talk.wasm : polishing + adding many AI personalities

Author: ggerganov

bindings/javascript/whisper.js

@@ -31,6 +31,15 @@ In order to run this demo efficiently, you need to have the following:
 - Speak phrases that are no longer than 10 seconds - this is the audio context of the AI
 - The web-page uses about 1.4GB of RAM
 
+Notice that this demo is using the smallest GPT-2 model, so the generated text responses are not always very good.
+Also, the prompting strategy can likely be improved to achieve better results.
+
+The demo is quite computationally heavy - it's not usual to run these transformer models in a browser. Typically, they
+run on powerful GPU hardware. So for better experience, you do need to have a powerful computer.
+
+Probably in the near future, mobile browsers will start to support the WASM SIMD capabilities and this will allow
+to run the demo on your phone or tablet. But for now it seems to be not supported (at least on iPhone).
+
 ## Feedback
 
 If you have any comments or ideas for improvement, please drop a comment in the following discussion:

examples/talk.wasm/README.md

@@ -988,7 +988,7 @@ std::atomic<bool> g_running(false);
 
 bool g_force_speak = false;
 std::string g_text_to_speak = "";
-std::string g_status = "idle";
+std::string g_status = "";
 std::string g_status_forced = "";
 
 std::string gpt2_gen_text(const std::string & prompt) {
@@ -997,7 +997,7 @@ std::string gpt2_gen_text(const std::string & prompt) {
     std::vector<float> embd_w;
 
     // tokenize the prompt
-    std::vector<gpt_vocab::id> embd_inp = ::gpt_tokenize(g_gpt2.vocab, g_gpt2.prompt_base + prompt);
+    std::vector<gpt_vocab::id> embd_inp = ::gpt_tokenize(g_gpt2.vocab, prompt);
 
     g_gpt2.n_predict = std::min(g_gpt2.n_predict, g_gpt2.model.hparams.n_ctx - (int) embd_inp.size());
 
@@ -1088,6 +1088,8 @@ void talk_main(size_t index) {
         printf("gpt-2: model loaded in %d ms\n", (int) (t_load_us/1000));
     }
 
+    printf("talk: using %d threads\n", N_THREAD);
+
     std::vector<float> pcmf32;
 
     auto & ctx = g_contexts[index];
@@ -1214,53 +1216,60 @@ void talk_main(size_t index) {
             printf("whisper: number of tokens: %d, '%s'\n", (int) tokens.size(), text_heard.c_str());
 
             std::string text_to_speak;
+            std::string prompt_base;
+
+            {
+                std::lock_guard<std::mutex> lock(g_mutex);
+                prompt_base = g_gpt2.prompt_base;
+            }
 
             if (tokens.size() > 0) {
-                text_to_speak = gpt2_gen_text(text_heard + "\n");
+                text_to_speak = gpt2_gen_text(prompt_base + text_heard + "\n");
                 text_to_speak = std::regex_replace(text_to_speak, std::regex("[^a-zA-Z0-9\\.,\\?!\\s\\:\\'\\-]"), "");
                 text_to_speak = text_to_speak.substr(0, text_to_speak.find_first_of("\n"));
 
                 std::lock_guard<std::mutex> lock(g_mutex);
 
                 // remove first 2 lines of base prompt
                 {
-                    const size_t pos = g_gpt2.prompt_base.find_first_of("\n");
+                    const size_t pos = prompt_base.find_first_of("\n");
                     if (pos != std::string::npos) {
-                        g_gpt2.prompt_base = g_gpt2.prompt_base.substr(pos + 1);
+                        prompt_base = prompt_base.substr(pos + 1);
                     }
                 }
                 {
-                    const size_t pos = g_gpt2.prompt_base.find_first_of("\n");
+                    const size_t pos = prompt_base.find_first_of("\n");
                     if (pos != std::string::npos) {
-                        g_gpt2.prompt_base = g_gpt2.prompt_base.substr(pos + 1);
+                        prompt_base = prompt_base.substr(pos + 1);
                     }
                 }
-                g_gpt2.prompt_base += text_heard + "\n" + text_to_speak + "\n";
+                prompt_base += text_heard + "\n" + text_to_speak + "\n";
             } else {
-                text_to_speak = gpt2_gen_text("");
+                text_to_speak = gpt2_gen_text(prompt_base);
                 text_to_speak = std::regex_replace(text_to_speak, std::regex("[^a-zA-Z0-9\\.,\\?!\\s\\:\\'\\-]"), "");
                 text_to_speak = text_to_speak.substr(0, text_to_speak.find_first_of("\n"));
 
                 std::lock_guard<std::mutex> lock(g_mutex);
 
-                const size_t pos = g_gpt2.prompt_base.find_first_of("\n");
+                const size_t pos = prompt_base.find_first_of("\n");
                 if (pos != std::string::npos) {
-                    g_gpt2.prompt_base = g_gpt2.prompt_base.substr(pos + 1);
+                    prompt_base = prompt_base.substr(pos + 1);
                 }
-                g_gpt2.prompt_base += text_to_speak + "\n";
+                prompt_base += text_to_speak + "\n";
             }
 
             printf("gpt-2: %s\n", text_to_speak.c_str());
 
             //printf("========================\n");
-            //printf("gpt-2: prompt_base:\n'%s'\n", g_gpt2.prompt_base.c_str());
+            //printf("gpt-2: prompt_base:\n'%s'\n", prompt_base.c_str());
             //printf("========================\n");
 
             {
                 std::lock_guard<std::mutex> lock(g_mutex);
                 t_last = std::chrono::high_resolution_clock::now();
                 g_text_to_speak = text_to_speak;
                 g_pcmf32.clear();
+                g_gpt2.prompt_base = prompt_base;
             }
 
             talk_set_status("speaking ...");
@@ -1376,4 +1385,11 @@ EMSCRIPTEN_BINDINGS(talk) {
             g_status_forced = status;
         }
     }));
+
+    emscripten::function("set_prompt", emscripten::optional_override([](const std::string & prompt) {
+        {
+            std::lock_guard<std::mutex> lock(g_mutex);
+            g_gpt2.prompt_base = prompt;
+        }
+    }));
 }