mlx-examples/llms/mlx_lm/models/mamba2.py

import math
from dataclasses import dataclass, field
from typing import Tuple, Union
import mlx.core as mx
import mlx.nn as nn

from .base import BaseModelArgs
from .cache import MambaCache

@dataclass
class ModelArgs(BaseModelArgs):
    model_type: str
    num_heads: int
    head_dim: int
    vocab_size: int
    hidden_size: int
    state_size: int
    num_hidden_layers: int
    layer_norm_epsilon: float
    expand: int
    conv_kernel: int
    n_groups: int
    use_bias: bool
    use_conv_bias: bool
    initializer_range: float
    residual_in_fp32: bool
    chunk_size: int
    tie_word_embeddings: bool
    time_step_limit: Tuple[float, float]
    time_step_rank: Union[int, str]
    time_step_min: float
    time_step_max: float
    time_step_floor: float
    norm_before_gate: bool = True

    def __post_init__(self):
        if not hasattr(self, "intermediate_size"):
            self.intermediate_size = int(self.expand * self.hidden_size)
        if not hasattr(self, "head_dim"):
            self.head_dim = self.hidden_size // self.num_heads
        if self.time_step_rank == "auto":
            self.time_step_rank = math.ceil(self.hidden_size / 16)


def ssd_forward_attn(
    x: mx.array,
    dt: mx.array,
    A: mx.array,
    B: mx.array,
    C: mx.array,
    D: mx.array,
    dt_bias: mx.array,
    dt_min: float,
    dt_max: float,
    prev_state=None,
) -> Tuple[mx.array, mx.array]:
    b, l, h, dh = x.shape
    _, _, g, _ = B.shape

    # Process dt
    if dt_bias is not None:
        dt = dt + dt_bias.reshape(1, 1, -1)
    dt = nn.softplus(dt)
    dt = mx.clip(dt, a_min=dt_min, a_max=dt_max)

    # Reshape tensors
    B_reshaped = mx.swapaxes(mx.swapaxes(B, 1, 3), 1, 2)
    C_reshaped = mx.swapaxes(C, 1, 2)

    # Compute CB
    CB = C_reshaped @ B_reshaped
    CB = mx.repeat(CB, repeats=h // g, axis=1)

    # Compute decay terms
    dtA = dt * A.reshape(1, 1, -1)
    dtA = mx.swapaxes(dtA, 1, 2)
    decay = mx.exp(segsum(dtA))

    # Create attention matrix
    surrogate_attention_matrix = mx.tril(CB * decay, 0)

    # Apply attention
    dtx = dt.reshape(b, l, h, 1) * x
    y = surrogate_attention_matrix @ dtx.swapaxes(1, 2)
    y = mx.swapaxes(y, 1, 2)

    # Compute next state
    decay_last = decay[:, :, -1, :].reshape(b, h, l).swapaxes(1, 2).reshape(b, l, h, 1)
    B_for_state = mx.repeat(B_reshaped, h // g, axis=1).swapaxes(2, 3)
    dtxdecay = dtx * decay_last
    dtxdecay = dtxdecay.swapaxes(1, 2).swapaxes(2, 3)
    
    # Calculate new state contribution
    new_state_contribution = dtxdecay @ B_for_state
    
    # Initialize or update state
    if prev_state is not None:
        decayed_prev_state = prev_state * decay[:, :, -1, :].reshape(b, h, 1, 1)
        next_state = decayed_prev_state + new_state_contribution
    else:
        next_state = new_state_contribution

    # Add skip connection if D is provided
    if D is not None:
        y += x * D.reshape(1, 1, h, 1)

    # Reshape output
    y = y.reshape(b, l, h * dh)

    return y, next_state


def segsum(x):
    # x shape: [b, h, l]
    b, h, l = x.shape
    indices = mx.arange(l)
    mask = indices[:, None] >= indices[None, :]  # [l, l] lower triangular mask
    # Expand x for broadcasting
    x_expanded = x.reshape(b, h, l, 1)  # [b, h, l, 1]
    # Apply mask and sum
    masked_x = x_expanded * mask.reshape(1, 1, l, l)  # [b, h, l, l]
    x_segsum = mx.sum(masked_x, axis=2, keepdims=True)  # [b, h, 1, l]
    return x_segsum


class Mamba2Block(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()
        self.args = args
        self.d_model = args.hidden_size
        self.d_state = args.state_size
        self.d_conv = args.conv_kernel
        self.expand = args.expand
        self.d_inner = int(self.expand * self.d_model)
        self.n_groups = args.n_groups
        self.n_heads = args.num_heads
        self.d_head = self.d_inner // self.n_heads
        self.chunk_size = args.chunk_size
        
        d_in_proj = 2 * self.d_inner + 2 * self.n_groups * self.d_state + self.n_heads
        self.in_proj = nn.Linear(self.d_model, d_in_proj, bias=args.use_bias)

        self.dt_bias = mx.random.normal((self.n_heads,)) * args.initializer_range
        self.A_log = mx.random.normal((self.n_heads,)) * args.initializer_range
        self.D = mx.random.normal((self.n_heads,)) * args.initializer_range

        conv_channels = self.d_inner + 2 * self.n_groups * self.d_state
        self.conv1d = nn.Conv1d(
            in_channels=conv_channels,
            out_channels=conv_channels,
            kernel_size=self.d_conv,
            groups=conv_channels,
            padding=self.d_conv - 1,
        )
        
        self.norm = nn.RMSNorm(self.d_inner, eps=args.layer_norm_epsilon)
        self.out_proj = nn.Linear(self.d_inner, self.d_model, bias=args.use_bias)

    def __call__(self, u: mx.array, cache=None):
        batch_size, seq_len, _ = u.shape
        
        if cache is None:
            cache = [None, None]
        else:
            conv_state, ssm_state = cache

        zxBCdt = self.in_proj(u)

        z, xBC, dt = mx.split(
            zxBCdt,
            [self.d_inner, 2 * self.d_inner + 2 * self.n_groups * self.d_state], 
            axis=-1
        )

        # Handle convolution with caching
        xBC = mx.swapaxes(xBC, 1, 2)  # [B, L, C] -> [B, C, L]
        
        if conv_state is not None and seq_len > 0:
            # Concatenate cached state with current input
            xBC_with_cache = mx.concatenate([conv_state, xBC], axis=2)
        elif seq_len > 0:
            # For the first call, pad with zeros
            padding = mx.zeros((batch_size, xBC.shape[1], self.d_conv - 1))
            xBC_with_cache = mx.concatenate([padding, xBC], axis=2)
        else:
            xBC_with_cache = conv_state if conv_state is not None else mx.zeros((batch_size, xBC.shape[1], 0))
        
        # Save state for next iteration
        if seq_len > 0:
            next_conv_state = xBC_with_cache[:, :, -(self.d_conv - 1):]
        else:
            next_conv_state = conv_state
        
        # Apply regular convolution using nn.Conv1d
        if seq_len > 0:
            # Use the standard Conv1d module for the actual computation
            xBC_conv = self.conv1d(xBC_with_cache)
            xBC = xBC_conv[:, :, -seq_len:]  # Take only the relevant output positions
            xBC = mx.swapaxes(xBC, 1, 2)  # [B, C, L] -> [B, L, C]
            xBC = xBC * mx.sigmoid(xBC)
        else:
            # Handle empty sequence case
            xBC = mx.swapaxes(xBC, 1, 2)  # [B, C, L] -> [B, L, C]
        
        x, B, C = mx.split(
            xBC, 
            [self.d_inner, self.d_inner + self.d_state * self.n_groups], 
            axis=-1
        )

        x = mx.reshape(x, (batch_size, seq_len, self.n_heads, self.d_head))
        B = mx.reshape(B, (batch_size, seq_len, self.n_groups, -1))
        C = mx.reshape(C, (batch_size, seq_len, self.n_groups, -1))

        y, next_ssm_state = ssd_forward_attn(
            x=x,
            dt=dt,
            A=-mx.exp(self.A_log),
            B=B,
            C=C,
            D=self.D,
            dt_bias=self.dt_bias,
            dt_min=self.args.time_step_min,
            dt_max=self.args.time_step_max,
            prev_state=ssm_state
        )

        if self.args.norm_before_gate:
            y = self.norm(y)
            y = y * nn.silu(z)
        else:
            y = y * nn.silu(z)
            y = self.norm(y)

        y = self.out_proj(y)

        cache[0] = next_conv_state
        cache[1] = next_ssm_state
        return y


class ResidualBlock(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()
        self.residual_in_fp32 = args.residual_in_fp32
        self.mixer = Mamba2Block(args)
        self.norm = nn.RMSNorm(args.hidden_size)

    def __call__(self, x: mx.array, cache):
        if self.residual_in_fp32:
            x = x.astype(mx.float32)
        normed = self.norm(x)
        output = self.mixer(normed, cache)
        return output + x


class Mamba2(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()
        self.args = args
        self.embeddings = nn.Embedding(args.vocab_size, args.hidden_size)
        self.layers = [ResidualBlock(args) for _ in range(args.num_hidden_layers)]
        self.norm_f = nn.RMSNorm(args.hidden_size, eps=args.layer_norm_epsilon)

    def __call__(self, x: mx.array, cache):
        x = self.embeddings(x)
        if cache is None:
            cache = [None] * len(self.layers)
        
        hidden = x
        for layer, c in zip(self.layers, cache):
            hidden = layer(hidden, c)
        return self.norm_f(hidden)


class Model(nn.Module):
    def __init__(self, args: ModelArgs):
        super().__init__()
        self.args = args
        self.model_type = args.model_type
        self.backbone = Mamba2(args)

        if not args.tie_word_embeddings:
            self.lm_head = nn.Linear(args.hidden_size, args.vocab_size, bias=False)

    def __call__(self, inputs: mx.array, cache=None):
        hidden = self.backbone(inputs, cache)
        
        if self.args.tie_word_embeddings:
            logits = self.backbone.embeddings.as_linear(hidden)
        else:
            logits = self.lm_head(hidden)
        
        return logits

    def make_cache(self):
        return [MambaCache() for _ in range(len(self.layers))]

    @property
    def layers(self):
        return self.backbone.layers