# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved. # # This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX # and OPT implementations in this library. It has been modified from its # original forms to accommodate minor architectural differences compared # to GPT-NeoX and OPT used by the Meta AI team that trained the model. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """ PyTorch LLaMA model.""" import math from typing import List, Optional, Tuple, Union import faiss from einops import rearrange import torch import torch.utils.checkpoint from torch import nn import torch.nn.functional as F from torch.linalg import vector_norm from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss from transformers.activations import ACT2FN from transformers.modeling_outputs import ( BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast, ) from transformers.modeling_utils import PreTrainedModel from transformers.utils import ( add_start_docstrings, add_start_docstrings_to_model_forward, logging, replace_return_docstrings, ) from .configuration_llama import ExtendedLlamaConfig logger = logging.get_logger(__name__) _CONFIG_FOR_DOC = "ExtendedLlamaConfig" # Copied from transformers.models.bart.modeling_bart._make_causal_mask def _make_causal_mask( input_ids_shape: torch.Size, dtype: torch.dtype, device: torch.device, past_key_values_length: int = 0 ): """ Make causal mask used for bi-directional self-attention. """ bsz, tgt_len = input_ids_shape mask = torch.full((tgt_len, tgt_len), torch.tensor(torch.finfo(dtype).min, device=device), device=device) mask_cond = torch.arange(mask.size(-1), device=device) mask.masked_fill_(mask_cond < (mask_cond + 1).view(mask.size(-1), 1), 0) mask = mask.to(dtype) if past_key_values_length > 0: mask = torch.cat([torch.zeros(tgt_len, past_key_values_length, dtype=dtype, device=device), mask], dim=-1) return mask[None, None, :, :].expand(bsz, 1, tgt_len, tgt_len + past_key_values_length) # Copied from transformers.models.bart.modeling_bart._expand_mask def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None): """ Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`. """ bsz, src_len = mask.size() tgt_len = tgt_len if tgt_len is not None else src_len expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype) inverted_mask = 1.0 - expanded_mask return inverted_mask.masked_fill(inverted_mask.to(torch.bool), torch.finfo(dtype).min) class LlamaRMSNorm(nn.Module): def __init__(self, hidden_size, eps=1e-6): """ LlamaRMSNorm is equivalent to T5LayerNorm """ super().__init__() self.weight = nn.Parameter(torch.ones(hidden_size)) self.variance_epsilon = eps def forward(self, hidden_states): input_dtype = hidden_states.dtype variance = hidden_states.to(torch.float32).pow(2).mean(-1, keepdim=True) hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon) return (self.weight * hidden_states).to(input_dtype) class LlamaRotaryEmbedding(torch.nn.Module): def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None): super().__init__() inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(device) / dim)) self.register_buffer("inv_freq", inv_freq) # Build here to make `torch.jit.trace` work. self.max_seq_len_cached = max_position_embeddings t = torch.arange( self.max_seq_len_cached, device=self.inv_freq.device, dtype=self.inv_freq.dtype, ) freqs = torch.einsum("i,j->ij", t, self.inv_freq) # Different from paper, but it uses a different permutation in order to obtain the same calculation emb = torch.cat((freqs, freqs), dim=-1) dtype = torch.get_default_dtype() self.register_buffer( "cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False ) self.register_buffer( "sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False ) def forward(self, x, seq_len=None): # x: [bs, num_attention_heads, seq_len, head_size] # This `if` block is unlikely to be run after we build sin/cos in `__init__`. Keep the logic here just in case. if seq_len > self.max_seq_len_cached: self.max_seq_len_cached = seq_len t = torch.arange( self.max_seq_len_cached, device=x.device, dtype=self.inv_freq.dtype ) freqs = torch.einsum("i,j->ij", t, self.inv_freq) # Different from paper, but it uses a different permutation in order to obtain the same calculation emb = torch.cat((freqs, freqs), dim=-1).to(x.device) self.register_buffer( "cos_cached", emb.cos()[None, None, :, :].to(x.dtype), persistent=False ) self.register_buffer( "sin_cached", emb.sin()[None, None, :, :].to(x.dtype), persistent=False ) return ( self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype), self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype), ) class LlamaDynamicScaledRotaryEmbedding(torch.nn.Module): def __init__(self, dim, max_position_embeddings=2048, base=10000, ntk=False, device=None): super().__init__() self.ntk = ntk self.base = base self.dim = dim self.max_position_embeddings = max_position_embeddings inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(device) / dim)) self.register_buffer("inv_freq", inv_freq) # Build here to make `torch.jit.trace` work. self.max_seq_len_cached = max_position_embeddings t = torch.arange(self.max_seq_len_cached, device=self.inv_freq.device, dtype=self.inv_freq.dtype) freqs = torch.einsum("i,j->ij", t, self.inv_freq) # Different from paper, but it uses a different permutation in order to obtain the same calculation emb = torch.cat((freqs, freqs), dim=-1) dtype = torch.get_default_dtype() self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False) self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False) def forward(self, x, seq_len=None): # x: [bs, num_attention_heads, seq_len, head_size] # This `if` block is unlikely to be run after we build sin/cos in `__init__`. Keep the logic here just in case. if seq_len > self.max_seq_len_cached: self.max_seq_len_cached = seq_len if self.ntk: base = self.base * ((self.ntk * seq_len / self.max_position_embeddings) - (self.ntk - 1)) ** (self.dim / (self.dim-2)) inv_freq = 1.0 / (base ** (torch.arange(0, self.dim, 2).float().to(x.device) / self.dim)) self.register_buffer("inv_freq", inv_freq) t = torch.arange(self.max_seq_len_cached, device=x.device, dtype=self.inv_freq.dtype) if not self.ntk: t *= self.max_position_embeddings / seq_len freqs = torch.einsum("i,j->ij", t, self.inv_freq) # Different from paper, but it uses a different permutation in order to obtain the same calculation emb = torch.cat((freqs, freqs), dim=-1).to(x.device) self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(x.dtype), persistent=False) self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(x.dtype), persistent=False) return ( self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype), self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype), ) class LlamaLinearScaledRotaryEmbedding(torch.nn.Module): def __init__(self, dim, max_position_embeddings=2048, base=10000, scale=1, device=None): super().__init__() self.scale = scale inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(device) / dim)) self.register_buffer("inv_freq", inv_freq) # Build here to make `torch.jit.trace` work. self.max_seq_len_cached = max_position_embeddings t = torch.arange(self.max_seq_len_cached, device=self.inv_freq.device, dtype=self.inv_freq.dtype) t /= self.scale freqs = torch.einsum("i,j->ij", t, self.inv_freq) # Different from paper, but it uses a different permutation in order to obtain the same calculation emb = torch.cat((freqs, freqs), dim=-1) dtype = torch.get_default_dtype() self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False) self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False) def forward(self, x, seq_len=None): # x: [bs, num_attention_heads, seq_len, head_size] # This `if` block is unlikely to be run after we build sin/cos in `__init__`. Keep the logic here just in case. if seq_len > self.max_seq_len_cached: self.max_seq_len_cached = seq_len t = torch.arange(self.max_seq_len_cached, device=x.device, dtype=self.inv_freq.dtype) t /= self.scale freqs = torch.einsum("i,j->ij", t, self.inv_freq) # Different from paper, but it uses a different permutation in order to obtain the same calculation emb = torch.cat((freqs, freqs), dim=-1).to(x.device) self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(x.dtype), persistent=False) self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(x.dtype), persistent=False) return ( self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype), self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype), ) class LlamaNTKScaledRotaryEmbedding(torch.nn.Module): def __init__(self, dim, max_position_embeddings=2048, base=10000, alpha=1, device=None): super().__init__() base = base * alpha ** (dim / (dim-2)) inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(device) / dim)) self.register_buffer("inv_freq", inv_freq) # Build here to make `torch.jit.trace` work. self.max_seq_len_cached = max_position_embeddings t = torch.arange(self.max_seq_len_cached, device=self.inv_freq.device, dtype=self.inv_freq.dtype) freqs = torch.einsum("i,j->ij", t, self.inv_freq) # Different from paper, but it uses a different permutation in order to obtain the same calculation emb = torch.cat((freqs, freqs), dim=-1) dtype = torch.get_default_dtype() self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(dtype), persistent=False) self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(dtype), persistent=False) def forward(self, x, seq_len=None): # x: [bs, num_attention_heads, seq_len, head_size] # This `if` block is unlikely to be run after we build sin/cos in `__init__`. Keep the logic here just in case. if seq_len > self.max_seq_len_cached: self.max_seq_len_cached = seq_len t = torch.arange(self.max_seq_len_cached, device=x.device, dtype=self.inv_freq.dtype) freqs = torch.einsum("i,j->ij", t, self.inv_freq) # Different from paper, but it uses a different permutation in order to obtain the same calculation emb = torch.cat((freqs, freqs), dim=-1).to(x.device) self.register_buffer("cos_cached", emb.cos()[None, None, :, :].to(x.dtype), persistent=False) self.register_buffer("sin_cached", emb.sin()[None, None, :, :].to(x.dtype), persistent=False) return ( self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype), self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype), ) def rotate_half(x): """Rotates half the hidden dims of the input.""" x1 = x[..., : x.shape[-1] // 2] x2 = x[..., x.shape[-1] // 2 :] return torch.cat((-x2, x1), dim=-1) def apply_rotary_pos_emb(q, k, cos, sin, position_ids): # The first two dimensions of cos and sin are always 1, so we can `squeeze` them. cos = cos.squeeze(1).squeeze(0) # [seq_len, dim] sin = sin.squeeze(1).squeeze(0) # [seq_len, dim] s_q = q.size(-2) #Since we apply rotary pos emb after reading from cache, queries may be shorter _q_position_ids = position_ids[:, -s_q:] _q_cos = cos[_q_position_ids].unsqueeze(1) _q_sin = sin[_q_position_ids].unsqueeze(1) q_embed = (q * _q_cos) + (rotate_half(q) * _q_sin) cos = cos[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim] sin = sin[position_ids].unsqueeze(1) # [bs, 1, seq_len, dim] k_embed = (k * cos) + (rotate_half(k) * sin) return q_embed, k_embed class LlamaMLP(nn.Module): def __init__( self, hidden_size: int, intermediate_size: int, hidden_act: str, ): super().__init__() self.gate_proj = nn.Linear(hidden_size, intermediate_size, bias=False) self.down_proj = nn.Linear(intermediate_size, hidden_size, bias=False) self.up_proj = nn.Linear(hidden_size, intermediate_size, bias=False) self.act_fn = ACT2FN[hidden_act] def forward(self, x): return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x)) class ExtendedLlamaAttention(nn.Module): """Multi-headed attention from 'Attention Is All You Need' paper""" def __init__(self, config: ExtendedLlamaConfig): super().__init__() self.config = config self.hidden_size = config.hidden_size self.num_heads = config.num_attention_heads self.head_dim = self.hidden_size // self.num_heads self.max_position_embeddings = config.max_position_embeddings self.num_hidden_layers = config.num_hidden_layers if (self.head_dim * self.num_heads) != self.hidden_size: raise ValueError( f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}" f" and `num_heads`: {self.num_heads})." ) self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False) self.k_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False) self.v_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False) self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False) self.rotary_emb = LlamaRotaryEmbedding(self.head_dim, max_position_embeddings=self.max_position_embeddings) def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int): return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous() def forward( self, hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor] = None, position_ids: Optional[torch.LongTensor] = None, past_key_value: Optional[Tuple[torch.Tensor]] = None, output_attentions: bool = False, use_cache: bool = False, long_range_past_key_value=None, faiss_indexes=None, mask_by_sim=False, sim_threshold=0.0, topk=None, current_layer=None, ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: bsz, q_len, _ = hidden_states.size() query_states = self.q_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) key_states = self.k_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) value_states = self.v_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) if past_key_value is not None: # reuse k, v, self_attention key_states = torch.cat([past_key_value[0], key_states], dim=2) value_states = torch.cat([past_key_value[1], value_states], dim=2) past_key_value = (key_states, value_states) if use_cache else None kv_seq_len = key_states.shape[-2] cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len) query_states_no_rotary = query_states.clone() # use queries wo positional info for memory retrieval query_states, key_states = apply_rotary_pos_emb( query_states, key_states, cos, sin, position_ids ) # [bsz, nh, t, hd] bsz, nh, s_q, hd = query_states.shape s_k = key_states.size(-2) attn_weights = torch.matmul( query_states, key_states.transpose(2, 3) ) / math.sqrt(self.head_dim) if attn_weights.size() != (bsz, self.num_heads, q_len, kv_seq_len): raise ValueError( f"Attention weights should be of size {(bsz, self.num_heads, q_len, kv_seq_len)}, but is" f" {attn_weights.size()}" ) if long_range_past_key_value is not None or faiss_indexes is not None: if long_range_past_key_value is not None: #manual memories k_cache, v_cache = long_range_past_key_value s_cache = k_cache.size(-2) k_cache = k_cache.to(key_states.device) v_cache = v_cache.to(key_states.device) q_n = query_states_no_rotary/vector_norm(query_states_no_rotary, ord=2, dim=-1, keepdim=True) k_n = k_cache/vector_norm(k_cache, ord=2, dim=-1, keepdim=True) sim = q_n.matmul(k_n.transpose(2,3)) if s_cache sim_threshold).bool(), 'b h s i -> b h (s i)').unsqueeze(-2).expand(-1, -1, s_q, -1) elif faiss_indexes is not None: #faiss indexes kn_index, kv_index = faiss_indexes q_n = query_states_no_rotary/vector_norm(query_states_no_rotary, ord=2, dim=-1, keepdim=True) one_hot_encodings = F.one_hot(torch.arange(0, nh*self.num_hidden_layers, device=query_states.device))*10 q_n = torch.concat([rearrange(q_n, 'b h s d -> b (h s) d', h=nh), one_hot_encodings[nh*current_layer:nh*(current_layer+1)].unsqueeze(0).repeat_interleave(repeats=query_states.size(-2), dim=-2)], dim=-1).squeeze() D, I = kn_index.search(q_n.to('cpu').numpy(), k=topk) selected_k=rearrange(torch.tensor(kv_index.reconstruct_batch(I.flatten()))[:,:hd], '(h s) d -> 1 h s d', h=nh).to(query_states.device) cos_m, sin_m = self.rotary_emb(value_states, seq_len=self.max_position_embeddings) # use max pos emb for memories cos_m = cos_m[:,:,-1,...].repeat(1,1,s_q * topk,1) sin_m = sin_m[:,:,-1,...].repeat(1,1,s_q * topk,1) _, selected_k = apply_rotary_pos_emb( torch.ones(selected_k.shape, device=key_states.device), selected_k, cos_m, sin_m, position_ids=torch.arange(s_q * topk, device=key_states.device).unsqueeze(0) ) # Apply rotary pos emb to selected memory keys, use dummy input for queries selected_v=rearrange(torch.tensor(kv_index.reconstruct_batch(I.flatten()))[:,hd:], '(h s) d -> 1 h s d', h=nh).to(query_states.device) attn_weight_cache = torch.matmul(query_states, selected_k.transpose(2, 3)) / math.sqrt(self.head_dim) if mask_by_sim: attn_weight_cache = attn_weight_cache.masked_fill(sim_mask, torch.finfo(selected_k.dtype).min) attn_weights = torch.cat([attn_weight_cache, attn_weights], dim=-1) value_states = torch.cat([selected_v, value_states], dim=-2) min_val = torch.finfo(attn_weights.dtype).min def _create_active_externalism_mask(k, s_q, device, min_val=min_val): mask = torch.ones(s_q, s_q * k, device=device, dtype=torch.float32) for i in range(s_q): mask[i, i * k : (i + 1) * k] = 0 filled = mask.masked_fill(mask.bool(), min_val) return filled if attention_mask is not None: if attention_mask.size() != (bsz, 1, q_len, kv_seq_len): raise ValueError( f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}" ) if long_range_past_key_value is not None: memory_mask = _create_active_externalism_mask(k=topk,s_q=s_q, device=attn_weights.device) attention_mask = torch.cat([memory_mask, attention_mask[:,:,:,-s_k:].squeeze(dim=[0,1])], dim=1).unsqueeze(dim=0).unsqueeze(dim=1) attn_weights = attn_weights + attention_mask attn_weights = torch.max( attn_weights, torch.tensor(torch.finfo(attn_weights.dtype).min, device=attn_weights.device) ) # upcast attention to fp32 attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype) attn_output = torch.matmul(attn_weights, value_states) if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim): raise ValueError( f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is" f" {attn_output.size()}" ) attn_output = attn_output.transpose(1, 2) attn_output = attn_output.reshape(bsz, q_len, self.hidden_size) attn_output = self.o_proj(attn_output) if not output_attentions: attn_weights = None if long_range_past_key_value is None and faiss_indexes is None: reshaped_idx=None return attn_output, attn_weights, past_key_value, reshaped_idx class ExtendedLlamaDecoderLayer(nn.Module): def __init__(self, config: ExtendedLlamaConfig): super().__init__() self.hidden_size = config.hidden_size self.self_attn = ExtendedLlamaAttention(config=config) self.mlp = LlamaMLP( hidden_size=self.hidden_size, intermediate_size=config.intermediate_size, hidden_act=config.hidden_act, ) self.input_layernorm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps) self.post_attention_layernorm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps) def forward( self, hidden_states: torch.Tensor, attention_mask: Optional[torch.Tensor] = None, position_ids: Optional[torch.LongTensor] = None, past_key_value: Optional[Tuple[torch.Tensor]] = None, output_attentions: Optional[bool] = False, use_cache: Optional[bool] = False, long_range_past_key_value:Optional[Tuple[torch.Tensor]] = None, faiss_indexes:Tuple=None, mask_by_sim:bool=False, sim_threshold:float=None, topk:int=None, current_layer=None ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]: """ Args: hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` attention_mask (`torch.FloatTensor`, *optional*): attention mask of size `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values. output_attentions (`bool`, *optional*): Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned tensors for more detail. use_cache (`bool`, *optional*): If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see `past_key_values`). past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states """ residual = hidden_states hidden_states = self.input_layernorm(hidden_states) # Self Attention hidden_states, self_attn_weights, present_key_value, selected_idx = self.self_attn( hidden_states=hidden_states, attention_mask=attention_mask, position_ids=position_ids, past_key_value=past_key_value, output_attentions=output_attentions, use_cache=use_cache, long_range_past_key_value=long_range_past_key_value, faiss_indexes=faiss_indexes, mask_by_sim=mask_by_sim, sim_threshold=sim_threshold, topk=topk, current_layer=current_layer, ) hidden_states = residual + hidden_states # Fully Connected residual = hidden_states hidden_states = self.post_attention_layernorm(hidden_states) hidden_states = self.mlp(hidden_states) hidden_states = residual + hidden_states outputs = (hidden_states,) if output_attentions: outputs += (self_attn_weights,) if use_cache: outputs += (present_key_value,) if output_attentions: outputs += (selected_idx,) return outputs LLAMA_START_DOCSTRING = r""" This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.) This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior. Parameters: config ([`LlamaConfig`]): Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the [`~PreTrainedModel.from_pretrained`] method to load the model weights. """ @add_start_docstrings( "The bare LLaMA Model outputting raw hidden-states without any specific head on top.", LLAMA_START_DOCSTRING, ) class LlamaPreTrainedModel(PreTrainedModel): config_class = ExtendedLlamaConfig base_model_prefix = "model" supports_gradient_checkpointing = True _no_split_modules = ["LlamaDecoderLayer"] _skip_keys_device_placement = "past_key_values" def _init_weights(self, module): std = self.config.initializer_range if isinstance(module, nn.Linear): module.weight.data.normal_(mean=0.0, std=std) if module.bias is not None: module.bias.data.zero_() elif isinstance(module, nn.Embedding): module.weight.data.normal_(mean=0.0, std=std) if module.padding_idx is not None: module.weight.data[module.padding_idx].zero_() def _set_gradient_checkpointing(self, module, value=False): if isinstance(module, ExtendedLlamaModel): module.gradient_checkpointing = value LLAMA_INPUTS_DOCSTRING = r""" Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide it. Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. [What are input IDs?](../glossary#input-ids) attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*): Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. [What are attention masks?](../glossary#attention-mask) Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and [`PreTrainedTokenizer.__call__`] for details. If `past_key_values` is used, optionally only the last `decoder_input_ids` have to be input (see `past_key_values`). If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`] and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information on the default strategy. - 1 indicates the head is **not masked**, - 0 indicates the head is **masked**. position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0, config.n_positions - 1]`. [What are position IDs?](../glossary#position-ids) past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`): Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`. Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `decoder_input_ids` of shape `(batch_size, sequence_length)`. inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*): Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert `input_ids` indices into associated vectors than the model's internal embedding lookup matrix. use_cache (`bool`, *optional*): If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see `past_key_values`). output_attentions (`bool`, *optional*): Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned tensors for more detail. output_hidden_states (`bool`, *optional*): Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for more detail. return_dict (`bool`, *optional*): Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. """ @add_start_docstrings( "The bare LLaMA Model outputting raw hidden-states without any specific head on top.", LLAMA_START_DOCSTRING, ) class ExtendedLlamaModel(LlamaPreTrainedModel): """ Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`LlamaDecoderLayer`] Args: config: LlamaConfig """ def __init__(self, config: ExtendedLlamaConfig): super().__init__(config) self.padding_idx = config.pad_token_id self.vocab_size = config.vocab_size self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx) self.layers = nn.ModuleList([ExtendedLlamaDecoderLayer(config) for _ in range(config.num_hidden_layers)]) self.norm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps) self.gradient_checkpointing = False # Initialize weights and apply final processing self.mask_by_sim = config.mask_by_sim self.sim_threshold = config.sim_threshold self.topk = config.topk self.use_active_externalism = config.use_active_externalism self.use_active_externalism_by_layer = config.use_active_externalism_by_layer self.post_init() def get_input_embeddings(self): return self.embed_tokens def set_input_embeddings(self, value): self.embed_tokens = value # Copied from transformers.models.bart.modeling_bart.BartDecoder._prepare_decoder_attention_mask def _prepare_decoder_attention_mask(self, attention_mask, input_shape, inputs_embeds, past_key_values_length): # create causal mask # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len] combined_attention_mask = None if input_shape[-1] > 1: combined_attention_mask = _make_causal_mask( input_shape, inputs_embeds.dtype, device=inputs_embeds.device, past_key_values_length=past_key_values_length, ) if attention_mask is not None: # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len] expanded_attn_mask = _expand_mask(attention_mask, inputs_embeds.dtype, tgt_len=input_shape[-1]).to( inputs_embeds.device ) combined_attention_mask = ( expanded_attn_mask if combined_attention_mask is None else expanded_attn_mask + combined_attention_mask ) return combined_attention_mask @add_start_docstrings_to_model_forward(LLAMA_INPUTS_DOCSTRING) def forward( self, input_ids: torch.LongTensor = None, attention_mask: Optional[torch.Tensor] = None, position_ids: Optional[torch.LongTensor] = None, past_key_values: Optional[List[torch.FloatTensor]] = None, inputs_embeds: Optional[torch.FloatTensor] = None, use_cache: Optional[bool] = None, output_attentions: Optional[bool] = None, output_hidden_states: Optional[bool] = None, return_dict: Optional[bool] = None, use_active_externalism:Optional[bool]=None, long_range_past_key_values:Optional[List[Tuple[torch.FloatTensor]]] = None, faiss_indexes:Tuple=None, topk:int=None, ) -> Union[Tuple, BaseModelOutputWithPast]: output_attentions = ( output_attentions if output_attentions is not None else self.config.output_attentions ) output_hidden_states = ( output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states ) use_cache = use_cache if use_cache is not None else self.config.use_cache return_dict = return_dict if return_dict is not None else self.config.use_return_dict use_active_externalism = (use_active_externalism if use_active_externalism is not None else self.use_active_externalism) topk = (topk if topk is not None else self.topk) # retrieve input_ids and inputs_embeds if input_ids is not None and inputs_embeds is not None: raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time") elif input_ids is not None: batch_size, seq_length = input_ids.shape elif inputs_embeds is not None: batch_size, seq_length, _ = inputs_embeds.shape else: raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds") seq_length_with_past = seq_length past_key_values_length = 0 if past_key_values is not None: past_key_values_length = past_key_values[0][0].shape[2] seq_length_with_past = seq_length_with_past + past_key_values_length if position_ids is None: device = input_ids.device if input_ids is not None else inputs_embeds.device position_ids = torch.arange( seq_length_with_past, dtype=torch.long, device=device #range of position ids is total seq length since we apply rotary pos emb after reading from cache ) position_ids = position_ids.unsqueeze(0).view(-1, seq_length_with_past) else: position_ids = position_ids.view(-1, seq_length).long() if inputs_embeds is None: inputs_embeds = self.embed_tokens(input_ids) # embed positions if attention_mask is None: attention_mask = torch.ones( (batch_size, seq_length_with_past), dtype=torch.bool, device=inputs_embeds.device ) attention_mask = self._prepare_decoder_attention_mask( attention_mask, (batch_size, seq_length), inputs_embeds, past_key_values_length ) hidden_states = inputs_embeds if self.gradient_checkpointing and self.training: if use_cache: logger.warning_once( "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..." ) use_cache = False # decoder layers all_hidden_states = () if output_hidden_states else None all_self_attns = () if output_attentions else None next_decoder_cache = () if use_cache else None all_idx = () if output_attentions else None for idx, decoder_layer in enumerate(self.layers): if output_hidden_states: all_hidden_states += (hidden_states,) past_key_value = past_key_values[idx] if past_key_values is not None else None long_range_past_key_value = (long_range_past_key_values[idx] if (long_range_past_key_values is not None and self.use_active_externalism_by_layer[idx] and use_active_externalism is True) else None) if long_range_past_key_value is not None and faiss_indexes is not None: raise NotImplementedError( 'Using faiss and passing key value pairs manually are mutually exclusive right now.') if self.gradient_checkpointing and self.training: def create_custom_forward(module): def custom_forward(*inputs): # None for past_key_value return module(*inputs, output_attentions, None) return custom_forward layer_outputs = torch.utils.checkpoint.checkpoint( create_custom_forward(decoder_layer), hidden_states, attention_mask, position_ids, None, ) else: layer_outputs = decoder_layer( hidden_states, attention_mask=attention_mask, position_ids=position_ids, past_key_value=past_key_value, output_attentions=output_attentions, use_cache=use_cache, topk=topk, long_range_past_key_value=long_range_past_key_value, faiss_indexes=faiss_indexes, mask_by_sim=self.mask_by_sim, sim_threshold=self.sim_threshold, current_layer=idx, ) hidden_states = layer_outputs[0] if use_cache: next_decoder_cache += (layer_outputs[2 if output_attentions else 1],) if output_attentions: all_self_attns += (layer_outputs[1],) all_idx += (layer_outputs[3],) # record which memories were retrieved hidden_states = self.norm(hidden_states) # add hidden states from the last decoder layer if output_hidden_states: all_hidden_states += (hidden_states,) next_cache = next_decoder_cache if use_cache else None if not return_dict: return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None) return BaseModelOutputWithPast( last_hidden_state=hidden_states, past_key_values=next_cache, hidden_states=all_hidden_states, attentions=(all_self_attns, all_idx) ) class ExtendedLlamaForCausalLM(LlamaPreTrainedModel): _tied_weights_keys = ["lm_head.weight"] def __init__(self, config, external_memories=None, **kwargs): super().__init__(config) self.model = ExtendedLlamaModel(config) self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False) self.use_active_externalism = config.use_active_externalism self.memory_type = config.memory_type self.memory_device = config.memory_device self._memories = None if external_memories is not None: self._memories = external_memories self.memories = None # Initialize weights and apply final processing self.post_init() def get_input_embeddings(self): return self.model.embed_tokens def set_input_embeddings(self, value): self.model.embed_tokens = value def get_output_embeddings(self): return self.lm_head def set_output_embeddings(self, new_embeddings): self.lm_head = new_embeddings def set_decoder(self, decoder): self.model = decoder def get_decoder(self): return self.model @add_start_docstrings_to_model_forward(LLAMA_INPUTS_DOCSTRING) @replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC) def forward( self, input_ids: torch.LongTensor = None, attention_mask: Optional[torch.Tensor] = None, position_ids: Optional[torch.LongTensor] = None, past_key_values: Optional[List[torch.FloatTensor]] = None, inputs_embeds: Optional[torch.FloatTensor] = None, labels: Optional[torch.LongTensor] = None, use_cache: Optional[bool] = None, output_attentions: Optional[bool] = None, output_hidden_states: Optional[bool] = None, return_dict: Optional[bool] = None, use_active_externalism: Optional[bool]=None, topk:int=None ) -> Union[Tuple, CausalLMOutputWithPast]: r""" Args: labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): Labels for computing the masked language modeling loss. Indices should either be in `[0, ..., config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. Returns: Example: ```python >>> from transformers import AutoTokenizer, LlamaForCausalLM >>> model = LlamaForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS) >>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER) >>> prompt = "Hey, are you conscious? Can you talk to me?" >>> inputs = tokenizer(prompt, return_tensors="pt") >>> # Generate >>> generate_ids = model.generate(inputs.input_ids, max_length=30) >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0] "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you." ```""" if self._memories is not None and self.memories is None: #init memories once on first call self.memories = self.generate_cache(self._memories, cache_type=self.memory_type) output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions output_hidden_states = ( output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states ) return_dict = return_dict if return_dict is not None else self.config.use_return_dict use_active_externalism = (use_active_externalism if use_active_externalism is not None else self.use_active_externalism ) topk = topk if topk is not None else None long_range_past_key_values = None faiss_indexes = None if hasattr(self, "memories") and isinstance(self.memories, list): long_range_past_key_values = self.memories faiss_indexes = None elif hasattr(self, "memories"): long_range_past_key_values = None faiss_indexes = self.memories # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn) outputs = self.model( input_ids=input_ids, attention_mask=attention_mask, position_ids=position_ids, past_key_values=past_key_values, inputs_embeds=inputs_embeds, use_cache=use_cache, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, long_range_past_key_values=long_range_past_key_values, faiss_indexes=faiss_indexes, use_active_externalism=use_active_externalism, topk=topk ) hidden_states = outputs[0] logits = self.lm_head(hidden_states) loss = None if labels is not None: # Shift so that tokens < n predict n shift_logits = logits[..., :-1, :].contiguous() shift_labels = labels[..., 1:].contiguous() # Flatten the tokens loss_fct = CrossEntropyLoss() shift_logits = shift_logits.view(-1, self.config.vocab_size) shift_labels = shift_labels.view(-1) # Enable model parallelism shift_labels = shift_labels.to(shift_logits.device) loss = loss_fct(shift_logits, shift_labels) if not return_dict: output = (logits,) + outputs[1:] return (loss,) + output if loss is not None else output return CausalLMOutputWithPast( loss=loss, logits=logits, past_key_values=outputs.past_key_values, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) def generate_cache(self, input_ids:torch.LongTensor, stride:int=512, max_len:int=2048, cache_type:str='manual'): if cache_type not in ['manual', 'faiss']: raise NotImplementedError(f"Cache type {cache_type} not implemented.") prev_end_loc=0 long_range_past_key_values = None faiss_indexes= None for b_idx in range(0, input_ids.size(-1), stride): #generate kv-pairs using stride end_loc = min(b_idx + max_len, input_ids.size(-1)) trg_len = end_loc - prev_end_loc subseq = input_ids[:, b_idx:end_loc].to(self.model.device) with torch.no_grad(): outputs = self.model(subseq, use_cache=True, use_active_externalism=False) to_cache = [( kv[0][:,:,-trg_len:], kv[1][:,:,-trg_len:]) for kv in outputs.past_key_values ] long_range_past_key_values, faiss_indexes = self.cache(to_cache, cache_type, long_range_past_key_values=long_range_past_key_values, faiss_indexes=faiss_indexes) prev_end_loc = end_loc if end_loc == input_ids.size(-1): break if long_range_past_key_values is not None: return long_range_past_key_values else: return faiss_indexes def cache(self, to_cache:List, cache_type:str='manual', long_range_past_key_values:List=None, faiss_indexes:faiss.IndexFlatIP=None, max_length_cache=100000, verbose=False): if long_range_past_key_values is not None and faiss_indexes is not None: raise NotImplementedError("Using faiss and passing key value pairs manually are mutually exclusive right now.") if cache_type=='faiss': #add one-hot encoding to match layer, head indices one_hot_encodings = F.one_hot(torch.arange(0, self.config.n_heads*self.config.num_hidden_layers))*10 if faiss_indexes is None: faiss_indexes = (faiss.IndexFlatIP(to_cache[0][0].size(-1)+one_hot_encodings.size(-1)), faiss.IndexFlatIP(to_cache[0][1].size(-1)*2)) kn_index, kv_index = faiss_indexes for b_idx, (k, v) in enumerate(to_cache): k_n = (k/vector_norm(k, ord=2, dim=-1, keepdim=True)).to('cpu') k_n = torch.concat([rearrange(k_n, 'b h s d -> b (h s) d', h=self.config.n_heads), one_hot_encodings[self.config.n_heads*b_idx:self.config.n_heads*(b_idx+1)].unsqueeze(0).repeat_interleave(repeats=k.size(-2), dim=-2)], dim=-1) kn_index.add(k_n.squeeze().numpy()) k= rearrange(k, 'b h s d -> b (h s) d', h=self.config.n_heads) v= rearrange(v, 'b h s d -> b (h s) d', h=self.config.n_heads) kv_index.add(torch.concat([v.squeeze(), k.squeeze()], dim=1).to('cpu').numpy()) else: if long_range_past_key_values is None: long_range_past_key_values = [(k.to(self.memory_device),v.to(self.memory_device)) for k,v in to_cache] else: long_range_past_key_values = [ ( torch.concat([kv[0], to_cache[ind][0].to(self.memory_device)], dim=2), torch.concat([kv[1], to_cache[ind][1].to(self.memory_device)], dim=2) ) for ind, kv in enumerate(long_range_past_key_values) ] if long_range_past_key_values is not None: #set a limit on manual memory length if long_range_past_key_values[0][0].size(-2) > max_length_cache: long_range_past_key_values = [ ( kv[0][:, :, -max_length_cache:], kv[1][:, :, -max_length_cache:] ) for kv in long_range_past_key_values] if verbose: if cache_type == 'faiss': print(f"{kn_index.ntotal} keys in faiss index") else: print(f"{long_range_past_key_values[0][0].size(-2)} cached kvs") return long_range_past_key_values, (kn_index, kv_index) if cache_type == 'faiss' else None def prepare_inputs_for_generation( self, input_ids, past_key_values=None, attention_mask=None, inputs_embeds=None, **kwargs ): if past_key_values: input_ids = input_ids[:, -1:] position_ids = kwargs.get("position_ids", None) if attention_mask is not None and position_ids is None: # create position_ids on the fly for batch generation position_ids = attention_mask.long().cumsum(-1) - 1 position_ids.masked_fill_(attention_mask == 0, 1) if past_key_values: position_ids = position_ids[:, -1].unsqueeze(-1) # if `inputs_embeds` are passed, we only want to use them in the 1st generation step if inputs_embeds is not None and past_key_values is None: model_inputs = {"inputs_embeds": inputs_embeds} else: model_inputs = {"input_ids": input_ids} model_inputs.update( { "position_ids": position_ids, "past_key_values": past_key_values, "use_cache": kwargs.get("use_cache"), "attention_mask": attention_mask, 'use_active_externalism': kwargs.get('use_active_externalism'), #add a few more kwargs for active externalism 'topk': kwargs.get('topk', None), } ) return model_inputs @staticmethod def _reorder_cache(past_key_values, beam_idx): reordered_past = () for layer_past in past_key_values: reordered_past += ( tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past), ) return reordered_past