Add lots of sliders

presets/Debug-deterministic.txt

@@ -1,2 +1 @@
 do_sample=False,
-max_new_tokens=tokens,

presets/Default.txt

@@ -1,4 +1,3 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_k=100,
 top_p=0.9,

presets/Individual Today.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=0.9,
 top_k=50,
 temperature=1.39,

presets/Kobold-Godlike.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=0.5,
 top_k=0,
 temperature=0.7,

presets/Kobold-Liminal Drift.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=1.0,
 top_k=0,
 temperature=0.66,

presets/Naive.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=1,
 typical_p=0.3,
 temperature=0.7,

presets/NovelAI-Best Guess.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=0.9,
 top_k=100,
 temperature=0.8,

presets/NovelAI-Decadence.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=1.0,
 top_k=100,
 temperature=2,

presets/NovelAI-Genesis.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=0.98,
 top_k=0,
 temperature=0.63,

presets/NovelAI-Lycaenidae.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=0.85,
 top_k=12,
 temperature=2,

presets/NovelAI-Ouroboros.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=1.0,
 top_k=100,
 temperature=1.07,

presets/NovelAI-Pleasing Results.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=1.0,
 top_k=0,
 temperature=0.44,

presets/NovelAI-Sphinx Moth.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=0.18,
 top_k=30,
 temperature=2.0,

presets/NovelAI-Storywriter.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=0.73,
 top_k=0,
 temperature=0.72,

presets/Pygmalion.txt

@@ -1,5 +1,4 @@
 do_sample=True,
-max_new_tokens=tokens,
 top_p=0.9,
 top_k=0,
 temperature=0.5,

presets/Verbose.txt

@@ -1,6 +1,5 @@
 num_beams=10,
-min_length=tokens,
-max_new_tokens=tokens,
+min_length=200,
 length_penalty =1.4,
 no_repeat_ngram_size=2,
 early_stopping=True,

server.py

@@ -150,6 +150,40 @@ def load_model(model_name):
     print(f"Loaded the model in {(time.time()-t0):.2f} seconds.")
     return model, tokenizer
 
+def load_model_wrapper(selected_model):
+    global model_name, model, tokenizer
+
+    if selected_model != model_name:
+        model_name = selected_model
+        model = tokenizer = None
+        if not args.cpu:
+            gc.collect()
+            torch.cuda.empty_cache()
+        model, tokenizer = load_model(model_name)
+
+def load_preset_values(preset_menu, return_dict=False):
+    settings = {
+        'do_sample': True,
+        'temperature': 1,
+        'top_p': 1,
+        'typical_p': 1,
+        'repetition_penalty': 1,
+        'top_k': 50,
+    }
+    with open(Path(f'presets/{preset_menu}.txt'), 'r') as infile:
+        preset = infile.read()
+    for i in preset.split(','):
+        i = i.strip().split('=')
+        if len(i) == 2 and i[0].strip() != 'tokens':
+            settings[i[0].strip()] = eval(i[1].strip())
+
+    settings['temperature'] = min(1.99, settings['temperature'])
+
+    if return_dict:
+        return settings
+    else:
+        return settings['do_sample'], settings['temperature'], settings['top_p'], settings['typical_p'], settings['repetition_penalty'], settings['top_k']
+
 # Removes empty replies from gpt4chan outputs
 def fix_gpt4chan(s):
     for i in range(10):
@@ -194,8 +228,8 @@ def formatted_outputs(reply, model_name):
     else:
         return reply
 
-def generate_reply(question, tokens, inference_settings, selected_model, eos_token=None, stopping_string=None):
-    global model, tokenizer, model_name, loaded_preset, preset
+def generate_reply(question, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, eos_token=None, stopping_string=None):
+    global model_name, model, tokenizer
 
     original_question = question
     if not (args.chat or args.cai_chat):
@@ -203,18 +237,6 @@ def generate_reply(question, tokens, inference_settings, selected_model, eos_tok
     if args.verbose:
         print(f"\n\n{question}\n--------------------\n")
 
-    if selected_model != model_name:
-        model_name = selected_model
-        model = tokenizer = None
-        if not args.cpu:
-            gc.collect()
-            torch.cuda.empty_cache()
-        model, tokenizer = load_model(model_name)
-    if inference_settings != loaded_preset:
-        with open(Path(f'presets/{inference_settings}.txt'), 'r') as infile:
-            preset = infile.read()
-        loaded_preset = inference_settings
-
     input_ids = encode(question, tokens)
     cuda = "" if (args.cpu or args.deepspeed) else ".cuda()"
     n = tokenizer.eos_token_id if eos_token is None else tokenizer.encode(eos_token, return_tensors='pt')[0][-1]
@@ -231,15 +253,29 @@ def generate_reply(question, tokens, inference_settings, selected_model, eos_tok
     else:
         stopping_criteria_list = None
 
-    generate_params = [f"eos_token_id={n}", "stopping_criteria=stopping_criteria_list"]
+    generate_params = [
+        f"eos_token_id={n}",
+        f"stopping_criteria=stopping_criteria_list",
+        f"do_sample={do_sample}",
+        f"temperature={temperature}",
+        f"top_p={top_p}",
+        f"typical_p={typical_p}",
+        f"repetition_penalty={repetition_penalty}",
+        f"top_k={top_k}",
+    ]
+
     if args.deepspeed:
         generate_params.append("synced_gpus=True")
+    if args.no_stream:
+        generate_params.append(f"max_new_tokens=tokens")
+    else:
+        generate_params.append(f"max_new_tokens=8")
 
     # Generate the entire reply at once
     if args.no_stream:
         t0 = time.time()
         with torch.no_grad():
-            output = eval(f"model.generate(input_ids, {','.join(generate_params)}, {preset}){cuda}")
+            output = eval(f"model.generate(input_ids, {','.join(generate_params)}){cuda}")
         reply = decode(output[0])
         t1 = time.time()
         print(f"Output generated in {(t1-t0):.2f} seconds ({(len(output[0])-len(input_ids[0]))/(t1-t0)/8:.2f} it/s, {len(output[0])-len(input_ids[0])} tokens)")
@@ -250,10 +286,9 @@ def generate_reply(question, tokens, inference_settings, selected_model, eos_tok
     # Generate the reply 1 token at a time
     else:
         yield formatted_outputs(original_question, model_name)
-        preset = preset.replace('max_new_tokens=tokens', 'max_new_tokens=8')
         for i in tqdm(range(tokens//8+1)):
             with torch.no_grad():
-                output = eval(f"model.generate(input_ids, {','.join(generate_params)}, {preset}){cuda}")
+                output = eval(f"model.generate(input_ids, {','.join(generate_params)}){cuda}")
             reply = decode(output[0])
             if not (args.chat or args.cai_chat):
                 reply = original_question + apply_extensions(reply[len(question):], "output")
@@ -285,6 +320,18 @@ def update_extensions_parameters(*kwargs):
                     params[param] = eval(f"kwargs[{i}]")
                     i += 1
 
+def get_available_models():
+    return sorted(set([item.replace('.pt', '') for item in map(lambda x : str(x.name), list(Path('models/').glob('*'))+list(Path('torch-dumps/').glob('*'))) if not item.endswith('.txt')]), key=str.lower)
+
+def get_available_presets():
+    return sorted(set(map(lambda x : '.'.join(str(x.name).split('.')[:-1]), Path('presets').glob('*.txt'))), key=str.lower)
+
+def get_available_characters():
+    return ["None"] + sorted(set(map(lambda x : '.'.join(str(x.name).split('.')[:-1]), Path('characters').glob('*.json'))), key=str.lower)
+
+def get_available_extensions():
+    return sorted(set(map(lambda x : x.parts[1], Path('extensions').glob('*/script.py'))), key=str.lower)
+
 def create_extensions_block():
     extensions_ui_elements = []
     default_values = []
@@ -307,62 +354,33 @@ def create_extensions_block():
     btn_extensions = gr.Button("Apply")
     btn_extensions.click(update_extensions_parameters, [*extensions_ui_elements], [])
 
-def get_available_models():
-    return sorted(set([item.replace('.pt', '') for item in map(lambda x : str(x.name), list(Path('models/').glob('*'))+list(Path('torch-dumps/').glob('*'))) if not item.endswith('.txt')]), key=str.lower)
+def create_settings_menus():
+    defaults = load_preset_values(settings[f'preset{suffix}'], return_dict=True)
 
-def get_available_presets():
-    return sorted(set(map(lambda x : '.'.join(str(x.name).split('.')[:-1]), Path('presets').glob('*.txt'))), key=str.lower)
+    with gr.Row():
+        with gr.Column():
+            with gr.Row():
+                model_menu = gr.Dropdown(choices=available_models, value=model_name, label='Model')
+                create_refresh_button(model_menu, lambda : None, lambda : {"choices": get_available_models()}, "refresh-button")
+        with gr.Column():
+            with gr.Row():
+                preset_menu = gr.Dropdown(choices=available_presets, value=settings[f'preset{suffix}'], label='Generation parameters preset')
+                create_refresh_button(preset_menu, lambda : None, lambda : {"choices": get_available_presets()}, "refresh-button")
 
-def get_available_characters():
-    return ["None"] + sorted(set(map(lambda x : '.'.join(str(x.name).split('.')[:-1]), Path('characters').glob('*.json'))), key=str.lower)
+    with gr.Accordion("Custom generation parameters", open=False):
+        with gr.Row():
+            with gr.Column():
+                do_sample = gr.Checkbox(value=defaults['do_sample'], label="do_sample")
+                temperature = gr.Slider(0.01, 1.99, value=defaults['temperature'], step=0.01, label="temperature")
+                top_p = gr.Slider(0.0,1.0,value=defaults['top_p'],step=0.01,label="top_p")
+            with gr.Column():
+                typical_p = gr.Slider(0.0,1.0,value=defaults['typical_p'],step=0.01,label="typical_p")
+                repetition_penalty = gr.Slider(1.0,5.0,value=defaults['repetition_penalty'],step=0.01,label="repetition_penalty")
+                top_k = gr.Slider(0,200,value=defaults['top_k'],step=1,label="top_k")
 
-def get_available_extensions():
-    return sorted(set(map(lambda x : x.parts[1], Path('extensions').glob('*/script.py'))), key=str.lower)
-
-available_models = get_available_models()
-available_presets = get_available_presets()
-available_characters = get_available_characters()
-available_extensions = get_available_extensions()
-extension_state = {}
-if args.extensions is not None:
-    for i,ext in enumerate(args.extensions.split(',')):
-        if ext in available_extensions:
-            print(f'Loading the extension "{ext}"... ', end='')
-            ext_string = f"extensions.{ext}.script"
-            exec(f"import {ext_string}")
-            extension_state[ext] = [True, i]
-            print(f'Ok.')
-
-# Choosing the default model
-if args.model is not None:
-    model_name = args.model
-else:
-    if len(available_models) == 0:
-        print("No models are available! Please download at least one.")
-        sys.exit(0)
-    elif len(available_models) == 1:
-        i = 0
-    else:
-        print("The following models are available:\n")
-        for i,model in enumerate(available_models):
-            print(f"{i+1}. {model}")
-        print(f"\nWhich one do you want to load? 1-{len(available_models)}\n")
-        i = int(input())-1
-        print()
-    model_name = available_models[i]
-model, tokenizer = load_model(model_name)
-loaded_preset = None
-
-# UI settings
-default_text = settings['prompt_gpt4chan'] if model_name.lower().startswith(('gpt4chan', 'gpt-4chan', '4chan')) else settings['prompt']
-description = f"\n\n# Text generation lab\nGenerate text using Large Language Models.\n"
-css = ".my-4 {margin-top: 0} .py-6 {padding-top: 2.5rem} #refresh-button {flex: none; margin: 0; padding: 0; min-width: 50px; border: none; box-shadow: none; border-radius: 0} #download-label, #upload-label {min-height: 0}"
-buttons = {}
-gen_events = []
-
-if args.chat or args.cai_chat:
-    history = {'internal': [], 'visible': []}
-    character = None
+    model_menu.change(load_model_wrapper, [model_menu], [])
+    preset_menu.change(load_preset_values, [preset_menu], [do_sample, temperature, top_p, typical_p, repetition_penalty, top_k])
+    return preset_menu, do_sample, temperature, top_p, typical_p, repetition_penalty, top_k
 
 # This gets the new line characters right.
 def clean_chat_message(text):
@@ -433,14 +451,14 @@ if args.chat or args.cai_chat:
 
     return reply, next_character_found, substring_found
 
-    def chatbot_wrapper(text, tokens, inference_settings, selected_model, name1, name2, context, check, history_size):
+def chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, name1, name2, context, check, history_size):
     original_text = text
     text = apply_extensions(text, "input")
     question = generate_chat_prompt(text, tokens, name1, name2, context, history_size)
     history['internal'].append(['', ''])
     history['visible'].append(['', ''])
     eos_token = '\n' if check else None
-        for reply in generate_reply(question, tokens, inference_settings, selected_model, eos_token=eos_token, stopping_string=f"\n{name1}:"):
+    for reply in generate_reply(question, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, eos_token=eos_token, stopping_string=f"\n{name1}:"):
         reply, next_character_found, substring_found = extract_message_from_reply(question, reply, name2, name1, check, extensions=True)
         history['internal'][-1] = [text, reply]
         history['visible'][-1] = [original_text, apply_extensions(reply, "output")]
@@ -450,10 +468,10 @@ if args.chat or args.cai_chat:
             break
     yield history['visible']
 
-    def impersonate_wrapper(text, tokens, inference_settings, selected_model, name1, name2, context, check, history_size):
+def impersonate_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, name1, name2, context, check, history_size):
     question = generate_chat_prompt(text, tokens, name1, name2, context, history_size, impersonate=True)
     eos_token = '\n' if check else None
-        for reply in generate_reply(question, tokens, inference_settings, selected_model, eos_token=eos_token, stopping_string=f"\n{name2}:"):
+    for reply in generate_reply(question, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, eos_token=eos_token, stopping_string=f"\n{name2}:"):
         reply, next_character_found, substring_found = extract_message_from_reply(question, reply, name1, name2, check, extensions=False)
         if not substring_found:
             yield apply_extensions(reply, "output")
@@ -461,19 +479,19 @@ if args.chat or args.cai_chat:
             break
     yield apply_extensions(reply, "output")
 
-    def cai_chatbot_wrapper(text, tokens, inference_settings, selected_model, name1, name2, context, check, history_size):
-        for _history in chatbot_wrapper(text, tokens, inference_settings, selected_model, name1, name2, context, check, history_size):
+def cai_chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, name1, name2, context, check, history_size):
+    for _history in chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, name1, name2, context, check, history_size):
         yield generate_chat_html(_history, name1, name2, character)
 
-    def regenerate_wrapper(text, tokens, inference_settings, selected_model, name1, name2, context, check, history_size):
+def regenerate_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, name1, name2, context, check, history_size):
     last = history['visible'].pop()
     history['internal'].pop()
     text = last[0]
     if args.cai_chat:
-            for i in cai_chatbot_wrapper(text, tokens, inference_settings, selected_model, name1, name2, context, check, history_size):
+        for i in cai_chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, name1, name2, context, check, history_size):
             yield i
     else:
-            for i in chatbot_wrapper(text, tokens, inference_settings, selected_model, name1, name2, context, check, history_size):
+        for i in chatbot_wrapper(text, tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, name1, name2, context, check, history_size):
             yield i
 
 def remove_last_message(name1, name2):
@@ -660,7 +678,53 @@ if args.chat or args.cai_chat:
     img.save(Path(f'img_me.png'))
     print(f'Profile picture saved to "img_me.png"')
 
+# Global variables
+available_models = get_available_models()
+available_presets = get_available_presets()
+available_characters = get_available_characters()
+available_extensions = get_available_extensions()
+extension_state = {}
+if args.extensions is not None:
+    for i,ext in enumerate(args.extensions.split(',')):
+        if ext in available_extensions:
+            print(f'Loading the extension "{ext}"... ', end='')
+            ext_string = f"extensions.{ext}.script"
+            exec(f"import {ext_string}")
+            extension_state[ext] = [True, i]
+            print(f'Ok.')
+
+# Choosing the default model
+if args.model is not None:
+    model_name = args.model
+else:
+    if len(available_models) == 0:
+        print("No models are available! Please download at least one.")
+        sys.exit(0)
+    elif len(available_models) == 1:
+        i = 0
+    else:
+        print("The following models are available:\n")
+        for i,model in enumerate(available_models):
+            print(f"{i+1}. {model}")
+        print(f"\nWhich one do you want to load? 1-{len(available_models)}\n")
+        i = int(input())-1
+        print()
+    model_name = available_models[i]
+model, tokenizer = load_model(model_name)
+loaded_preset = None
+
+# UI settings
+default_text = settings['prompt_gpt4chan'] if model_name.lower().startswith(('gpt4chan', 'gpt-4chan', '4chan')) else settings['prompt']
+description = f"\n\n# Text generation lab\nGenerate text using Large Language Models.\n"
+css = ".my-4 {margin-top: 0} .py-6 {padding-top: 2.5rem} #refresh-button {flex: none; margin: 0; padding: 0; min-width: 50px; border: none; box-shadow: none; border-radius: 0} #download-label, #upload-label {min-height: 0}"
+buttons = {}
+gen_events = []
+
 suffix = '_pygmalion' if 'pygmalion' in model_name.lower() else ''
+history = {'internal': [], 'visible': []}
+character = None
+
+if args.chat or args.cai_chat:
     with gr.Blocks(css=css+".h-\[40vh\] {height: 66.67vh} .gradio-container {max-width: 800px; margin-left: auto; margin-right: auto} .w-screen {width: unset}", analytics_enabled=False) as interface:
         if args.cai_chat:
             display = gr.HTML(value=generate_chat_html([], "", "", character))
@@ -681,15 +745,11 @@ if args.chat or args.cai_chat:
 
         with gr.Row():
             with gr.Column():
-                length_slider = gr.Slider(minimum=settings['max_new_tokens_min'], maximum=settings['max_new_tokens_max'], step=1, label='max_new_tokens', value=settings['max_new_tokens'])
-                with gr.Row():
-                    model_menu = gr.Dropdown(choices=available_models, value=model_name, label='Model')
-                    create_refresh_button(model_menu, lambda : None, lambda : {"choices": get_available_models()}, "refresh-button")
+                max_new_tokens = gr.Slider(minimum=settings['max_new_tokens_min'], maximum=settings['max_new_tokens_max'], step=1, label='max_new_tokens', value=settings['max_new_tokens'])
             with gr.Column():
                 history_size_slider = gr.Slider(minimum=settings['history_size_min'], maximum=settings['history_size_max'], step=1, label='Chat history size in prompt (0 for no limit)', value=settings['history_size'])
-                with gr.Row():
-                    preset_menu = gr.Dropdown(choices=available_presets, value=settings[f'preset{suffix}'], label='Generation parameters preset')
-                    create_refresh_button(preset_menu, lambda : None, lambda : {"choices": get_available_presets()}, "refresh-button")
+
+        preset_menu, do_sample, temperature, top_p, typical_p, repetition_penalty, top_k = create_settings_menus()
 
         name1 = gr.Textbox(value=settings[f'name1{suffix}'], lines=1, label='Your name')
         name2 = gr.Textbox(value=settings[f'name2{suffix}'], lines=1, label='Bot\'s name')
@@ -727,7 +787,7 @@ if args.chat or args.cai_chat:
         if args.extensions is not None:
             create_extensions_block()
 
-        input_params = [textbox, length_slider, preset_menu, model_menu, name1, name2, context, check, history_size_slider]
+        input_params = [textbox, max_new_tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k, name1, name2, context, check, history_size_slider]
         if args.cai_chat:
             gen_events.append(buttons["Generate"].click(cai_chatbot_wrapper, input_params, display, show_progress=args.no_stream, api_name="textgen"))
             gen_events.append(textbox.submit(cai_chatbot_wrapper, input_params, display, show_progress=args.no_stream))
@@ -768,25 +828,19 @@ elif args.notebook:
             markdown = gr.Markdown()
         with gr.Tab('HTML'):
             html = gr.HTML()
+
         buttons["Generate"] = gr.Button("Generate")
         buttons["Stop"] = gr.Button("Stop")
 
-        length_slider = gr.Slider(minimum=settings['max_new_tokens_min'], maximum=settings['max_new_tokens_max'], step=1, label='max_new_tokens', value=settings['max_new_tokens'])
-        with gr.Row():
-            with gr.Column():
-                with gr.Row():
-                    model_menu = gr.Dropdown(choices=available_models, value=model_name, label='Model')
-                    create_refresh_button(model_menu, lambda : None, lambda : {"choices": get_available_models()}, "refresh-button")
-            with gr.Column():
-                with gr.Row():
-                    preset_menu = gr.Dropdown(choices=available_presets, value=settings['preset'], label='Generation parameters preset')
-                    create_refresh_button(preset_menu, lambda : None, lambda : {"choices": get_available_presets()}, "refresh-button")
+        max_new_tokens = gr.Slider(minimum=settings['max_new_tokens_min'], maximum=settings['max_new_tokens_max'], step=1, label='max_new_tokens', value=settings['max_new_tokens'])
+
+        preset_menu, do_sample, temperature, top_p, typical_p, repetition_penalty, top_k = create_settings_menus()
 
         if args.extensions is not None:
             create_extensions_block()
 
-        gen_events.append(buttons["Generate"].click(generate_reply, [textbox, length_slider, preset_menu, model_menu], [textbox, markdown, html], show_progress=args.no_stream, api_name="textgen"))
-        gen_events.append(textbox.submit(generate_reply, [textbox, length_slider, preset_menu, model_menu], [textbox, markdown, html], show_progress=args.no_stream))
+        gen_events.append(buttons["Generate"].click(generate_reply, [textbox, max_new_tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k], [textbox, markdown, html], show_progress=args.no_stream, api_name="textgen"))
+        gen_events.append(textbox.submit(generate_reply, [textbox, max_new_tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k], [textbox, markdown, html], show_progress=args.no_stream))
         buttons["Stop"].click(None, None, None, cancels=gen_events)
 
 else:
@@ -795,19 +849,15 @@ else:
         with gr.Row():
             with gr.Column():
                 textbox = gr.Textbox(value=default_text, lines=15, label='Input')
-                length_slider = gr.Slider(minimum=settings['max_new_tokens_min'], maximum=settings['max_new_tokens_max'], step=1, label='max_new_tokens', value=settings['max_new_tokens'])
-                with gr.Row():
-                    preset_menu = gr.Dropdown(choices=available_presets, value=settings['preset'], label='Generation parameters preset')
-                    create_refresh_button(preset_menu, lambda : None, lambda : {"choices": get_available_presets()}, "refresh-button")
-                with gr.Row():
-                    model_menu = gr.Dropdown(choices=available_models, value=model_name, label='Model')
-                    create_refresh_button(model_menu, lambda : None, lambda : {"choices": get_available_models()}, "refresh-button")
+                max_new_tokens = gr.Slider(minimum=settings['max_new_tokens_min'], maximum=settings['max_new_tokens_max'], step=1, label='max_new_tokens', value=settings['max_new_tokens'])
                 buttons["Generate"] = gr.Button("Generate")
                 with gr.Row():
                     with gr.Column():
                         buttons["Continue"] = gr.Button("Continue")
                     with gr.Column():
                         buttons["Stop"] = gr.Button("Stop")
+
+                preset_menu, do_sample, temperature, top_p, typical_p, repetition_penalty, top_k = create_settings_menus()
                 if args.extensions is not None:
                     create_extensions_block()
 
@@ -819,13 +869,17 @@ else:
                 with gr.Tab('HTML'):
                     html = gr.HTML()
 
-        gen_events.append(buttons["Generate"].click(generate_reply, [textbox, length_slider, preset_menu, model_menu], [output_textbox, markdown, html], show_progress=args.no_stream, api_name="textgen"))
-        gen_events.append(textbox.submit(generate_reply, [textbox, length_slider, preset_menu, model_menu], [output_textbox, markdown, html], show_progress=args.no_stream))
-        gen_events.append(buttons["Continue"].click(generate_reply, [output_textbox, length_slider, preset_menu, model_menu], [output_textbox, markdown, html], show_progress=args.no_stream))
+        gen_events.append(buttons["Generate"].click(generate_reply, [textbox, max_new_tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k], [output_textbox, markdown, html], show_progress=args.no_stream, api_name="textgen"))
+        gen_events.append(textbox.submit(generate_reply, [textbox, max_new_tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k], [output_textbox, markdown, html], show_progress=args.no_stream))
+        gen_events.append(buttons["Continue"].click(generate_reply, [output_textbox, max_new_tokens, do_sample, max_new_tokens, temperature, top_p, typical_p, repetition_penalty, top_k], [output_textbox, markdown, html], show_progress=args.no_stream))
         buttons["Stop"].click(None, None, None, cancels=gen_events)
 
 interface.queue()
 if args.listen:
-    interface.launch(share=args.share, server_name="0.0.0.0", server_port=args.listen_port)
+    interface.launch(prevent_thread_lock=True, share=args.share, server_name="0.0.0.0", server_port=args.listen_port)
 else:
-    interface.launch(share=args.share, server_port=args.listen_port)
+    interface.launch(prevent_thread_lock=True, share=args.share, server_port=args.listen_port)
+
+# I think that I will need this later
+while True:
+    time.sleep(0.5)