, true);
     }
 
     bool llama_generate(

interfaces/python/fastllama.py

@@ -321,22 +321,27 @@ def load_state(self, filepath: str) -> bool:
         fn.restype = ctypes.c_bool
         return bool(fn(self.ctx, bytes(filepath, 'utf-8')))
 
-    def ingest(self, prompt: str, is_system_prompt: bool = False) -> bool:
+    def ingest(self, prompt: str, progress_fn=Callable[[int,int], None], is_system_prompt: bool = False) -> bool:
         """
         Ingests a prompt into the model.
 
         :param prompt: The prompt to be ingested.
         :param is_system_prompt: Flag to indicate if the prompt is a system prompt. Default is False.
         :return: True if successful, False otherwise.
         """
+
+        def callback_fn(s: ctypes.c_size_t, t: ctypes.c_size_t):
+            progress_fn(int(s),int(t))
+        ctype_callback_fn = ctypes.CFUNCTYPE(None, ctypes.c_size_t, ctypes.c_size_t)
+
         if is_system_prompt:
             ingest_fn = self.lib.llama_ingest_system_prompt
         else:
             ingest_fn = self.lib.llama_ingest
 
         ingest_fn.argtypes = [c_llama_model_context_ptr, ctypes.c_char_p]
         ingest_fn.restype = ctypes.c_bool
-        return bool(ingest_fn(self.ctx, bytes(prompt, 'utf-8')))
+        return bool(ingest_fn(self.ctx, bytes(prompt, 'utf-8'), ctype_callback_fn(callback_fn)))
 
     def generate(
             self,

lib/bridge.cpp

@@ -183,7 +183,7 @@ namespace fastllama {
         return m_model.dump_vocab(filepath);
     }
 
-    bool FastLlama::ingest(std::string prompt, bool is_system_prompt) {
+    bool FastLlama::ingest(std::string prompt,std::function<void(size_t const&, size_t const&)> fn, bool is_system_prompt) {
         m_model.logger.reset();
         if (!m_model.is_valid) {
             m_model.logger.log_err("FastLlama::ingest", "tried to ingest using invalid model");

src/alpaca.cpp

@@ -33,7 +33,7 @@ int main() {
 
         prompt = "### Instruction:\n\n" + prompt + "\n\n ### Response:\n\n";
 
-        if (!bridge.ingest(prompt)) return 2;
+        if (!bridge.ingest(prompt,[](size_t const& s,size_t const& t){})) return 2;
 
         auto gen_res = bridge.generate([](std::string const& s) {
             std::cout<<s;