githubmirror/text-generation-webui

modules/text_generation.py
547 строк19 KB
oobabooga
Fix --idle-timeout issues with encode/decode and parallel generation
25 мар 2026, 16:37
25 мар 2026, 16:37368f373
import ast
import copy
import html
import pprint
import random
import time

import numpy as np

import modules.shared as shared
from modules import models
from modules.callbacks import Iteratorize
from modules.extensions import apply_extensions
from modules.html_generator import generate_basic_html
from modules.logging_colors import logger
from modules.utils import check_model_loaded


def generate_reply(*args, **kwargs):
    models.load_model_if_idle_unloaded()

    state = args[1] if len(args) > 1 else kwargs.get('state', {})
    use_parallel = (
        state.get('stop_event') is not None
        and shared.model.__class__.__name__ in ['Exllamav3Model', 'LlamaServer', 'TensorRTLLMModel']
        and (shared.model.__class__.__name__ != 'LlamaServer' or shared.args.parallel > 1)
    )

    if not use_parallel:
        shared.generation_lock.acquire()

    with models._generation_count_lock:
        models.active_generation_count += 1

    try:
        for result in _generate_reply(*args, **kwargs):
            yield result
    finally:
        with models._generation_count_lock:
            models.active_generation_count -= 1

        models.last_generation_time = time.time()
        if not use_parallel:
            shared.generation_lock.release()


def _generate_reply(question, state, stopping_strings=None, is_chat=False, escape_html=False, for_ui=False):
    # Find the appropriate generation function
    generate_func = apply_extensions('custom_generate_reply')
    if generate_func is None:
        model_is_loaded, error_message = check_model_loaded()
        if not model_is_loaded:
            yield ''
            return

        if shared.model.__class__.__name__ in ['LlamaServer', 'Exllamav3Model', 'TensorRTLLMModel']:
            generate_func = generate_reply_custom
        else:
            generate_func = generate_reply_HF

    if generate_func != generate_reply_HF and shared.args.verbose:
        logger.info("PROMPT=")
        print_prompt(question)

    # Prepare the input
    original_question = question
    if not is_chat:
        state = apply_extensions('state', state)
        question = apply_extensions('input', question, state)

    # Find the stopping strings
    all_stop_strings = []
    for st in (stopping_strings, state['custom_stopping_strings']):
        if type(st) is str:
            st = ast.literal_eval(f"[{st}]")

        if type(st) is list and len(st) > 0:
            all_stop_strings += st

    shared.stop_everything = False
    reply = ''
    is_stream = state['stream']
    if len(all_stop_strings) > 0 and not state['stream']:
        original_logits_processor = state.get('logits_processor')
        stop_event_ref = state.pop('stop_event', None)
        state = copy.deepcopy(state)
        if stop_event_ref is not None:
            state['stop_event'] = stop_event_ref
        if original_logits_processor is not None:
            state['logits_processor'] = original_logits_processor
        state['stream'] = True

    # Generate
    last_update = -1
    latency_threshold = 1 / 1000
    for reply in generate_func(question, original_question, state, stopping_strings, is_chat=is_chat):
        cur_time = time.monotonic()
        reply, stop_found = apply_stopping_strings(reply, all_stop_strings)
        if escape_html:
            reply = html.escape(reply)

        if is_stream:
            # Limit number of tokens/second to make text readable in real time
            if state['max_tokens_second'] > 0:
                diff = 1 / state['max_tokens_second'] - (cur_time - last_update)
                if diff > 0:
                    time.sleep(diff)

                last_update = time.monotonic()
                yield reply

            # Limit updates to avoid lag in the Gradio UI
            # API updates are not limited
            else:
                # If 'generate_func' takes less than 0.001 seconds to yield the next token
                # (equivalent to more than 1000 tok/s), assume that the UI is lagging behind and skip yielding
                if (cur_time - last_update) > latency_threshold:
                    yield reply
                last_update = time.monotonic()

        stop_event = state.get('stop_event')
        if stop_found or shared.stop_everything or (stop_event and stop_event.is_set()):
            break

    if not is_chat:
        reply = apply_extensions('output', reply, state)

    yield reply


def encode(prompt, add_special_tokens=True, add_bos_token=True, truncation_length=None):
    if shared.tokenizer is None:
        models.load_model_if_idle_unloaded()
        if shared.tokenizer is None:
            raise ValueError('No tokenizer is loaded')

    # llama.cpp case
    if shared.model.__class__.__name__ == 'LlamaServer':
        input_ids = shared.tokenizer.encode(str(prompt), add_bos_token=add_bos_token)
        input_ids = np.array(input_ids).reshape(1, len(input_ids))

        if truncation_length is not None:
            input_ids = input_ids[:, -truncation_length:]

        return input_ids

    # All other model types
    else:
        import torch

        from modules.torch_utils import get_device

        if shared.model.__class__.__name__ in ['Exllamav3Model', 'TensorRTLLMModel']:
            input_ids = shared.tokenizer.encode(str(prompt))
            if shared.model.__class__.__name__ not in ['Exllamav3Model']:
                input_ids = np.array(input_ids).reshape(1, len(input_ids))
        else:
            input_ids = shared.tokenizer.encode(str(prompt), return_tensors='pt', add_special_tokens=add_special_tokens)
            if hasattr(shared.tokenizer, 'bos_token_id') and shared.tokenizer.bos_token_id is not None:
                if add_bos_token:
                    # Add BOS token if missing
                    if (len(input_ids[0]) > 0 and input_ids[0][0] != shared.tokenizer.bos_token_id) or len(input_ids[0]) == 0:
                        bos_tensor = torch.tensor([[shared.tokenizer.bos_token_id]])
                        input_ids = torch.cat((bos_tensor, input_ids), 1)

                # Always prevent double BOS tokens (regardless of add_bos_token setting)
                while len(input_ids[0]) > 1 and input_ids[0][0] == shared.tokenizer.bos_token_id and input_ids[0][1] == shared.tokenizer.bos_token_id:
                    input_ids = input_ids[:, 1:]

        if truncation_length is not None:
            input_ids = input_ids[:, -truncation_length:]

        if shared.model.__class__.__name__ in ['Exllamav3Model', 'TensorRTLLMModel'] or shared.args.cpu:
            return input_ids
        else:
            device = get_device()
            if device:
                return input_ids.to(device)

            return input_ids


def decode(output_ids, skip_special_tokens=True):
    if shared.tokenizer is None:
        models.load_model_if_idle_unloaded()
        if shared.tokenizer is None:
            raise ValueError('No tokenizer is loaded')

    return shared.tokenizer.decode(output_ids, skip_special_tokens=skip_special_tokens)


def get_encoded_length(prompt):
    length_after_extensions = apply_extensions('tokenized_length', prompt)
    if length_after_extensions is not None:
        return length_after_extensions

    return len(encode(prompt)[0])


def get_token_ids(prompt):
    tokens = encode(prompt)[0]
    decoded_tokens = [shared.tokenizer.decode([int(i)]) for i in tokens]

    output = ''
    for row in list(zip(tokens, decoded_tokens)):
        output += f"{str(int(row[0])).ljust(5)}  -  {repr(row[1])}\n"

    return output


def get_max_prompt_length(state):
    return state['truncation_length'] - state['max_new_tokens']


def generate_reply_wrapper(question, state, stopping_strings=None):
    """
    Returns formatted outputs for the UI
    """
    reply = question if not shared.is_seq2seq else ''
    yield formatted_outputs(reply, shared.model_name)

    for reply in generate_reply(question, state, stopping_strings, is_chat=False, escape_html=True, for_ui=True):
        if not shared.is_seq2seq:
            reply = question + reply

        yield formatted_outputs(reply, shared.model_name)


def formatted_outputs(reply, model_name):
    return html.unescape(reply), generate_basic_html(reply)


def set_manual_seed(seed):
    seed = int(seed)
    if seed == -1:
        seed = random.randint(1, 2**31)

    if shared.args.loader != 'llama.cpp':
        import torch
        from transformers import is_torch_npu_available, is_torch_xpu_available

        torch.manual_seed(seed)
        if torch.cuda.is_available():
            torch.cuda.manual_seed_all(seed)
        elif is_torch_xpu_available():
            torch.xpu.manual_seed_all(seed)
        elif is_torch_npu_available():
            torch.npu.manual_seed_all(seed)

    return seed


def stop_everything_event():
    shared.stop_everything = True


def apply_stopping_strings(reply, all_stop_strings):
    stop_found = False
    for string in all_stop_strings:
        idx = reply.find(string)
        if idx != -1:
            reply = reply[:idx]
            stop_found = True
            break

    if not stop_found:
        # If something like "\nYo" is generated just before "\nYou:"
        # is completed, trim it
        for string in all_stop_strings:
            for j in range(len(string) - 1, 0, -1):
                if reply[-j:] == string[:j]:
                    reply = reply[:-j]
                    break
            else:
                continue

            break

    return reply, stop_found


def get_reply_from_output_ids(output_ids, state=None, starting_from=0):
    import torch

    if torch.cuda.is_available():
        torch.cuda.synchronize()

    reply = decode(output_ids[starting_from:], state['skip_special_tokens'] if state else True)

    # Handle tokenizers that do not add the leading space for the first token
    if (hasattr(shared.tokenizer, 'convert_ids_to_tokens') and len(output_ids) > starting_from) and not reply.startswith(' '):
        first_token = shared.tokenizer.convert_ids_to_tokens(int(output_ids[starting_from]))
        if isinstance(first_token, (bytes,)):
            # try to decode the bytes to a string
            # if it fails, which means it's not a string in this turn, just ignore it
            try:
                first_token = first_token.decode('utf8')
            except UnicodeDecodeError:
                first_token = ''

        if first_token.startswith('▁'):
            reply = ' ' + reply

    return reply


def generate_reply_HF(question, original_question, state, stopping_strings=None, is_chat=False):
    import torch
    import transformers
    from transformers import LogitsProcessorList

    from modules.grammar.grammar_utils import initialize_grammar
    from modules.grammar.logits_process import (
        GrammarConstrainedLogitsProcessor
    )
    from modules.torch_utils import clear_torch_cache, get_device
    from modules.transformers_loader import (
        Stream,
        _StopEverythingStoppingCriteria
    )

    if shared.args.loader == 'Transformers':
        clear_torch_cache()

    seed = set_manual_seed(state['seed'])

    generate_params = {}
    for k in [
        'temperature',
        'dynatemp_low',
        'dynatemp_high',
        'dynatemp_exponent',
        'smoothing_factor',
        'smoothing_curve',
        'min_p',
        'top_p',
        'top_k',
        'typical_p',
        'xtc_threshold',
        'xtc_probability',
        'tfs',
        'top_a',
        'top_n_sigma',
        'adaptive_target',
        'adaptive_decay',
        'dry_multiplier',
        'dry_allowed_length',
        'dry_base',
        'repetition_penalty',
        'frequency_penalty',
        'presence_penalty',
        'encoder_repetition_penalty',
        'no_repeat_ngram_size',
        'repetition_penalty_range',
        'penalty_alpha',
        'guidance_scale',
        'mirostat_mode',
        'mirostat_tau',
        'mirostat_eta',
        'max_new_tokens',
        'do_sample',
        'dynamic_temperature',
        'temperature_last',
        'dry_sequence_breakers',
    ]:
        if k in state:
            generate_params[k] = state[k]

    for k in ['epsilon_cutoff', 'eta_cutoff']:
        if state[k] > 0:
            generate_params[k] = state[k] * 1e-4

    if state['prompt_lookup_num_tokens'] > 0:
        generate_params['prompt_lookup_num_tokens'] = state['prompt_lookup_num_tokens']

    if state['ban_eos_token']:
        generate_params['suppress_tokens'] = [shared.tokenizer.eos_token_id]

    if state['static_cache']:
        generate_params['cache_implementation'] = 'static'

    if isinstance(state['sampler_priority'], list) and len(state['sampler_priority']) > 0:
        generate_params['sampler_priority'] = state['sampler_priority']
    elif isinstance(state['sampler_priority'], str) and state['sampler_priority'].strip() != '':
        generate_params['sampler_priority'] = [x.strip() for x in state['sampler_priority'].replace('\n', ',').split(',') if x.strip()]

    if state['custom_token_bans']:
        to_ban = [int(x.strip()) for x in state['custom_token_bans'].split(',') if x.strip()]
        if len(to_ban) > 0:
            if generate_params.get('suppress_tokens', None):
                generate_params['suppress_tokens'] += to_ban
            else:
                generate_params['suppress_tokens'] = to_ban

    if state['negative_prompt'] != '':
        generate_params['negative_prompt_ids'] = encode(state['negative_prompt'])

    generate_params.update({'use_cache': not shared.args.no_cache})

    # Encode the input
    input_ids = encode(question, add_bos_token=state['add_bos_token'], truncation_length=get_max_prompt_length(state))
    output = input_ids[0]
    if state['auto_max_new_tokens']:
        generate_params['max_new_tokens'] = state['truncation_length'] - input_ids.shape[-1]

    # Add the encoded tokens to generate_params
    question, input_ids, inputs_embeds = apply_extensions('tokenizer', state, question, input_ids, None)
    original_input_ids = input_ids
    generate_params.update({'inputs': input_ids})
    if inputs_embeds is not None:
        generate_params.update({'inputs_embeds': inputs_embeds})

    # Stopping criteria / eos token
    eos_token_ids = [shared.tokenizer.eos_token_id] if shared.tokenizer.eos_token_id is not None else []
    generate_params['eos_token_id'] = eos_token_ids
    generate_params['stopping_criteria'] = transformers.StoppingCriteriaList()
    generate_params['stopping_criteria'].append(_StopEverythingStoppingCriteria())

    # Logits processor
    processor = state.get('logits_processor', LogitsProcessorList([]))
    if not isinstance(processor, LogitsProcessorList):
        processor = LogitsProcessorList([processor])

    # Grammar
    if state['grammar_string'].strip() != '':
        grammar = initialize_grammar(state['grammar_string'])
        grammar_processor = GrammarConstrainedLogitsProcessor(grammar)
        processor.append(grammar_processor)

    apply_extensions('logits_processor', processor, input_ids)
    generate_params['logits_processor'] = processor

    if shared.args.verbose:
        logger.info("GENERATE_PARAMS=")
        filtered_params = {key: value for key, value in generate_params.items() if not isinstance(value, torch.Tensor)}
        pprint.PrettyPrinter(indent=4, sort_dicts=False).pprint(filtered_params)
        print()

        logger.info("PROMPT=")
        print_prompt(decode(input_ids[0], skip_special_tokens=False))

    t0 = time.time()
    try:
        if not is_chat and not shared.is_seq2seq:
            yield ''

        # Generate the entire reply at once.
        if not state['stream']:
            with torch.no_grad():
                output = shared.model.generate(**generate_params)[0]
                device = get_device()
                if device:
                    output = output.to(device)

            starting_from = 0 if shared.is_seq2seq else len(input_ids[0])
            yield get_reply_from_output_ids(output, state, starting_from=starting_from)

        # Stream the reply 1 token at a time.
        # This is based on the trick of using 'stopping_criteria' to create an iterator.
        else:

            def generate_with_callback(callback=None, *args, **kwargs):
                kwargs['stopping_criteria'].append(Stream(callback_func=callback))
                with torch.no_grad():
                    shared.model.generate(**kwargs)

            def generate_with_streaming(**kwargs):
                return Iteratorize(generate_with_callback, [], kwargs, callback=None)

            with generate_with_streaming(**generate_params) as generator:
                cumulative_reply = ''
                starting_from = 0 if shared.is_seq2seq else len(input_ids[0])
                for output in generator:
                    if output[-1] in eos_token_ids:
                        break

                    new_content = get_reply_from_output_ids(output, state, starting_from=starting_from)
                    # check the partial unicode character
                    if chr(0xfffd) in new_content:
                        continue

                    cumulative_reply += new_content
                    starting_from = len(output)
                    yield cumulative_reply

    except Exception:
        logger.exception("Failed to generate reply (HF)")
    finally:
        t1 = time.time()
        original_tokens = len(original_input_ids[0])
        new_tokens = len(output) - (original_tokens if not shared.is_seq2seq else 0)
        logger.info(f'Output generated in {(t1-t0):.2f} seconds ({new_tokens/(t1-t0):.2f} tokens/s, {new_tokens} tokens, context {original_tokens}, seed {seed})')
        return


def generate_reply_custom(question, original_question, state, stopping_strings=None, is_chat=False):
    """
    For models that do not use the transformers library for sampling
    """

    stop_event_ref = state.pop('stop_event', None)
    state = copy.deepcopy(state)
    if stop_event_ref is not None:
        state['stop_event'] = stop_event_ref
    state['seed'] = set_manual_seed(state['seed'])
    t0 = time.time()
    reply = ''
    try:
        if not is_chat:
            yield ''

        if not state['stream']:
            reply = shared.model.generate(question, state)
            yield reply
        else:
            for reply in shared.model.generate_with_streaming(question, state):
                yield reply

    except Exception:
        logger.exception("Failed to generate reply (custom)")
    finally:
        t1 = time.time()

        if hasattr(shared.model, 'last_prompt_token_count'):
            original_tokens = shared.model.last_prompt_token_count
            new_tokens = len(encode(reply)[0]) if reply else 0
        else:
            original_tokens = len(encode(original_question)[0])
            new_tokens = len(encode(original_question + reply)[0]) - original_tokens

        logger.info(f'Output generated in {(t1-t0):.2f} seconds ({new_tokens/(t1-t0):.2f} tokens/s, {new_tokens} tokens, context {original_tokens}, seed {state["seed"]})')
        return


def print_prompt(prompt, max_chars=-1):
    DARK_YELLOW = "\033[38;5;3m"
    RESET = "\033[0m"

    if max_chars > 0 and len(prompt) > max_chars:
        half_chars = max_chars // 2
        hidden_len = len(prompt[half_chars:-half_chars])
        hidden_msg = f"{DARK_YELLOW}[...{hidden_len} characters hidden...]{RESET}"
        print(prompt[:half_chars] + hidden_msg + prompt[-half_chars:])
    else:
        print(prompt)

    print()