fish-speech/tools/llama/generate.py

import os
import time
from pathlib import Path
from typing import Optional, Tuple

import click
import numpy as np
import torch
import torch._dynamo.config
import torch._inductor.config
from hydra import compose, initialize
from hydra.utils import instantiate
from loguru import logger
from tqdm import tqdm
from transformers import AutoTokenizer

from fish_speech.text.parser import clean_text

os.environ["TOKENIZERS_PARALLELISM"] = "false"
torch._inductor.config.coordinate_descent_tuning = True
torch._inductor.config.triton.unique_kernel_names = True
torch._inductor.config.fx_graph_cache = True  # Experimental feature to reduce compilation times, will be on by default in future


from fish_speech.models.text2semantic.llama import Transformer
from fish_speech.text import g2p
from fish_speech.text.symbols import pad as pad_symbol
from fish_speech.text.symbols import pu_symbols


def multinomial_sample_one_no_sync(
    probs_sort,
):  # Does multinomial sampling without a cuda synchronization
    q = torch.empty_like(probs_sort).exponential_(1)
    return torch.argmax(probs_sort / q, dim=-1, keepdim=True).to(dtype=torch.int)


def logits_to_probs(
    logits,
    previous_tokens: Optional[torch.Tensor] = None,
    temperature: float = 1.0,
    top_k: Optional[int] = None,
    top_p: Optional[int] = None,
    repetition_penalty: float = 1.0,
):
    if previous_tokens is not None and repetition_penalty != 1.0:
        previous_tokens = previous_tokens.long()
        score = torch.gather(logits, dim=0, index=previous_tokens)
        score = torch.where(
            score < 0, score * repetition_penalty, score / repetition_penalty
        )
        logits.scatter_(dim=0, index=previous_tokens, src=score)

    if top_p is not None and top_p < 1.0:
        sorted_logits, sorted_indices = torch.sort(logits, descending=True)
        cum_probs = torch.cumsum(
            torch.nn.functional.softmax(sorted_logits, dim=-1), dim=-1
        )
        sorted_indices_to_remove = cum_probs > top_p
        sorted_indices_to_remove[0] = False  # keep at least one option
        indices_to_remove = sorted_indices_to_remove.scatter(
            dim=0, index=sorted_indices, src=sorted_indices_to_remove
        )
        logits = logits.masked_fill(indices_to_remove, -float("Inf"))

    logits = logits / max(temperature, 1e-5)

    if top_k is not None:
        v, _ = torch.topk(logits, min(top_k, logits.size(-1)))
        pivot = v.select(-1, -1).unsqueeze(-1)
        logits = torch.where(logits < pivot, -float("Inf"), logits)

    probs = torch.nn.functional.softmax(logits, dim=-1)
    return probs


def sample(
    logits,
    previous_tokens: Optional[torch.Tensor] = None,
    **sampling_kwargs,
) -> Tuple[torch.Tensor, torch.Tensor]:
    probs = logits_to_probs(
        logits=logits[0, -1], previous_tokens=previous_tokens, **sampling_kwargs
    )
    idx_next = multinomial_sample_one_no_sync(probs)
    return idx_next, probs


def decode_one_token(
    model: Transformer,
    x: torch.Tensor,
    input_pos: torch.Tensor,
    previous_tokens: torch.Tensor = None,
    **sampling_kwargs,
) -> torch.Tensor:
    assert input_pos.shape[-1] == 1

    logits = model.forward_generate(x, input_pos)
    codebooks = [
        sample(
            logits.token_logits,
            previous_tokens=None,  # Disable repetition penalty for the token codebook
            **sampling_kwargs,
        )[0]
    ]

    # Disable <s> and </s> tokens for codebooks
    if model.config.num_codebooks != 0:
        logits.codebook_logits[:, :, :, :2] = -float("Inf")

        for i in range(model.config.num_codebooks):
            codebooks.append(
                sample(
                    logits.codebook_logits[:, :, i],
                    previous_tokens=previous_tokens[i + 1]
                    if previous_tokens is not None
                    else None,
                    **sampling_kwargs,
                )[0]
            )

    return torch.stack(codebooks, dim=0)


def prefill(
    model: Transformer, x: torch.Tensor, input_pos: torch.Tensor, **sampling_kwargs
) -> torch.Tensor:
    # input_pos: [B, S]
    logits = model.forward_generate(x, input_pos)
    codebooks = [
        sample(
            logits.token_logits,
            previous_tokens=None,
            **sampling_kwargs,
        )[0]
    ]

    # Disable <s> and </s> tokens for codebooks
    if model.config.num_codebooks != 0:
        logits.codebook_logits[:, :, :, :2] = -float("Inf")

        for i in range(model.config.num_codebooks):
            codebooks.append(
                sample(
                    logits.codebook_logits[:, :, i],
                    previous_tokens=None,
                    **sampling_kwargs,
                )[0]
            )

    return torch.stack(codebooks, dim=0)


def decode_n_tokens(
    model: Transformer,
    cur_token: torch.Tensor,
    input_pos: torch.Tensor,
    num_new_tokens: int,
    eos_token_id: int = 2,
    **sampling_kwargs,
):
    previous_tokens = torch.zeros(
        (model.config.num_codebooks + 1, num_new_tokens),
        dtype=torch.int,
        device=cur_token.device,
    )

    for i in tqdm(range(num_new_tokens)):
        # We need to get windowed repeat penalty
        win_size = 16
        if i < win_size:
            window = previous_tokens[:, :win_size]
        else:
            window = previous_tokens[:, i - win_size : i]

        with torch.backends.cuda.sdp_kernel(
            enable_flash=False, enable_mem_efficient=False, enable_math=True
        ):  # Actually better for Inductor to codegen attention here
            next_token = decode_one_token(
                model,
                cur_token,
                input_pos,
                window,
                **sampling_kwargs,
            )

        input_pos += 1
        cur_token = next_token.view(1, model.config.num_codebooks + 1, -1)
        previous_tokens[:, i : i + 1] = next_token.view(
            model.config.num_codebooks + 1, -1
        )

        # TODO: use tokenizer's eos
        if (cur_token[0, 0, -1] == eos_token_id).any():
            break

    return previous_tokens[:, : i + 1]


@torch.no_grad()
def generate(
    *,
    model: Transformer,
    prompt: torch.Tensor,
    max_new_tokens: int,
    eos_token_id: int = 2,
    **sampling_kwargs,
) -> torch.Tensor:
    """
    Takes a conditioning sequence (prompt) as input and continues to generate as many tokens as requested.
    """

    # create an empty tensor of the expected final shape and fill in the current tokens
    T = prompt.size(1)

    if max_new_tokens:
        if T + max_new_tokens > model.config.max_seq_len:
            max_new_tokens = model.config.max_seq_len - T
            logger.info(f"Truncating max_new_tokens to {max_new_tokens}")

        T_new = T + max_new_tokens
    else:
        T_new = model.config.max_seq_len
        max_new_tokens = T_new - T

    device, dtype = prompt.device, prompt.dtype
    with torch.device(device):
        model.setup_caches(max_batch_size=1, max_seq_len=T_new)

    codebook_dim = 1 + model.config.num_codebooks
    # create an empty tensor of the expected final shape and fill in the current tokens
    empty = torch.empty((codebook_dim, T_new), dtype=dtype, device=device)
    empty[:, :T] = prompt
    seq = empty
    input_pos = torch.arange(0, T, device=device)

    next_token = prefill(
        model, prompt.view(1, codebook_dim, -1), input_pos, **sampling_kwargs
    )
    seq[:, T : T + 1] = next_token

    input_pos = torch.tensor([T], device=device, dtype=torch.int)
    x = decode_n_tokens(
        model,
        next_token.view(1, codebook_dim, -1),
        input_pos,
        max_new_tokens - 1,
        eos_token_id=eos_token_id,
        **sampling_kwargs,
    )
    # x = torch.cat(generated_tokens, dim=1)
    seq = seq[:, : T + 1 + x.size(1)]
    seq[:, T + 1 :] = x

    return seq


def encode_tokens(
    tokenizer,
    string,
    bos=True,
    device="cuda",
    prompt_text=None,
    prompt_tokens=None,
    use_g2p=False,
    speaker=None,
):
    if prompt_text is not None:
        string = prompt_text + " " + string

    if use_g2p:
        prompt = g2p(string)
        prompt = [
            (f"<p:{i}>" if i not in pu_symbols and i != pad_symbol else i)
            for _, i in prompt
        ]
        string = " ".join(prompt)
    else:
        string = clean_text(string)

    if speaker is not None:
        string = f"[SPK: {speaker}] {string}"

    string = f"[INST] {string} [/INST]"

    tokens = tokenizer.encode(
        string,
        max_length=10**6,
        add_special_tokens=bos,
        truncation=False,
    )
    tokens = torch.tensor([tokens], dtype=torch.int, device=device)

    # Codebooks
    zeros = torch.zeros((4, tokens.size(1)), dtype=torch.int, device=device)
    prompt = torch.cat((tokens, zeros), dim=0)

    if prompt_tokens is None:
        return prompt

    # Get prompt tokens
    assert prompt_tokens.ndim == 2
    data = prompt_tokens + 2

    zeros = (
        torch.zeros((1, data.size(1)), dtype=torch.int, device=device)
        + tokenizer.pad_token_id
    )  # 32311 is the <pad> token
    data = torch.cat((zeros, data), dim=0)
    prompt = torch.cat((prompt, data), dim=1)

    return prompt


def load_model(config_name, checkpoint_path, device, precision):
    with initialize(version_base="1.3", config_path="../../fish_speech/configs"):
        cfg = compose(config_name=config_name)

    with torch.device("meta"):
        model: Transformer = instantiate(cfg.model.model)

    if "int8" in str(checkpoint_path):
        logger.info("Using int8 weight-only quantization!")
        from quantize import WeightOnlyInt8QuantHandler

        simple_quantizer = WeightOnlyInt8QuantHandler(model)
        model = simple_quantizer.convert_for_runtime()

    if "int4" in str(checkpoint_path):
        logger.info("Using int4 quantization!")
        path_comps = checkpoint_path.name.split(".")
        assert path_comps[-2].startswith("g")
        groupsize = int(path_comps[-2][1:])
        from quantize import WeightOnlyInt4QuantHandler

        simple_quantizer = WeightOnlyInt4QuantHandler(model, groupsize)
        model = simple_quantizer.convert_for_runtime()

    checkpoint = torch.load(str(checkpoint_path), map_location="cpu")
    if "state_dict" in checkpoint:
        checkpoint = checkpoint["state_dict"]

    if any(k.startswith("model.") for k in checkpoint):
        checkpoint = {
            k.replace("model.", ""): v
            for k, v in checkpoint.items()
            if k.startswith("model.")
        }

    model.load_state_dict(checkpoint, assign=True)

    model = model.to(device=device, dtype=precision)
    logger.info("Restored model from checkpoint")

    return model.eval()


@click.command()
@click.option("--text", type=str, default="你说的对, 但是原神是一款由米哈游自主研发的开放世界手游.")
@click.option("--prompt-text", type=str, default=None)
@click.option(
    "--prompt-tokens", type=click.Path(path_type=Path, exists=True), default=None
)
@click.option("--num-samples", type=int, default=1)
@click.option("--max_new_tokens", type=int, default=0)
@click.option("--top-k", type=int, default=None)
@click.option("--top-p", type=float, default=0.5)
@click.option("--repetition-penalty", type=float, default=1.5)
@click.option("--temperature", type=float, default=0.7)
@click.option(
    "--checkpoint-path",
    type=click.Path(path_type=Path, exists=True),
    default="results/text2semantic_400m_finetune/step_000002000.pth",
)
@click.option("--config-name", type=str, default="text2semantic_finetune")
@click.option("--tokenizer", type=str, default="fishaudio/speech-lm-v1")
@click.option("--compile/--no-compile", default=False)
@click.option("--use-g2p/--no-g2p", default=True)
@click.option("--seed", type=int, default=42)
@click.option("--speaker", type=str, default=None)
def main(
    text: str,
    prompt_text: Optional[str],
    prompt_tokens: Optional[Path],
    num_samples: int,
    max_new_tokens: int,
    top_k: int,
    top_p: int,
    repetition_penalty: float,
    temperature: float,
    checkpoint_path: Path,
    config_name: str,
    tokenizer: str,
    compile: bool,
    use_g2p: bool,
    seed: int,
    speaker: Optional[str],
) -> None:
    device = "cuda"
    precision = torch.bfloat16

    logger.info("Loading model ...")
    t0 = time.time()
    model = load_model(config_name, checkpoint_path, device, precision)
    model_size = sum(p.numel() for p in model.parameters() if p.requires_grad)

    torch.cuda.synchronize()
    logger.info(f"Time to load model: {time.time() - t0:.02f} seconds")

    tokenizer = AutoTokenizer.from_pretrained(tokenizer)
    prompt_tokens = (
        torch.from_numpy(np.load(prompt_tokens)).to(device)
        if prompt_tokens is not None
        else None
    )
    encoded = encode_tokens(
        tokenizer,
        text,
        prompt_text=prompt_text,
        prompt_tokens=prompt_tokens,
        bos=True,
        device=device,
        use_g2p=use_g2p,
        speaker=speaker,
    )
    prompt_length = encoded.size(1)
    logger.info(f"Encoded prompt shape: {encoded.shape}")

    torch.manual_seed(seed)
    if compile:
        global decode_one_token
        decode_one_token = torch.compile(
            decode_one_token, mode="reduce-overhead", fullgraph=True
        )

    for i in range(num_samples):
        torch.cuda.synchronize()

        t0 = time.perf_counter()
        y = generate(
            model=model,
            prompt=encoded,
            max_new_tokens=max_new_tokens,
            eos_token_id=tokenizer.eos_token_id,
            temperature=temperature,
            top_k=top_k,
            top_p=top_p,
            repetition_penalty=repetition_penalty,
        )

        if i == 0 and compile:
            logger.info(f"Compilation time: {time.perf_counter() - t0:.2f} seconds")

        torch.cuda.synchronize()
        t = time.perf_counter() - t0

        tokens_generated = y.size(1) - prompt_length
        tokens_sec = tokens_generated / t
        logger.info(
            f"Generated {tokens_generated} tokens in {t:.02f} seconds, {tokens_sec:.02f} tokens/sec"
        )
        logger.info(f"Bandwidth achieved: {model_size * tokens_sec / 1e9:.02f} GB/s")
        logger.info(
            f"GPU Memory used: {torch.cuda.max_memory_reserved() / 1e9:.02f} GB"
        )

        codes = y[1:, prompt_length:-1]
        codes = codes - 2
        assert (codes >= 0).all(), "Codes should be >= 0"

        np.save(f"codes_{i}.npy", codes.cpu().numpy())
        logger.info(f"Saved codes to codes_{i}.npy")


if __name__ == "__main__":
    main()