adapter_v2.py

# Copyright Lightning AI. Licensed under the Apache License 2.0, see LICENSE file.

"""Implementation of the paper:

LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model
https://arxiv.org/abs/2304.15010

Port for LitGPT
"""

from dataclasses import dataclass
from typing import Any, Dict, List, Optional, Tuple, Type, Union

import torch
import torch.nn as nn
from typing_extensions import Self

import litgpt
from litgpt.adapter import GPT as BaseModel
from litgpt.adapter import Block as BaseBlock
from litgpt.adapter import CausalSelfAttention as BaseCausalSelfAttention
from litgpt.adapter import Config as BaseConfig
from litgpt.model import KVCache
from litgpt.utils import map_old_state_dict_weights
from litgpt.model import KVCache, apply_rope
from litgpt.smoe import AdapterV2SMoE

from transformers import PreTrainedModel

@dataclass
class Config(BaseConfig):
    @property
    def mlp_class(self) -> Type:
        return getattr(litgpt.adapter_v2, self.mlp_class_name)

@dataclass
class ConfigSMOE(BaseConfig):
    use_smoe: bool=False
    num_experts: int=4
    top_k: int=1
    alpha: int=0
    model_type: str = "gpt"
    def __init__(self, *args, **kwargs):
        super(ConfigSMOE, self).__init__(*args, **kwargs)
    
    @property
    def mlp_class(self) -> Type:
        return getattr(litgpt.adapter_v2, self.mlp_class_name)
    def load_extra(self, extra_config):
        for k in list(extra_config.keys()):
            setattr(self, k, extra_config[k])

def adapter_filter(key: str, value: Any) -> bool:
    
    adapter_substrings = (
        # regular adapter v1 parameters
        "adapter_wte",
        "gating_factor",
        # adapter v2: new bias and scale used in Linear
        "adapter_scale",
        "adapter_bias",
        # adapter v2: Norm parameters are now trainable
        "norm_1",
        "norm_2",
        "ln_f",
        # smoe: gating mechanism
        "gate",
        )
    return any(s in key for s in adapter_substrings)


class AdapterV2Linear(torch.nn.Module):
    def __init__(self, in_features: int, out_features: int, **kwargs) -> None:
        super().__init__()
        self.linear = torch.nn.Linear(in_features, out_features, **kwargs)
        self.adapter_bias = torch.nn.Parameter(torch.zeros(out_features), requires_grad=False)
        self.adapter_scale = torch.nn.Parameter(torch.ones(out_features), requires_grad=False)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        # breakpoint()
        return self.adapter_scale * (self.linear(x) + self.adapter_bias)

    def reset_parameters(self) -> None:
        nn.init.zeros_(self.adapter_bias)
        nn.init.ones_(self.adapter_scale)

class GPT(BaseModel, PreTrainedModel):
    config_class=ConfigSMOE

    def __init__(self, config: ConfigSMOE) -> None:
        # Skip the parent class __init__ altogether and replace it to avoid useless allocations
        nn.Module.__init__(self)
        # super().__init__(config)
        assert config.padded_vocab_size is not None
        self.config = config
        if config.use_smoe:
            print("🐙 Run AdapterV2SMoE")
            self.lm_head = AdapterV2SMoE(
                in_features=config.n_embd,
                out_features=config.padded_vocab_size,
                num_experts=config.num_experts,
                top_k=config.top_k,
                bias=config.lm_head_bias
            )
            self.transformer = nn.ModuleDict(
                dict(
                    wte=nn.Embedding(config.padded_vocab_size, config.n_embd),
                    h=nn.ModuleList(BlockSMoE(config, i) for i in range(config.n_layer)),
                    ln_f=config.norm_class(config.n_embd, eps=config.norm_eps),
                )
            )
        else:
            print("🐙 Run AdapterV2Linear")
            self.lm_head = AdapterV2Linear(config.n_embd, config.padded_vocab_size, bias=config.lm_head_bias)
            self.transformer = nn.ModuleDict(
                dict(
                    wte=nn.Embedding(config.padded_vocab_size, config.n_embd),
                    h=nn.ModuleList(Block(config, i) for i in range(config.n_layer)),
                    ln_f=config.norm_class(config.n_embd, eps=config.norm_eps),
                )
            )
        self.max_seq_length = self.config.block_size
        self.mask_cache: Optional[torch.Tensor] = None

    def forward(
        self, idx: torch.Tensor, input_pos: Optional[torch.Tensor] = None, lm_head_chunk_size: int = 0
    ) -> Union[torch.Tensor, List[torch.Tensor]]:
        T = idx.size(1)
        if self.max_seq_length < T:
            raise ValueError(f"Cannot forward sequence of length {T}, max seq length is only {self.max_seq_length}.")

        if input_pos is not None:  # use the kv cache
            cos = self.cos.index_select(0, input_pos)
            sin = self.sin.index_select(0, input_pos)
            if self.mask_cache is None:
                raise TypeError("You need to call `gpt.set_kv_cache()`")
            mask = self.mask_cache.index_select(2, input_pos)
        else:
            cos = self.cos[:T]
            sin = self.sin[:T]
            mask = None

        x = self.transformer.wte(idx)  # token embeddings of shape (b, t, n_embd)
        if self.config.scale_embeddings:
            x = x * (self.config.n_embd**0.5)
        for block in self.transformer.h:
            x = block(x, cos, sin, mask, input_pos)
        x = self.transformer.ln_f(x)
        if self.config.use_smoe:
            if lm_head_chunk_size > 0:
                outputs = []
                routers = []
                for x_i in x.split(lm_head_chunk_size, dim = 1):
                    output, router = self.lm_head(x_i)
                    outputs.append(output)
                    routers.append(router)
                return outputs, routers
            output, router = self.lm_head(x)
            return output, router #(b, t, vocab_size)
        else:
            if lm_head_chunk_size > 0:
                # chunk the lm head logits to reduce the peak memory used by autograd
                return [self.lm_head(x_i) for x_i in x.split(lm_head_chunk_size, dim=1)]
            return self.lm_head(x)  # (b, t, vocab_size)

    @classmethod
    def from_name(cls, name: str, **kwargs: Any) -> Self:
        return cls(Config.from_name(name, **kwargs))

    def _init_weights(self, module: nn.Module) -> None:
        """Meant to be used with `gpt.apply(gpt._init_weights)`. Unused method left for completeness."""
        super()._init_weights(module)
        if isinstance(module, AdapterV2Linear):
            module.reset_parameters()

    def _load_from_state_dict(self, state_dict: Dict, prefix: str, *args: Any, **kwargs: Any) -> None:
        """For compatibility with base checkpoints."""
        mapping = {"lm_head.weight": "lm_head.linear.weight", "lm_head.bias": "lm_head.linear.bias"}
        state_dict = map_old_state_dict_weights(state_dict, mapping, prefix)
        super()._load_from_state_dict(state_dict, prefix, *args, **kwargs)


class Block(BaseBlock):
    """The implementation is identical to `litgpt.model.Block` with the exception that
    we replace the attention layer where adaption is implemented."""

    def __init__(self, config: Config, block_idx: int) -> None:
        # Skip the parent class __init__ altogether and replace it to avoid useless allocations
        nn.Module.__init__(self)
        self.norm_1 = config.norm_class(config.n_embd, eps=config.norm_eps)
        if config.use_smoe:
            self.attn = CausalSelfAttentionSMoE(config, block_idx)
        else:
            self.attn = CausalSelfAttention(config, block_idx)
        if not config.shared_attention_norm:
            self.norm_2 = config.norm_class(config.n_embd, eps=config.norm_eps)
        self.mlp = config.mlp_class(config)

        self.config = config

class BlockSMoE(Block):
    def forward(
        self,
        x: torch.Tensor,
        cos: torch.Tensor,
        sin: torch.Tensor,
        mask: Optional[torch.Tensor] = None,
        input_pos: Optional[torch.Tensor] = None,
    ) -> torch.Tensor:
        x_normed = self.norm_1(x)
        attention_output, _ = self.attn(x_normed, cos, sin, mask, input_pos)
        if self.config.parallel_residual:
            x_normed = x_normed if self.config.shared_attention_norm else self.norm_2(x)
            x = self.mlp(x_normed) + attention_output + x
        else:
            x = attention_output + x
            x = self.mlp(self.norm_2(x)) + x
        return x


class CausalSelfAttention(BaseCausalSelfAttention):
    """A modification of `litgpt.adapter.CausalSelfAttention` that uses the Adapter V2 Linear class"""

    def __init__(self, config: Config, block_idx: int) -> None:
        # Skip the parent class __init__ altogether and replace it to avoid useless allocations
        nn.Module.__init__(self)
        shape = (config.n_head + 2 * config.n_query_groups) * config.head_size
        # key, query, value projections for all heads, but in a batch
        if config.use_smoe:
            self.attn = AdapterV2SMoE(
                in_features=config.n_embd,
                out_features=shape,
                num_experts=config.num_experts,
                top_k=config.top_k,
                bias=config.bias
            )
            # output projection
            # if `head_size` is explicitly specified in the config, `n_emd` might not be equal to `head_size * n_head`
            self.proj = AdapterV2SMoE(
                in_features=config.head_size * config.n_head,
                out_features=config.n_embd,
                num_experts=config.num_experts,
                top_k=config.top_k,
                bias=config.bias
            )
            # disabled by default
        else:
            self.attn = AdapterV2Linear(in_features=config.n_embd, out_features=shape, bias=config.bias)
            # output projection
            # if `head_size` is explicitly specified in the config, `n_emd` might not be equal to `head_size * n_head`
            self.proj = AdapterV2Linear(config.head_size * config.n_head, config.n_embd, bias=config.bias)
            # disabled by default
        self.kv_cache: Optional[KVCache] = None

        if block_idx >= config.adapter_start_layer:
            # adapter embedding layer
            self.adapter_wte = nn.Embedding(config.adapter_prompt_length, config.n_embd)
            # gate for adaption
            self.gating_factor = torch.nn.Parameter(torch.zeros(1, 1, config.n_head, 1))
            # kv cache for inference
            self.adapter_kv_cache: Optional[Tuple[torch.Tensor, torch.Tensor]] = None
        self.block_idx = block_idx

        self.config = config

    def _load_from_state_dict(self, state_dict: Dict, prefix: str, *args: Any, **kwargs: Any) -> None:
        """For compatibility with base checkpoints."""
        mapping = {
            "attn.weight": "attn.linear.weight",
            "attn.bias": "attn.linear.bias",
            "proj.weight": "proj.linear.weight",
            "proj.bias": "proj.linear.bias",
        }
        state_dict = map_old_state_dict_weights(state_dict, mapping, prefix)
        # For compatibility with older checkpoints
        if (key := prefix + "gating_factor") in state_dict and state_dict[key].size(1) == self.config.n_head:
            state_dict[key] = state_dict[key].permute(0, 2, 1, 3)
        super()._load_from_state_dict(state_dict, prefix, *args, **kwargs)

class CausalSelfAttentionSMoE(CausalSelfAttention):
    def forward(
        self,
        x: torch.Tensor,
        cos: torch.Tensor,
        sin: torch.Tensor,
        mask: Optional[torch.Tensor] = None,
        input_pos: Optional[torch.Tensor] = None,
    ) -> torch.Tensor:
        B, T, C = x.size()  # batch size, sequence length, embedding dimensionality (n_embd)
        
        # breakpoint()
        qkv, _ = self.attn(x)

        # assemble into a number of query groups to support MHA, MQA and GQA together (see `config.n_query_groups`)
        q_per_kv = self.config.n_head // self.config.n_query_groups
        total_qkv = q_per_kv + 2  # each group has 1+ queries, 1 key, and 1 value
        qkv = qkv.view(B, T, self.config.n_query_groups, total_qkv, self.config.head_size)
        qkv = qkv.permute(0, 2, 3, 1, 4)  # (B, n_query_groups, total_qkv, T, hs)

        # split batched computation into three
        q, k, v = qkv.split((q_per_kv, 1, 1), dim=2)

        # maybe repeat k and v if for the non multi-head attention cases
        # training: flash attention requires it
        # inference: multi-query would require a full kv cache so avoid it to limit its memory usage
        if self.config.n_query_groups != self.config.n_head and (input_pos is None or self.config.n_query_groups != 1):
            k = k.expand(B, self.config.n_query_groups, q_per_kv, T, self.config.head_size)
            v = v.expand(B, self.config.n_query_groups, q_per_kv, T, self.config.head_size)

        q = q.reshape(B, -1, T, self.config.head_size)  # (B, nh_q, T, hs)
        k = k.reshape(B, -1, T, self.config.head_size)  # (B, nh_k, T, hs)
        v = v.reshape(B, -1, T, self.config.head_size)  # (B, nh_v, T, hs)

        q_roped = apply_rope(q[..., : self.config.rope_n_elem], cos, sin)
        k_roped = apply_rope(k[..., : self.config.rope_n_elem], cos, sin)
        q = torch.cat((q_roped, q[..., self.config.rope_n_elem :]), dim=-1)
        k = torch.cat((k_roped, k[..., self.config.rope_n_elem :]), dim=-1)

        if input_pos is not None:
            if not isinstance(self.kv_cache, KVCache):
                raise TypeError("You need to call `gpt.set_kv_cache()`")
            k, v = self.kv_cache(input_pos, k, v)

        y = self.scaled_dot_product_attention(q, k, v, mask)

        y = y.reshape(B, T, self.config.head_size * self.config.n_head)  # re-assemble all head outputs side by side

        # output projection
        return self.proj(y)

class GptNeoxMLP(litgpt.model.GptNeoxMLP):
    def __init__(self, config: Config) -> None:
        nn.Module.__init__(self)
        if config.use_smoe:
            self.fc = AdapterV2SMoE(
                in_features=config.n_embd,
                out_features=config.intermediate_size,
                num_experts=config.num_experts,
                top_k=config.top_k,
                bias=config.bias
            )
            # output projection
            # if `head_size` is explicitly specified in the config, `n_emd` might not be equal to `head_size * n_head`
            self.proj = AdapterV2SMoE(
                in_features=config.intermediate_size,
                out_features=config.n_embd,
                num_experts=config.num_experts,
                top_k=config.top_k,
                bias=config.bias
            )
        else:
            self.fc = AdapterV2Linear(config.n_embd, config.intermediate_size, bias=config.bias)
            self.proj = AdapterV2Linear(config.intermediate_size, config.n_embd, bias=config.bias)

        self.config = config

    def _load_from_state_dict(self, state_dict: Dict, prefix: str, *args: Any, **kwargs: Any) -> None:
        """For compatibility with base checkpoints."""
        mapping = {
            "fc.weight": "fc.linear.weight",
            "fc.bias": "fc.linear.bias",
            "proj.weight": "proj.linear.weight",
            "proj.bias": "proj.linear.bias",
        }
        state_dict = map_old_state_dict_weights(state_dict, mapping, prefix)
        super()._load_from_state_dict(state_dict, prefix, *args, **kwargs)


class LLaMAMLP(litgpt.model.LLaMAMLP):
    def __init__(self, config: Config) -> None:
        nn.Module.__init__(self)
        self.fc_1 = AdapterV2Linear(config.n_embd, config.intermediate_size, bias=config.bias)
        self.fc_2 = AdapterV2Linear(config.n_embd, config.intermediate_size, bias=config.bias)
        self.proj = AdapterV2Linear(config.intermediate_size, config.n_embd, bias=config.bias)

        self.config = config

    def _load_from_state_dict(self, state_dict: Dict, prefix: str, *args: Any, **kwargs: Any) -> None:
        """For compatibility with base checkpoints."""
        mapping = {
            "fc_1.weight": "fc_1.linear.weight",
            "fc_1.bias": "fc_1.linear.bias",
            "fc_2.weight": "fc_2.linear.weight",
            "fc_2.bias": "fc_2.linear.bias",
            "proj.weight": "proj.linear.weight",
            "proj.bias": "proj.linear.bias",
        }
        state_dict = map_old_state_dict_weights(state_dict, mapping, prefix)
        super()._load_from_state_dict(state_dict, prefix, *args, **kwargs)


class GemmaMLP(LLaMAMLP):
    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x_fc_1 = self.fc_1(x)
        x_fc_2 = self.fc_2(x)
        x = torch.nn.functional.gelu(x_fc_1, approximate=self.config.gelu_approximate) * x_fc_2
        return self.proj(x)


class LLaMAMoE(litgpt.model.LLaMAMoE):
    def __init__(self, config: Config) -> None:
        nn.Module.__init__(self)
        self.gate = AdapterV2Linear(config.n_embd, config.n_expert, bias=False)
        self.experts = nn.ModuleList(LLaMAMLP(config) for _ in range(config.n_expert))

        self.config = config

    def _load_from_state_dict(self, state_dict: Dict, prefix: str, *args: Any, **kwargs: Any) -> None:
        """For compatibility with base checkpoints."""
        mapping = {"gate.weight": "gate.linear.weight"}
        state_dict = map_old_state_dict_weights(state_dict, mapping, prefix)
        super()._load_from_state_dict(state_dict, prefix, *args, **kwargs)


def mark_only_adapter_v2_as_trainable(model: GPT) -> None:
    """Sets requires_grad=False for all non-adapter weights"""
    for name, param in model.named_parameters():
        param.requires_grad = adapter_filter(name, param)