chatbot with llm using mlx

mlx_chatbot.py

import time
import json
import argparse
from pathlib import Path
from dataclasses import dataclass
from typing import Optional, Tuple
from os.path import abspath, expanduser
from sentencepiece import SentencePieceProcessor

import mlx.core as mx
import mlx.nn as nn
from mlx.utils import tree_unflatten


@dataclass
class ModelArgs:
    dim: int
    n_layers: int
    head_dim: int
    hidden_dim: int
    n_heads: int
    n_kv_heads: int
    norm_eps: float
    vocab_size: int
    rope_theta: float
    rope_traditional: bool = True


class RMSNorm(nn.Module):
    def __init__(self, dims: int, eps: float = 1e-5):
        super().__init__()
        self.weight = mx.ones((dims,))
        self.eps = eps

    def _norm(self, x):
        return x * mx.rsqrt(x.square().mean(-1, keepdims=True) + self.eps)

    def __call__(self, x):
        output = self._norm(x.astype(mx.float32)).astype(x.dtype)
        return self.weight * output


class RoPE(nn.RoPE):
    def __init__(self, dims: int, traditional: bool = False, base: float = 1e4):
        super().__init__(dims, traditional)
        self.base = base

    def __call__(self, x, offset: int = 0):
        shape = x.shape
        x = mx.reshape(x, (-1, shape[-2], shape[-1]))
        N = x.shape[1] + offset
        cos_theta, sin_theta = RoPE.create_cos_sin_theta(
            N, self.dims, offset=offset, base=self.base, dtype=x.dtype
        )
        rope = (
            self._compute_traditional_rope if self.traditional else self._compute_rope
        )
        rx = rope(cos_theta, sin_theta, x)

        return mx.reshape(rx, shape)


class Attention(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()
        self.args = args

        self.n_heads: int = args.n_heads
        self.n_kv_heads: int = args.n_kv_heads

        self.repeats = self.n_heads // self.n_kv_heads

        self.scale = self.args.head_dim ** -0.5

        self.wq = nn.Linear(args.dim, args.n_heads * args.head_dim, bias=False)
        self.wk = nn.Linear(args.dim, args.n_kv_heads * args.head_dim, bias=False)
        self.wv = nn.Linear(args.dim, args.n_kv_heads * args.head_dim, bias=False)
        self.wo = nn.Linear(args.n_heads * args.head_dim, args.dim, bias=False)
        self.rope = RoPE(
            args.head_dim, traditional=args.rope_traditional, base=args.rope_theta
        )

    def __call__(
            self,
            x: mx.array,
            mask: Optional[mx.array] = None,
            cache: Optional[Tuple[mx.array, mx.array]] = None,
    ) -> mx.array:
        B, L, D = x.shape

        queries, keys, values = self.wq(x), self.wk(x), self.wv(x)

        queries = queries.reshape(B, L, self.n_heads, -1).transpose(0, 2, 1, 3)
        keys = keys.reshape(B, L, self.n_kv_heads, -1).transpose(0, 2, 1, 3)
        values = values.reshape(B, L, self.n_kv_heads, -1).transpose(0, 2, 1, 3)

        def repeat(a):
            a = mx.concatenate([mx.expand_dims(a, 2)] * self.repeats, axis=2)
            return a.reshape([B, self.n_heads, L, -1])

        keys, values = map(repeat, (keys, values))

        if cache is not None:
            key_cache, value_cache = cache
            queries = self.rope(queries, offset=key_cache.shape[2])
            keys = self.rope(keys, offset=key_cache.shape[2])
            keys = mx.concatenate([key_cache, keys], axis=2)
            values = mx.concatenate([value_cache, values], axis=2)
        else:
            queries = self.rope(queries)
            keys = self.rope(keys)

        scores = (queries * self.scale) @ keys.transpose(0, 1, 3, 2)
        if mask is not None:
            scores += mask
        scores = mx.softmax(scores.astype(mx.float32), axis=-1).astype(scores.dtype)
        output = (scores @ values).transpose(0, 2, 1, 3).reshape(B, L, -1)
        return self.wo(output), (keys, values)


class FeedForward(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()

        self.w1 = nn.Linear(args.dim, args.hidden_dim, bias=False)
        self.w2 = nn.Linear(args.hidden_dim, args.dim, bias=False)
        self.w3 = nn.Linear(args.dim, args.hidden_dim, bias=False)

    def __call__(self, x) -> mx.array:
        return self.w2(nn.silu(self.w1(x)) * self.w3(x))


class TransformerBlock(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()
        self.n_heads = args.n_heads
        self.dim = args.dim
        self.attention = Attention(args)
        self.feed_forward = FeedForward(args=args)
        self.attention_norm = RMSNorm(args.dim, eps=args.norm_eps)
        self.ffn_norm = RMSNorm(args.dim, eps=args.norm_eps)
        self.args = args

    def __call__(
            self,
            x: mx.array,
            mask: Optional[mx.array] = None,
            cache: Optional[Tuple[mx.array, mx.array]] = None,
    ) -> mx.array:
        r, cache = self.attention(self.attention_norm(x), mask, cache)
        h = x + r
        r = self.feed_forward(self.ffn_norm(h))
        out = h + r
        return out, cache


class Llama(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()
        self.args = args
        self.vocab_size = args.vocab_size
        self.tok_embeddings = nn.Embedding(args.vocab_size, args.dim)
        self.layers = [TransformerBlock(args=args) for _ in range(args.n_layers)]
        self.norm = RMSNorm(args.dim, eps=args.norm_eps)
        self.output = nn.Linear(args.dim, args.vocab_size, bias=False)

    def __call__(self, x):
        mask = nn.MultiHeadAttention.create_additive_causal_mask(x.shape[1])
        mask = mask.astype(self.tok_embeddings.weight.dtype)

        x = self.tok_embeddings(x)
        for l in self.layers:
            x, _ = l(x, mask)
        x = self.norm(x)
        return self.output(x)

    def generate(self, x, temp=1.0):
        def sample(logits):
            if temp == 0:
                return mx.argmax(logits, axis=-1)
            else:
                return mx.random.categorical(logits * (1 / temp))

        cache = []

        # Make an additive causal mask. We will need that to process the prompt.
        mask = nn.MultiHeadAttention.create_additive_causal_mask(x.shape[1])
        mask = mask.astype(self.tok_embeddings.weight.dtype)

        # First we process the prompt x the same was as in __call__ but
        # save the caches in cache
        x = self.tok_embeddings(x)
        for l in self.layers:
            x, c = l(x, mask=mask)
            # We store the per layer cache in a simple python list
            cache.append(c)
        x = self.norm(x)
        # We only care about the last logits that generate the next token
        y = self.output(x[:, -1])
        y = sample(y)

        # y now has size [1]
        # Since MLX is lazily evaluated nothing is computed yet.
        # Calling y.item() would force the computation to happen at
        # this point but we can also choose not to do that and let the
        # user choose when to start the computation.
        yield y

        # Now we parsed the prompt and generated the first token we
        # need to feed it back into the model and loop to generate the
        # rest.
        while True:
            # Unsqueezing the last dimension to add a sequence length
            # dimension of 1
            x = y[:, None]

            x = self.tok_embeddings(x)
            for i in range(len(cache)):
                # We are overwriting the arrays in the cache list. When
                # the computation will happen, MLX will be discarding the
                # old cache the moment it is not needed anymore.
                x, cache[i] = self.layers[i](x, mask=None, cache=cache[i])
            x = self.norm(x)
            y = sample(self.output(x[:, -1]))

            yield y


def tic():
    return time.time()


def toc(msg, start):
    end = time.time()
    return f"[INFO] {msg}: {end - start:.3f} s"


def generate(question: str, regen: bool = False) -> bool:
    if not regen:
        print("-----Bot-----")
        print("Bot: ", end="")
    x = mx.array([[tokenizer.bos_id()] + tokenizer.encode(question)])
    skip = 0
    prompt_processing = None
    tokens = []
    start = tic()
    start_idx = 0
    fin_idx = 0
    next_turn = False
    for token in model.generate(x, arguments.temp):
        tokens.append(token)

        if len(tokens) == 1:
            # Actually perform the computation to measure the prompt processing time
            mx.eval(token)
            prompt_processing = toc("Prompt processing", start)

        if len(tokens) >= arguments.num_tokens:
            break

        elif (len(tokens) % arguments.write_every) == 0:
            # It is perfectly ok to eval things we have already eval-ed.
            mx.eval(tokens)
            s = tokenizer.decode([t.item() for t in tokens if 32000 > t.item()])
            if "</fin>" in s:
                fin_idx = s.index("</fin>")
                if s != "</fin>":
                    next_turn = True
                break
            if "<" not in s:
                print(s[skip:], end="", flush=True)
            elif s.startswith("<"):
                print(s[skip-5:-5], end="", flush=True)
            elif "</s" in s:
                print(s[skip:-4], end="", flush=True)
            else:
                print(s[skip:-5], end="", flush=True)
            skip = len(s)
            if "<fin>" in s and skip <= s.index("<fin>") + 5:
                skip += 5
                start_idx = 5
            elif "</s>" in s and skip <= s.index("</s>") + 4:
                skip += 4
                start_idx = 4

    mx.eval(tokens)
    full_gen = toc("Full generation", start)
    s = tokenizer.decode([t.item() for t in tokens if 32000 > t.item()])[start_idx:fin_idx]
    if next_turn:
        chat_log.append({
            "role": "assistant",
            "content": s,
        })
        print(s[skip:], end="", flush=True)
        print()
        print(prompt_processing)
        print(full_gen)

    return next_turn


def gen_dialogue(preset: str = "") -> str:
    global chat_log
    string_dialogue = preset
    if len(chat_log) > log_limit * 2:
        # string_dialogue = ""
        chat_log = chat_log[:log_limit * -2]
    for dict_message in chat_log:
        if dict_message["role"] == "user":
            if arguments.ko:
                string_dialogue += "User: " + dict_message["content"] + "\\n\\n"
            else:
                if dict_message == chat_log[0]:
                    start = "\n"
                else:
                    start = "\n<s>\n[INST]"
                string_dialogue += start + dict_message["content"] + " [/INST]"
        else:
            if arguments.ko:
                string_dialogue += "Assistant: " + dict_message["content"] + "\\n\\n"
            else:
                string_dialogue += "\n" + dict_message["content"] + "\n</s>"
    return string_dialogue


def generate_llm_response():
    input("Press enter to start generation")
    if arguments.ko:
        dialogue_preset = ""
    else:
        dialogue_preset = "<s>[INST]<<SYS>>\n" \
                          "You are a helpful assistant.\n" \
                          "You do not respond as 'User' or pretend to be 'User'.\n" \
                          "You only respond once as 'Assistant'.\n" \
                          "You only respond about the user's question.\n" \
                          "You do not go off topic. \n" \
                          "You must finish your response with '</fin>'.\n" \
                          "You have not start your response with '<fin>'.\n" \
                          f"You must finish your response in {arguments.num_tokens - 1} words.\n" \
                          "<</SYS>>\n"
    # string_dialogue = gen_dialogue(dialogue_preset)
    while True:
        print("-----User-----")
        question = input("User: ")
        chat_log.append({
            "role": "user",
            "content": question,
        })
        string_dialogue = gen_dialogue(dialogue_preset)
        if question == "Good bye LLM!":
            print("-----Bot-----")
            print("Bot: Good bye Human!")
            break
        regen = True
        if arguments.ko:
            while True:
                if regen := generate(f"{string_dialogue} ", not regen):
                    break
        else:
            while True:
                if regen := generate(f"{string_dialogue}\n", not regen):
                    break


def sanitize_config(config, weights):
    config.pop("model_type", None)
    n_heads = config["n_heads"]
    if "n_kv_heads" not in config:
        config["n_kv_heads"] = n_heads
    if "head_dim" not in config:
        config["head_dim"] = config["dim"] // n_heads
    if "hidden_dim" not in config:
        config["hidden_dim"] = weights["layers.0.feed_forward.w1.weight"].shape[0]
    if config.get("vocab_size", -1) < 0:
        config["vocab_size"] = weights["output.weight"].shape[-1]
    if "rope_theta" not in config:
        config["rope_theta"] = 10000
    unused = ["multiple_of", "ffn_dim_multiplier"]
    for k in unused:
        config.pop(k, None)
    return config


def load_model(model_path):
    model_path = Path(model_path)
    weights = mx.load(str(model_path / "weights.npz"))
    with open(model_path / "config.json", "r") as f:
        config = sanitize_config(json.loads(f.read()), weights)
        quantization = config.pop("quantization", None)
    loaded_model = Llama(ModelArgs(**config))
    if quantization is not None:
        nn.QuantizedLinear.quantize_module(loaded_model, **quantization)
    loaded_model.update(tree_unflatten(list(weights.items())))
    loaded_tokenizer = SentencePieceProcessor(model_file=str(model_path / "tokenizer.model"))
    return loaded_model, loaded_tokenizer


if __name__ == '__main__':
    parser = argparse.ArgumentParser(description="Llama inference script")
    parser.add_argument(
        'model', help='Path to the model directory containing the MLX weights'
    )
    parser.add_argument(
        '--num-tokens', '-n', type=int, default=100, help='How many tokens to generate'
    )
    parser.add_argument(
        '--write-every', type=int, default=1, help='After how many tokens to detokenize'
    )
    parser.add_argument(
        '--temp', type=float, default=0.8, help='The sampling temperature'
    )
    parser.add_argument('--seed', type=int, default=0, help='The PRNG seed')
    parser.add_argument('--log-limit', type=int, default=5, help='The number of saved chat logs to use in LLM.')
    parser.add_argument('--ko', '-k', action='store_true', default=False)

    arguments = parser.parse_args()

    mx.random.seed(arguments.seed)

    log_limit = arguments.log_limit

    # chat_log = [{"role": "assistant", "content": "How may I assist you today?"}]
    chat_log = []

    print('[INFO] Loading model from disk.')
    model, tokenizer = load_model(abspath(expanduser(arguments.model)))
    generate_llm_response()