model.py

import argparse
import json
from dataclasses import dataclass
from pathlib import Path
from typing import List, Optional, Tuple

import mlx.core as mx
import mlx.nn as nn
import numpy
import numpy as np
from mlx.utils import tree_unflatten
from transformers import AutoConfig, AutoTokenizer, PreTrainedTokenizerBase


class TransformerEncoderLayer(nn.Module):
    """
    A transformer encoder layer with (the original BERT) post-normalization.
    """

    def __init__(
        self,
        dims: int,
        num_heads: int,
        mlp_dims: Optional[int] = None,
        layer_norm_eps: float = 1e-12,
    ):
        super().__init__()
        mlp_dims = mlp_dims or dims * 4
        self.attention = nn.MultiHeadAttention(dims, num_heads, bias=True)
        self.ln1 = nn.LayerNorm(dims, eps=layer_norm_eps)
        self.ln2 = nn.LayerNorm(dims, eps=layer_norm_eps)
        self.linear1 = nn.Linear(dims, mlp_dims)
        self.linear2 = nn.Linear(mlp_dims, dims)
        self.gelu = nn.GELU()

    def __call__(self, x, mask):
        attention_out = self.attention(x, x, x, mask)
        add_and_norm = self.ln1(x + attention_out)

        ff = self.linear1(add_and_norm)
        ff_gelu = self.gelu(ff)
        ff_out = self.linear2(ff_gelu)
        x = self.ln2(ff_out + add_and_norm)

        return x


class TransformerEncoder(nn.Module):
    def __init__(
        self, num_layers: int, dims: int, num_heads: int, mlp_dims: Optional[int] = None
    ):
        super().__init__()
        self.layers = [
            TransformerEncoderLayer(dims, num_heads, mlp_dims)
            for i in range(num_layers)
        ]

    def __call__(self, x, mask):
        for layer in self.layers:
            x = layer(x, mask)

        return x


class BertEmbeddings(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size)
        self.token_type_embeddings = nn.Embedding(
            config.type_vocab_size, config.hidden_size
        )
        self.position_embeddings = nn.Embedding(
            config.max_position_embeddings, config.hidden_size
        )
        self.norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)

    def __call__(
        self, input_ids: mx.array, token_type_ids: mx.array = None
    ) -> mx.array:
        words = self.word_embeddings(input_ids)
        position = self.position_embeddings(
            mx.broadcast_to(mx.arange(input_ids.shape[1]), input_ids.shape)
        )

        if token_type_ids is None:
            # If token_type_ids is not provided, default to zeros
            token_type_ids = mx.zeros_like(input_ids)

        token_types = self.token_type_embeddings(token_type_ids)

        embeddings = position + words + token_types
        return self.norm(embeddings)


class Bert(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.embeddings = BertEmbeddings(config)
        self.encoder = TransformerEncoder(
            num_layers=config.num_hidden_layers,
            dims=config.hidden_size,
            num_heads=config.num_attention_heads,
            mlp_dims=config.intermediate_size,
        )
        self.pooler = nn.Linear(config.hidden_size, config.hidden_size)

    def __call__(
        self,
        input_ids: mx.array,
        token_type_ids: mx.array = None,
        attention_mask: mx.array = None,
    ) -> Tuple[mx.array, mx.array]:
        x = self.embeddings(input_ids, token_type_ids)

        if attention_mask is not None:
            # convert 0's to -infs, 1's to 0's, and make it broadcastable
            attention_mask = mx.log(attention_mask)
            attention_mask = mx.expand_dims(attention_mask, (1, 2))

        y = self.encoder(x, attention_mask)
        return y, mx.tanh(self.pooler(y[:, 0]))


def load_model(
    bert_model: str, weights_path: str
) -> Tuple[Bert, PreTrainedTokenizerBase]:
    if not Path(weights_path).exists():
        raise ValueError(f"No model weights found in {weights_path}")
    
    # First check if there's a local config
    config_path = Path(weights_path).parent / "config.json"
    if config_path.exists():
        with open(config_path, "r") as f:
            config_dict = json.load(f)
        config = AutoConfig.for_model(**config_dict)
        print(f"Loaded local config from {config_path}")
    else:
        # If no local config, use the HuggingFace one
        config = AutoConfig.from_pretrained(bert_model)
        print(f"Loaded config from HuggingFace for {bert_model}")

    # Create and update the model
    print(f"Creating model with vocab_size={config.vocab_size}, hidden_size={config.hidden_size}")
    model = Bert(config)
    model.load_weights(weights_path)

    tokenizer = AutoTokenizer.from_pretrained(bert_model)

    return model, tokenizer


def run(bert_model: str, mlx_model: str, batch: List[str]):
    import time
    
    # Time model loading
    load_start = time.time()
    model, tokenizer = load_model(bert_model, mlx_model)
    load_time = time.time() - load_start
    print(f"[MLX] Model loaded in {load_time:.2f} seconds")
    
    # Time tokenization
    print(f"[MLX] Tokenizing batch of {len(batch)} sentences")
    token_start = time.time()
    tokens = tokenizer(batch, return_tensors="np", padding=True)
    token_time = time.time() - token_start
    print(f"[MLX] Tokenization completed in {token_time:.4f} seconds")
    
    print(f"[MLX] Tokens shape: input_ids={tokens['input_ids'].shape}")
    tokens = {key: mx.array(v) for key, v in tokens.items()}
    
    # Time inference
    print(f"[MLX] Running model inference")
    infer_start = time.time()
    output, pooled = model(**tokens)
    mx.eval(output, pooled)  # Force evaluation of lazy arrays
    infer_time = time.time() - infer_start
    print(f"[MLX] Inference completed in {infer_time:.4f} seconds")
    
    return output, pooled


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Run the BERT model using MLX.")
    parser.add_argument(
        "--bert-model",
        type=str,
        default="bert-base-uncased",
        help="The huggingface name of the BERT model to save.",
    )
    parser.add_argument(
        "--mlx-model",
        type=str,
        default="weights/bert-base-uncased.npz",
        help="The path of the stored MLX BERT weights (npz file).",
    )
    parser.add_argument(
        "--text",
        type=str,
        default="This is an example of BERT working in MLX",
        help="The text to generate embeddings for.",
    )
    args = parser.parse_args()
    run(args.bert_model, args.mlx_model, args.text)