# coding=utf-8 # Copyright 2023 Mistral AI 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 VMistral model.""" from dataclasses import dataclass import inspect import math import warnings from typing import List, Optional, Tuple, Union import torch import torch.nn.functional as F import torch.utils.checkpoint from torch import nn from torch.nn import CrossEntropyLoss from transformers.activations import ACT2FN from transformers.modeling_attn_mask_utils import _prepare_4d_causal_attention_mask from transformers.utils import ( add_start_docstrings, add_start_docstrings_to_model_forward, is_flash_attn_2_available, replace_return_docstrings, ) from einops import rearrange, repeat from transformers import PreTrainedModel from transformers.utils import logging from transformers.modeling_outputs import ModelOutput from .configuration_vmistral import VMistralConfig from .vision import SiglipVisionModel if is_flash_attn_2_available(): from flash_attn import flash_attn_func, flash_attn_varlen_func from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input # noqa _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters) logger = logging.get_logger(__name__) _CONFIG_FOR_DOC = "VMistralConfig" VMistral_PRETRAINED_MODEL_ARCHIVE_LIST = [ "HuggingFaceM4/VLM_WebSight_finetuned" ] @dataclass class VMistralBaseModelOutputWithPast(ModelOutput): """ Base class for VMistral model's outputs that may also contain a past key/values (to speed up sequential decoding). Args: last_hidden_state (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`): Sequence of hidden-states at the output of the last layer of the model. If `past_key_values` is used only the last hidden-state of the sequences of shape `(batch_size, 1, hidden_size)` is output. 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 optionally if `config.is_encoder_decoder=True` 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 optionally if `config.is_encoder_decoder=True` in the cross-attention blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`): Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`. Hidden-states of the model at the output of each layer plus the optional initial embedding outputs. attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`): Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length, sequence_length)`. Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads. image_hidden_states (`tuple(torch.FloatTensor)`, *optional*): Tuple of `torch.FloatTensor` (one for the output of the image embeddings, `(batch_size, num_images, sequence_length, hidden_size)`. image_hidden_states of the model produced by the vision encoder, and optionally by the perceiver """ last_hidden_state: torch.FloatTensor = None past_key_values: Optional[Tuple[Tuple[torch.FloatTensor]]] = None hidden_states: Optional[Tuple[torch.FloatTensor]] = None attentions: Optional[Tuple[torch.FloatTensor]] = None image_hidden_states: Optional[Tuple[torch.FloatTensor]] = None @dataclass class VMistralCausalLMOutputWithPast(ModelOutput): """ Base class for VMistral causal language model (or autoregressive) outputs. Args: loss (`torch.FloatTensor` of shape `(1,)`, *optional*, returned when `labels` is provided): Language modeling loss (for next-token prediction). logits (`torch.FloatTensor` of shape `(batch_size, sequence_length, config.vocab_size)`): Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax). 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)`) Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see `past_key_values` input) to speed up sequential decoding. hidden_states (`tuple(torch.FloatTensor)`, *optional*, returned when `output_hidden_states=True` is passed or when `config.output_hidden_states=True`): Tuple of `torch.FloatTensor` (one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape `(batch_size, sequence_length, hidden_size)`. Hidden-states of the model at the output of each layer plus the optional initial embedding outputs. attentions (`tuple(torch.FloatTensor)`, *optional*, returned when `output_attentions=True` is passed or when `config.output_attentions=True`): Tuple of `torch.FloatTensor` (one for each layer) of shape `(batch_size, num_heads, sequence_length, sequence_length)`. Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads. image_hidden_states (`tuple(torch.FloatTensor)`, *optional*): Tuple of `torch.FloatTensor` (one for the output of the image embeddings, `(batch_size, num_images, sequence_length, hidden_size)`. image_hidden_states of the model produced by the vision encoder, and optionally by the perceiver """ loss: Optional[torch.FloatTensor] = None logits: torch.FloatTensor = None past_key_values: Optional[List[torch.FloatTensor]] = None hidden_states: Optional[Tuple[torch.FloatTensor]] = None attentions: Optional[Tuple[torch.FloatTensor]] = None image_hidden_states: Optional[Tuple[torch.FloatTensor]] = None def expand_inputs_for_generation( input_ids, expand_size=1, is_encoder_decoder=False, attention_mask=None, encoder_outputs=None, **model_kwargs, ): expanded_return_idx = ( torch.arange(input_ids.shape[0]).view(-1, 1).repeat(1, expand_size).view(-1).to(input_ids.device) ) input_ids = input_ids.index_select(0, expanded_return_idx) model_kwargs["pixel_values"] = model_kwargs.get("pixel_values", None) model_kwargs["image_hidden_states"] = model_kwargs.get("image_hidden_states", None) if "token_type_ids" in model_kwargs: token_type_ids = model_kwargs["token_type_ids"] model_kwargs["token_type_ids"] = token_type_ids.index_select(0, expanded_return_idx) if attention_mask is not None: model_kwargs["attention_mask"] = attention_mask.index_select(0, expanded_return_idx) if model_kwargs["pixel_values"] is not None: model_kwargs["pixel_values"] = model_kwargs["pixel_values"].index_select(0, expanded_return_idx) elif model_kwargs["image_hidden_states"] is not None: model_kwargs["image_hidden_states"] = model_kwargs["image_hidden_states"].index_select(0, expanded_return_idx) return input_ids, model_kwargs def update_model_kwargs_for_generation(outputs, model_kwargs): # must have this key set to at least None if "past_key_values" in outputs: model_kwargs["past_key_values"] = outputs.past_key_values else: model_kwargs["past_key_values"] = None # update token_type_ids with last value if "token_type_ids" in model_kwargs: token_type_ids = model_kwargs["token_type_ids"] model_kwargs["token_type_ids"] = torch.cat([token_type_ids, token_type_ids[:, -1].unsqueeze(-1)], dim=-1) # update attention masks if "attention_mask" in model_kwargs: attention_mask = model_kwargs["attention_mask"] model_kwargs["attention_mask"] = torch.cat( [attention_mask, attention_mask.new_ones((attention_mask.shape[0], 1))], dim=-1 ) # Get the precomputed image_hidden_states model_kwargs["image_hidden_states"] = outputs.image_hidden_states return model_kwargs def prepare_inputs_for_generation(input_ids, past_key_values=None, **kwargs): token_type_ids = kwargs.get("token_type_ids", None) # only last token for inputs_ids if past is defined in kwargs if past_key_values: input_ids = input_ids[:, -1].unsqueeze(-1) if token_type_ids is not None: token_type_ids = token_type_ids[:, -1].unsqueeze(-1) attention_mask = kwargs.get("attention_mask", None) 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) pixel_values = kwargs.get("pixel_values", None) image_hidden_states = kwargs.get("image_hidden_states", None) return { "input_ids": input_ids, "past_key_values": past_key_values, "use_cache": kwargs.get("use_cache"), "position_ids": position_ids, "attention_mask": attention_mask, "token_type_ids": token_type_ids, "pixel_values": pixel_values, "image_hidden_states": image_hidden_states, } def freeze_model(model, module_exceptions=[]): mapping = { "LayerNorm": nn.LayerNorm, "Linear": nn.Linear, "Embedding": nn.Embedding, } module_exceptions_mapped = [mapping[m] for m in module_exceptions] for module in model.modules(): if module_exceptions and any([isinstance(module, t) for t in module_exceptions_mapped]): module.requires_grad_(True) # Explicitly setting it to true to avoid any mistakes else: module.requires_grad_(False) return model class DecoupledEmbedding(nn.Embedding): # Derived from https://pytorch.org/docs/stable/_modules/torch/nn/modules/sparse.html#Embedding """ Implements a decoupling of parameters to allow freezing (or not) a subset of the embeddings. In practise, the regular `weight` can be trained or frozen (i.e. `partially_freeze=True`), and if `num_additional_embeddings` > 0, then it will create `num_additional_embeddings` additional parameters that are always trained. If `num_additional_embeddings=0`, then the module defaults back to the regular behavior of `nn.Embedding`. """ def __init__( self, num_embeddings, num_additional_embeddings, embedding_dim, partially_freeze=False, device=None, dtype=None, padding_idx=None, **kwargs, ) -> None: """ num_additional_embeddings: int. Number of additional embeddings. Only useful when you `partially_freeze=True`. partially_freeze: bool. If True, the regular `weight` will be frozen. `additional_weight` is never frozen. Note: there are a lot of other parameters to initialize a standard `nn.Embedding` such as `padding_idx`, `max_norm` or `norm_type`. We are not supporting these. """ if padding_idx is not None and padding_idx > num_embeddings: raise ValueError(f"padding_idx must be within num_embeddings. Got {padding_idx} and {num_embeddings}") super().__init__( num_embeddings=num_embeddings, embedding_dim=embedding_dim, device=device, dtype=dtype, padding_idx=padding_idx, **kwargs, ) self.num_embeddings = num_embeddings self.padding_idx = padding_idx self.num_additional_embeddings = num_additional_embeddings self.partially_freeze = partially_freeze if partially_freeze: self.weight.requires_grad_(False) if self.num_additional_embeddings > 0: self.additional_embedding = nn.Embedding( num_embeddings=self.num_additional_embeddings, embedding_dim=embedding_dim, device=device, dtype=dtype, ) def forward(self, input_ids): """ we have 2 embeddings, with different indices - one pretrained self.weight and another self.additional_embedding.weight that is being trained. in order to make a lookup of the input ids, we: 1. find out the indices of the entries belonging to the 2nd embedding 2. extract those values while subtracting the size of the first embedding (num_embeddings), since the 2nd embedding starts from 0 and not num_embeddings 3. perform the 2nd embedding lookup 4. now we handle the 1st embedding, we overwrite indices belonging to the 2nd embedding with a padding index 5. perform the 1st embedding lookup 6. now we overwrite the values in the 1st embedding lookup with the values of the 2nd embedding lookup note: for the 1st embedding lookup we could have looked up only the low indices and not do the padding, but then we have to create a new tensor and populate it with 2 tensors that are spread out across various indices - i.e. not a simple concat - I haven't benchmarked the complex case if it's any faster, given that seqlens are usually relatively short it's probably not faster or if faster not by much - but might be a good idea to measure. """ if self.num_additional_embeddings == 0: return self.additional_embedding(input_ids) # Clone so that we don't modify the original input_ids later on input_ids = input_ids.clone() additional_vocab_indices = torch.where(input_ids >= self.num_embeddings) input_ids_additional_vocab = input_ids[additional_vocab_indices] additional_embeddings = self.additional_embedding(input_ids_additional_vocab - self.num_embeddings) # for successful lookup replace input_ids with 0, the results of these will be discarded anyway input_ids[additional_vocab_indices] = 0 full_vector = F.embedding(input_ids, self.weight) # overwrite the records with high indices full_vector[additional_vocab_indices] = additional_embeddings return full_vector def extra_repr(self) -> str: return "num_embeddings={}, num_additional_embeddings={}, embedding_dim={}, partially_freeze={}".format( self.num_embeddings, self.num_additional_embeddings, self.embedding_dim, self.partially_freeze, ) class DecoupledLinear(nn.Linear): # Derived from https://pytorch.org/docs/stable/_modules/torch/nn/modules/linear.html#Linear """ Implements a decoupling of parameters to allow freezing (or not) a subset of the parameters. In practise, the regular `weight` can be trained or frozen (i.e. `partially_freeze=True`), and if `out_additional_features` > 0, then it will create `out_additional_features * in_features` additional parameters that are always trained. If `out_additional_features=0`, then the module defaults back to the regular behavior of `nn.Linear`. """ def __init__( self, in_features: int, out_features: int, out_additional_features: int = 0, bias: bool = True, partially_freeze: bool = True, device=None, dtype=None, ) -> None: """ out_additional_features: int. Number of additional trainable dimensions. Only makes sense when `partially_freeze=True`. partially_freeze: bool. If True, the regular `weight` will be frozen and extra parameters (if any) will be trainable. If False, default to the regular behavior of nn.Linear. """ super().__init__(in_features, out_features, bias, device, dtype) self.out_additional_features = out_additional_features self.partially_freeze = partially_freeze self.in_features = in_features self.out_features = out_features if partially_freeze: self.weight.requires_grad_(False) if bias: self.bias.requires_grad_(False) if out_additional_features > 0: self.additional_fc = nn.Linear( in_features=in_features, out_features=out_additional_features, bias=bias, device=device, dtype=dtype, ) def forward(self, input: torch.Tensor) -> torch.Tensor: output = F.linear(input, self.weight, self.bias) if self.out_additional_features > 0: additional_features = self.additional_fc(input) output = torch.cat((output, additional_features), -1) return output def extra_repr(self) -> str: """Overwriting `nn.Linear.extra_repr` to include new parameters.""" return "in_features={}, out_features={}, out_additional_features={}, bias={}, partially_freeze={}".format( self.in_features, self.out_features, self.out_additional_features, self.bias is not None, self.partially_freeze, ) class SwiGLU(nn.Module): def __init__(self, embed_dim) -> None: super().__init__() self.fc1 = nn.Linear(embed_dim, embed_dim, bias=False) self.fc2 = nn.Linear(embed_dim, embed_dim, bias=False) def forward(self, x: torch.Tensor) -> torch.Tensor: x_1 = self.fc1(x) x_1 = torch.mul(x_1, torch.sigmoid(x_1)) x_2 = self.fc2(x) x = torch.mul(x_1, x_2) return x class ModalityProjection(nn.Module): def __init__(self, embed_dim_in, embed_dim_out) -> None: super().__init__() self.fc1 = nn.Linear(embed_dim_in, embed_dim_out, bias=False) self.act = SwiGLU(embed_dim_out) self.fc2 = nn.Linear(embed_dim_out, embed_dim_out, bias=False) def forward(self, x: torch.Tensor) -> torch.Tensor: x = self.fc1(x) x = self.act(x) x = self.fc2(x) return x class PerceiverResampler(nn.Module): def __init__( self, embed_dim: int, depth: int, n_heads: int, head_dim: int, n_latents: int, qk_layer_norms: bool ) -> None: """ Instantiates a Perceiver Resampler that operates over a sequence of embeddings (say from a ResNet or ViT or MAE) of a given dimension, performs `depth` blocks of cross-attention with a fixed `n_latents` inputs, then returns a Tensor of shape [bsz, n_latents, embed_dim]. :param embed_dim: Dimensionality of embeddings being fed to the Perceiver Resampler (also dimensionality of latent embeddings *returned* by the Perceiver Resampler. Could be e.g., VIT embed_dim, ResNet pool dim, and so on. :param depth: Depth of the Perceiver Resampler (Transformer w/ cross attention). Should be shallow (< 3). :param n_heads: Number of heads in each Transformer block (for multi-headed self-attention). :param head_dim: Dimensionality of each head projection in the Transformer block. :param n_latents: Number of latent embeddings to resample ("compress") the input sequence to (usually < 128). """ super().__init__() self.embed_dim, self.n_heads, self.head_dim, self.n_latents = embed_dim, n_heads, head_dim, n_latents self.qk_layer_norms = qk_layer_norms # Create Latents for Perceiver self.latents = nn.Parameter(torch.ones(self.n_latents, self.embed_dim)) self.intermediate_dim = self.embed_dim * 4 # Create Transformer Blocks self.blocks = nn.ModuleList( [ nn.ModuleList( [ PerceiverAttention(self.embed_dim, self.n_heads, self.head_dim, self.qk_layer_norms), MLP(self.embed_dim, self.intermediate_dim), ] ) for _ in range(depth) ] ) self.layer_norm = nn.LayerNorm(self.embed_dim) def forward(self, context: torch.Tensor) -> torch.Tensor: """Resample arbitrary length context & *compress* down to self.n_latents latent embeddings""" latents = repeat(self.latents, "seq embed -> bsz seq embed", bsz=context.shape[0]) # Feed through Perceiver Attention blocks... for attn, ff in self.blocks: latents = attn(context, latents) + latents latents = ff(latents) + latents return self.layer_norm(latents) class PerceiverAttention(nn.Module): def __init__(self, embed_dim: int, n_heads: int, head_dim: int, qk_layer_norms: bool) -> None: """Perceiver Cross-Attention Module --> let long-form inputs be `context`, resampled embeddings be `latents`""" super().__init__() self.embed_dim, self.n_heads, self.head_dim = embed_dim, n_heads, head_dim self.qk_layer_norms = qk_layer_norms # Normalization & Scaling self.context_layer_norm = nn.LayerNorm(self.embed_dim) self.latents_layer_norm = nn.LayerNorm(self.embed_dim) if self.qk_layer_norms: self.q_layer_norm = nn.LayerNorm(self.head_dim) self.k_layer_norm = nn.LayerNorm(self.head_dim) self.qk_scale = self.head_dim**-0.5 # Q, K, V Projection (no bias -- detail from Perceiver/Flamingo Papers). self.q_proj = nn.Linear(self.embed_dim, self.n_heads * self.head_dim, bias=False) self.k_proj = nn.Linear(self.embed_dim, self.n_heads * self.head_dim, bias=False) self.v_proj = nn.Linear(self.embed_dim, self.n_heads * self.head_dim, bias=False) self.output_proj = nn.Linear(self.n_heads * self.head_dim, self.embed_dim, bias=False) def forward(self, context: torch.Tensor, latents: torch.Tensor) -> torch.Tensor: """ Runs Perceiver Self-Attention, with special (context, latents) appended along the `seq` dimension! :param context: Tensor of shape [bsz, seq, embed_dim] representing long-form context to resample. :param latents: Tensor of shape [bsz, n_latents, embed_dim] representing fixed length latents to compress to. :return: Tensor of shape [bsz, n_latents, embed_dim] representing attention over latents w/ cross from context. """ context = self.context_layer_norm(context) latents = self.latents_layer_norm(latents) # Query, Key, Value Projections --> Note that in Flamingo, latents are *concatenated* with context prior to attn! # Note: This results in queries w/ `seq = n_latents`, and keys, values with `seq = len(context) + n_latents` q = self.q_proj(latents) k = self.k_proj(torch.cat([context, latents], dim=-2)) v = self.v_proj(torch.cat([context, latents], dim=-2)) # Multiheaded Self-Attention w/ stable softmax (subtract per-row max -- `amax` -- before softmax call) # =>> `attn` should be a 2D matrix of shape [n_latents x (context + n_latents)] q, k, v = [rearrange(x, "bsz seq (heads embed) -> bsz heads seq embed", heads=self.n_heads) for x in (q, k, v)] if self.qk_layer_norms: q = self.q_layer_norm(q) k = self.k_layer_norm(k) scores = torch.einsum("... i d, ... j d -> ... i j", q * self.qk_scale, k) stabilized_scores = scores - (scores.amax(dim=-1, keepdim=True).detach()) attn = stabilized_scores.softmax(dim=-1) # Attend & project back to output... resampled = torch.einsum("... i j, ... j d -> ... i d", attn, v) return self.output_proj( rearrange(resampled, "bsz heads seq embed -> bsz seq (heads embed)", heads=self.n_heads) ) class MLP(nn.Module): def __init__(self, embed_dim, intermediate_size): """Simple MLP block with intermediate_size and embedding size""" super().__init__() self.embed_dim = embed_dim self.ln = nn.LayerNorm(self.embed_dim) self.fc = nn.Linear(self.embed_dim, intermediate_size, bias=False) self.act = nn.ReLU() self.c_proj = nn.Linear(intermediate_size, self.embed_dim, bias=False) def forward(self, hidden_states: Optional[Tuple[torch.FloatTensor]]) -> torch.FloatTensor: hidden_states = self.ln(hidden_states) hidden_states = self.fc(hidden_states) hidden_states = self.act(hidden_states) hidden_states = self.c_proj(hidden_states) return hidden_states # Copied from transformers.models.llama.modeling_llama._get_unpad_data def _get_unpad_data(attention_mask): seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32) indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten() max_seqlen_in_batch = seqlens_in_batch.max().item() cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.torch.int32), (1, 0)) return ( indices, cu_seqlens, max_seqlen_in_batch, ) # Copied from transformers.models.llama.modeling_llama.LlamaRMSNorm with Llama->Mistral class MistralRMSNorm(nn.Module): def __init__(self, hidden_size, eps=1e-6): """ MistralRMSNorm 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 hidden_states = hidden_states.to(torch.float32) variance = hidden_states.pow(2).mean(-1, keepdim=True) hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon) return self.weight * hidden_states.to(input_dtype) # Copied from transformers.models.llama.modeling_llama.LlamaRotaryEmbedding with Llama->Mistral class MistralRotaryEmbedding(nn.Module): def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None): super().__init__() self.dim = dim self.max_position_embeddings = max_position_embeddings self.base = base inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim)) self.register_buffer("inv_freq", inv_freq, persistent=False) # Build here to make `torch.jit.trace` work. self._set_cos_sin_cache( seq_len=max_position_embeddings, device=self.inv_freq.device, dtype=torch.get_default_dtype() ) def _set_cos_sin_cache(self, seq_len, device, dtype): self.max_seq_len_cached = seq_len t = torch.arange(self.max_seq_len_cached, device=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) self.register_buffer("cos_cached", emb.cos().to(dtype), persistent=False) self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False) def forward(self, x, seq_len=None): # x: [bs, num_attention_heads, seq_len, head_size] if seq_len > self.max_seq_len_cached: self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype) return ( self.cos_cached[:seq_len].to(dtype=x.dtype), self.sin_cached[:seq_len].to(dtype=x.dtype), ) # Copied from transformers.models.llama.modeling_llama.rotate_half 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) # Copied from transformers.models.llama.modeling_llama.apply_rotary_pos_emb def apply_rotary_pos_emb(q, k, cos, sin, position_ids): cos = cos[position_ids].unsqueeze(1) # [seq_len, dim] -> [batch_size, 1, seq_len, head_dim] sin = sin[position_ids].unsqueeze(1) q_embed = (q * cos) + (rotate_half(q) * sin) k_embed = (k * cos) + (rotate_half(k) * sin) return q_embed, k_embed class MistralMLP(nn.Module): def __init__(self, config): super().__init__() self.config = config self.hidden_size = config.hidden_size self.intermediate_size = config.intermediate_size self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False) self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False) self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False) self.act_fn = ACT2FN[config.hidden_act] def forward(self, x): return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x)) # Copied from transformers.models.llama.modeling_llama.repeat_kv def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor: """ This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch, num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim) """ batch, num_key_value_heads, slen, head_dim = hidden_states.shape if n_rep == 1: return hidden_states hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim) return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim) class MistralAttention(nn.Module): """ Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer and "Generating Long Sequences with Sparse Transformers". """ def __init__(self, config: VMistralConfig, qk_layer_norms: bool = False): 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.num_key_value_heads = config.num_key_value_heads self.num_key_value_groups = self.num_heads // self.num_key_value_heads self.max_position_embeddings = config.max_position_embeddings self.rope_theta = config.rope_theta self.is_causal = True 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_key_value_heads * self.head_dim, bias=False) self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=False) self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False) self.qk_layer_norms = qk_layer_norms if self.qk_layer_norms: self.q_layer_norm = MistralRMSNorm(self.head_dim, eps=config.rms_norm_eps) self.k_layer_norm = MistralRMSNorm(self.head_dim, eps=config.rms_norm_eps) self.rotary_emb = MistralRotaryEmbedding( self.head_dim, max_position_embeddings=self.max_position_embeddings, base=self.rope_theta, ) self.attention_dropout = config.attention_dropout 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, key_value_states: Optional[torch.Tensor] = None, 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, **kwargs, ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: if "padding_mask" in kwargs: warnings.warn( "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use" " `attention_mask` instead.`" ) 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_key_value_heads, self.head_dim).transpose(1, 2) ) value_states = ( self.v_proj(hidden_states).view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) ) kv_seq_len = key_states.shape[-2] if past_key_value is not None: kv_seq_len += past_key_value[0].shape[-2] cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len) query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids) 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 if self.qk_layer_norms: query_states = self.q_layer_norm(query_states) key_states = self.k_layer_norm(key_states) # repeat k/v heads if n_kv_heads < n_heads key_states = repeat_kv(key_states, self.num_key_value_groups) value_states = repeat_kv(value_states, self.num_key_value_groups) 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 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()}" ) attn_weights = attn_weights + attention_mask # 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).contiguous() 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 return attn_output, attn_weights, past_key_value class MistralFlashAttention2(MistralAttention): """ Mistral flash attention module. This module inherits from `MistralAttention` as the weights of the module stays untouched. The only required change would be on the forward pass where it needs to correctly call the public API of flash attention and deal with padding tokens in case the input contains any of them. """ 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, **kwargs, ): if "padding_mask" in kwargs: warnings.warn( "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use" " `attention_mask` instead.`" ) # overwrite attention_mask with padding_mask attention_mask = kwargs.pop("padding_mask") bsz, q_len, _ = hidden_states.size() query_states = self.q_proj(hidden_states) key_states = self.k_proj(hidden_states) value_states = self.v_proj(hidden_states) query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) kv_seq_len = key_states.shape[-2] if past_key_value is not None: kv_seq_len += past_key_value[0].shape[-2] # Because the input can be padded, the absolute sequence length depends on the max position id. rotary_seq_len = max(kv_seq_len, position_ids[:, -1].max().item()) + 1 cos, sin = self.rotary_emb(value_states, seq_len=rotary_seq_len) query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids) use_sliding_windows = ( _flash_supports_window_size and hasattr(self.config, "sliding_window") is not None and kv_seq_len > self.config.sliding_window ) if not _flash_supports_window_size: logger.warning_once( "The current flash attention version does not support sliding window attention, for a more memory" " efficient implementation make sure to upgrade flash-attn library." ) if past_key_value is not None: # Activate slicing cache only if the config has a value `sliding_windows` attribute if hasattr(self.config, "sliding_window") and kv_seq_len > self.config.sliding_window: slicing_tokens = kv_seq_len - self.config.sliding_window past_key = past_key_value[0] past_value = past_key_value[1] past_key = past_key[:, :, slicing_tokens:, :].contiguous() past_value = past_value[:, :, slicing_tokens:, :].contiguous() if past_key.shape[-2] != self.config.sliding_window - 1: raise ValueError( "past key must have a shape of (`batch_size, num_heads, self.config.sliding_window-1," f" head_dim`), got {past_key.shape}" ) past_key_value = (past_key, past_value) if attention_mask is not None: attention_mask = attention_mask[:, slicing_tokens:] attention_mask = torch.cat([attention_mask, torch.ones_like(attention_mask[:, -1:])], dim=-1) 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 # repeat k/v heads if n_kv_heads < n_heads key_states = repeat_kv(key_states, self.num_key_value_groups) value_states = repeat_kv(value_states, self.num_key_value_groups) dropout_rate = 0.0 if not self.training else self.attention_dropout # In PEFT, usually we cast the layer norms in float32 for training stability reasons # therefore the input hidden states gets silently casted in float32. Hence, we need # cast them back in float16 just to be sure everything works as expected. input_dtype = query_states.dtype if input_dtype == torch.float32: # Handle the case where the model is quantized if hasattr(self.config, "_pre_quantization_dtype"): target_dtype = self.config._pre_quantization_dtype else: target_dtype = self.q_proj.weight.dtype logger.warning_once( "The input hidden states seems to be silently casted in float32, this might be related to the fact" " you have upcasted embedding or layer norm layers in float32. We will cast back the input in" f" {target_dtype}." ) query_states = query_states.to(target_dtype) key_states = key_states.to(target_dtype) value_states = value_states.to(target_dtype) # Reashape to the expected shape for Flash Attention query_states = query_states.transpose(1, 2) key_states = key_states.transpose(1, 2) value_states = value_states.transpose(1, 2) attn_output = self._flash_attention_forward( query_states, key_states, value_states, attention_mask, q_len, dropout=dropout_rate, use_sliding_windows=use_sliding_windows, ) attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous() attn_output = self.o_proj(attn_output) if not output_attentions: attn_weights = None return attn_output, attn_weights, past_key_value def _flash_attention_forward( self, query_states, key_states, value_states, attention_mask, query_length, dropout=0.0, softmax_scale=None, use_sliding_windows=False, ): """ Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token first unpad the input, then computes the attention scores and pad the final attention scores. Args: query_states (`torch.Tensor`): Input query states to be passed to Flash Attention API key_states (`torch.Tensor`): Input key states to be passed to Flash Attention API value_states (`torch.Tensor`): Input value states to be passed to Flash Attention API attention_mask (`torch.Tensor`): The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the position of padding tokens and 1 for the position of non-padding tokens. dropout (`int`, *optional*): Attention dropout softmax_scale (`float`, *optional*): The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim) use_sliding_windows (`bool`, *optional*): Whether to activate sliding window attention. """ # Contains at least one padding token in the sequence if attention_mask is not None: batch_size = query_states.shape[0] query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input( query_states, key_states, value_states, attention_mask, query_length ) cu_seqlens_q, cu_seqlens_k = cu_seq_lens max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens if not use_sliding_windows: attn_output_unpad = flash_attn_varlen_func( query_states, key_states, value_states, cu_seqlens_q=cu_seqlens_q, cu_seqlens_k=cu_seqlens_k, max_seqlen_q=max_seqlen_in_batch_q, max_seqlen_k=max_seqlen_in_batch_k, dropout_p=dropout, softmax_scale=softmax_scale, causal=self.is_causal, ) else: attn_output_unpad = flash_attn_varlen_func( query_states, key_states, value_states, cu_seqlens_q=cu_seqlens_q, cu_seqlens_k=cu_seqlens_k, max_seqlen_q=max_seqlen_in_batch_q, max_seqlen_k=max_seqlen_in_batch_k, dropout_p=dropout, softmax_scale=softmax_scale, causal=self.is_causal, window_size=(self.config.sliding_window, self.config.sliding_window), ) attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length) else: if not use_sliding_windows: attn_output = flash_attn_func( query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=self.is_causal, ) else: attn_output = flash_attn_func( query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=self.is_causal, window_size=(self.config.sliding_window, self.config.sliding_window), ) return attn_output def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length): batch_size, kv_seq_len, num_heads, head_dim = key_layer.shape # On the first iteration we need to properly re-create the padding mask # by slicing it on the proper place if kv_seq_len != attention_mask.shape[-1]: attention_mask_num_tokens = attention_mask.shape[-1] attention_mask = attention_mask[:, attention_mask_num_tokens - kv_seq_len :] indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask) key_layer = index_first_axis(key_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k) value_layer = index_first_axis(value_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k) if query_length == kv_seq_len: query_layer = index_first_axis( query_layer.reshape(batch_size * kv_seq_len, num_heads, head_dim), indices_k ) cu_seqlens_q = cu_seqlens_k max_seqlen_in_batch_q = max_seqlen_in_batch_k indices_q = indices_k elif query_length == 1: max_seqlen_in_batch_q = 1 cu_seqlens_q = torch.arange( batch_size + 1, dtype=torch.int32, device=query_layer.device ) # There is a memcpy here, that is very bad. indices_q = cu_seqlens_q[:-1] query_layer = query_layer.squeeze(1) else: # The -q_len: slice assumes left padding. attention_mask = attention_mask[:, -query_length:] query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask) return ( query_layer, key_layer, value_layer, indices_q, (cu_seqlens_q, cu_seqlens_k), (max_seqlen_in_batch_q, max_seqlen_in_batch_k), ) class MistralDecoderLayer(nn.Module): def __init__(self, config: VMistralConfig): super().__init__() self.hidden_size = config.hidden_size self.self_attn = ( MistralAttention(config=config) if not getattr(config, "_flash_attn_2_enabled", False) else MistralFlashAttention2(config) ) self.mlp = MistralMLP(config) self.input_layernorm = MistralRMSNorm(config.hidden_size, eps=config.rms_norm_eps) self.post_attention_layernorm = MistralRMSNorm(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, **kwargs, ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]: if "padding_mask" in kwargs: warnings.warn( "Passing `padding_mask` is deprecated and will be removed in v4.37. Please make sure use" " `attention_mask` instead.`" ) """ 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, sequence_length)` where padding elements are indicated by 0. 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 = 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, ) 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,) return outputs MISTRAL_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 ([`VMistralConfig`]): 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 Mistral Model outputting raw hidden-states without any specific head on top.", MISTRAL_START_DOCSTRING, ) class VMistralPreTrainedModel(PreTrainedModel): config_class = VMistralConfig base_model_prefix = "model" supports_gradient_checkpointing = True _no_split_modules = ["MistralDecoderLayer"] _skip_keys_device_placement = "past_key_values" _supports_sdpa = False def _init_weights(self, module): # important: this ported version of the model isn't meant for training from scratch - only # inference and fine-tuning - so the proper init weights code has been removed - the m4 code # base should be used for training from scratch and it contains the correct code. 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_() # @classmethod # def override_vision_model_wrapper(cls, model, config, vision_model_name, vision_model_params, torch_dtype): # # this can be called via from_pretrained from a class w/ head or w/o head so we extract the beheaded model version # beheaded_model = model.model if hasattr(model, "model") else model # cls.override_vision_model(beheaded_model, vision_model_name, vision_model_params, torch_dtype) # beheaded_model.freeze_relevant_params(config) MISTRAL_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 Mistral Model outputting raw hidden-states without any specific head on top.", MISTRAL_START_DOCSTRING, ) class VMistralModel(VMistralPreTrainedModel): """ Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`MistralDecoderLayer`] Args: config: VMistralConfig """ def __init__(self, config: VMistralConfig, vision_model=None): super().__init__(config) self.config = config self.padding_idx = config.pad_token_id self.vocab_size = config.vocab_size self.sliding_window = config.sliding_window self.embed_tokens = DecoupledEmbedding( num_embeddings=config.vocab_size, num_additional_embeddings=config.additional_vocab_size, embedding_dim=config.hidden_size, partially_freeze=config.freeze_text_layers, padding_idx=self.padding_idx, ) # Load an uninitialized model and later in from_pretrained will load the pre-trained model - # this solves the losing of weights in `from_pretrained` on the main model self.vision_model = SiglipVisionModel(config.vision_config) # Dim projection - projecting from the vision dim to the text dim self.modality_projection = ModalityProjection( embed_dim_in=self.config.vision_config.hidden_size, embed_dim_out=self.config.hidden_size ) # Perceiver Resampler if config.use_resampler: self.perceiver_resampler = PerceiverResampler( config.hidden_size, config.perceiver_config.resampler_depth, config.perceiver_config.resampler_n_heads, config.perceiver_config.resampler_head_dim, config.perceiver_config.resampler_n_latents, config.perceiver_config.qk_layer_norms_perceiver, ) if config.use_resampler: self.image_seq_len = config.perceiver_config.resampler_n_latents else: self.image_seq_len = ( config.vision_config.image_size // config.vision_config.patch_size ) ** 2 # TODO: pretty sure that does not work for CLIP models since there is the CLS token self.image_token_id = self.config.image_token_id self.layers = nn.ModuleList([MistralDecoderLayer(config) for _ in range(config.num_hidden_layers)]) self.gradient_checkpointing = False self.norm = MistralRMSNorm(config.hidden_size, eps=config.rms_norm_eps) # Initialize weights and apply final processing self.post_init() self.freeze_relevant_params(config) def freeze_relevant_params(self, config=None): if config is None: config = self.config if config.freeze_text_layers: self.freeze_text_layers(config.freeze_text_module_exceptions) if config.freeze_vision_layers: freeze_model(self.vision_model, module_exceptions=config.freeze_vision_module_exceptions) def freeze_text_layers(self, module_exceptions): for module in [self.layers, self.norm]: freeze_model(module, module_exceptions=module_exceptions) def get_input_embeddings(self): return self.embed_tokens def set_input_embeddings(self, value): self.embed_tokens = value def inputs_merger( self, input_ids: torch.LongTensor = None, inputs_embeds: Optional[torch.Tensor] = None, image_hidden_states: Optional[torch.Tensor] = None, ): """ This method aims at merging the token embeddings with the image hidden states into one single sequence of vectors that are fed to the transformer LM. The merging happens as follows: - The text token sequence is: `tok_1 tok_2 tok_3 ... tok_4`. - We get the image hidden states for the image through the vision encoder (and potentially the perceiver), and that hidden state is then projected into the text embedding space. We thus have a sequence of image hidden states of size (1, image_seq_len, hidden_dim), where 1 is for batch_size of 1 image and hidden_dim is the hidden_dim of the LM transformer. - The merging happens so that we obtain the following sequence: `vector_tok_1 vector_tok_2 vector_tok_3 vector_fake_tok_around_image {sequence of image_seq_len image hidden states} vector_fake_toke_around_image vector_tok_4`. That sequence is fed to the LM. - To fit the format of that sequence, `input_ids`, `input_embeds`, `attention_mask` are all 3 adapted to insert the image hidden states. """ batch_size = input_ids.size(0) if inputs_embeds is not None: new_inputs_embeds = inputs_embeds.clone() if image_hidden_states is not None: vision_pipeline_output_seq_len = image_hidden_states.shape[1] vision_hidden_size = image_hidden_states.shape[2] # Get the number of images for each example num_images = (input_ids == self.image_token_id).sum(dim=-1) // self.image_seq_len cum_num_images = num_images.cumsum(dim=-1) for batch_idx in range(batch_size): # Get the number of images for this particular example example_num_images = num_images[batch_idx] # Get the image_hidden_states corresponding to True images for the example, so get rid of the padding images. start = 0 if batch_idx == 0 else cum_num_images[batch_idx - 1] end = cum_num_images[batch_idx] example_true_image_hidden_states = image_hidden_states[start:end] if ( new_inputs_embeds[batch_idx][input_ids[batch_idx] == self.image_token_id].shape[0] != example_num_images * vision_pipeline_output_seq_len ): raise ValueError( "new_inputs_embeds to replace has shape[0]:" f" {new_inputs_embeds[batch_idx][input_ids[batch_idx] == self.image_token_id].shape[0]} but" " should have shape[0]:" f" {example_num_images}*{vision_pipeline_output_seq_len}={example_num_images * vision_pipeline_output_seq_len} " ) # Insert the image_hidden_states new_inputs_embeds[batch_idx][input_ids[batch_idx] == self.image_token_id] = ( example_true_image_hidden_states.view( example_num_images * vision_pipeline_output_seq_len, vision_hidden_size, ) ) return_dict = {} if inputs_embeds is not None: return_dict["inputs_embeds"] = new_inputs_embeds return return_dict @add_start_docstrings_to_model_forward(MISTRAL_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, pixel_values: Optional[torch.FloatTensor] = None, image_hidden_states: 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, ) -> Union[Tuple, VMistralBaseModelOutputWithPast]: device = input_ids.device if input_ids is not None else inputs_embeds.device 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 # 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( past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device ) position_ids = position_ids.unsqueeze(0).view(-1, seq_length) else: position_ids = position_ids.view(-1, seq_length).long() if inputs_embeds is None: inputs_embeds = self.embed_tokens(input_ids) # START VISUAL INPUTS INTEGRATION if pixel_values is not None and image_hidden_states is not None: raise ValueError("You cannot specify both pixel_values and image_hidden_states at the same time") elif pixel_values is not None: pixel_values = pixel_values.to(dtype=self.dtype, device=input_ids.device) # fp16 compatibility batch_size, num_images = pixel_values.size(0), pixel_values.size(1) pixel_values = pixel_values.contiguous().view(batch_size * num_images, *pixel_values.shape[2:]) # Remove padding images - padding images are full 0. real_images_inds = pixel_values.sum(dim=(-1, -2, -3)) != 0.0 pixel_values = pixel_values[real_images_inds] # Get sequence from the vision encoder image_hidden_states = self.vision_model(pixel_values=pixel_values).last_hidden_state # Modality projection image_hidden_states = self.modality_projection(image_hidden_states) if self.config.use_resampler: image_hidden_states = self.perceiver_resampler(image_hidden_states) elif image_hidden_states is not None: image_hidden_states = image_hidden_states.to(dtype=self.dtype, device=input_ids.device) if past_key_values is None: # When we generate, we don't want to replace the potential image_token_id that we generated by images # that simply don't exist new_inp = self.inputs_merger( input_ids=input_ids, inputs_embeds=inputs_embeds, image_hidden_states=image_hidden_states, ) inputs_embeds = new_inp["inputs_embeds"] # Can do add some token types embeddings here (image token vs text token) # something like inputs_embeds += self.token_types(token_types) # embed positions if ( attention_mask is not None and hasattr(self.config, "_flash_attn_2_enabled") and self.config._flash_attn_2_enabled and past_key_values is not None ): is_padding_right = attention_mask[:, -1].sum().item() != batch_size if is_padding_right: raise ValueError( "You are attempting to perform batched generation with padding_side='right'" " this may lead to unexpected behaviour for Flash Attention version of Mistral. Make sure to " " call `tokenizer.padding_side = 'left'` before tokenizing the input. " ) if getattr(self.config, "_flash_attn_2_enabled", False): # 2d mask is passed through the layers attention_mask = attention_mask if (attention_mask is not None and 0 in attention_mask) else None else: # 4d mask is passed through the layers attention_mask = _prepare_4d_causal_attention_mask( attention_mask, (batch_size, seq_length), inputs_embeds, past_key_values_length, sliding_window=self.config.sliding_window, ) attention_mask[attention_mask == -float("inf")] = torch.finfo(self.dtype).min 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 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 if self.gradient_checkpointing and self.training: layer_outputs = self._gradient_checkpointing_func( decoder_layer.__call__, hidden_states, attention_mask, position_ids, past_key_value, output_attentions, use_cache, ) 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, ) 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],) 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, image_hidden_states] if v is not None ) return VMistralBaseModelOutputWithPast( last_hidden_state=hidden_states, past_key_values=next_cache, hidden_states=all_hidden_states, attentions=all_self_attns, image_hidden_states=image_hidden_states, ) class VMistralForVisionText2Text(VMistralPreTrainedModel): _tied_weights_keys = ["lm_head.weight"] def __init__(self, config, vision_model=None): super().__init__(config) self.model = VMistralModel(config, vision_model=vision_model) self.image_token_id = self.config.image_token_id self.lm_head = DecoupledLinear( in_features=config.hidden_size, out_features=config.vocab_size, out_additional_features=config.additional_vocab_size, bias=False, partially_freeze=config.freeze_lm_head, ) # 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 def tie_weights(self): """ Overwrite `transformers.modeling_utils.PreTrainedModel.tie_weights` to handle the case of DecoupledLinear and DecoupledEmbedding. """ output_embeddings = self.get_output_embeddings() input_embeddings = self.get_input_embeddings() if getattr(self.config, "tie_word_embeddings", True): output_embeddings.weight = input_embeddings.weight if input_embeddings.num_additional_embeddings > 0: assert output_embeddings.out_additional_features == input_embeddings.num_additional_embeddings output_embeddings.additional_fc.weight = input_embeddings.additional_embedding.weight if hasattr(output_embeddings, "out_features") and hasattr(input_embeddings, "num_embeddings"): output_embeddings.out_features = input_embeddings.num_embeddings if hasattr(output_embeddings, "out_additional_features") and hasattr( input_embeddings, "num_additional_embeddings" ): output_embeddings.out_additional_features = input_embeddings.num_additional_embeddings @add_start_docstrings_to_model_forward(MISTRAL_INPUTS_DOCSTRING) @replace_return_docstrings(output_type=VMistralCausalLMOutputWithPast, 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, pixel_values: Optional[torch.FloatTensor] = None, image_hidden_states: 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, ) -> Union[Tuple, VMistralCausalLMOutputWithPast]: 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: """ 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 # 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, pixel_values=pixel_values, image_hidden_states=image_hidden_states, use_cache=use_cache, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) hidden_states = outputs[0] logits = self.lm_head(hidden_states) logits = logits.float() loss = None if labels is not None: labels = labels.to(logits.device) # Shift so that tokens < n predict n if attention_mask is not None: shift_attention_mask = attention_mask[..., 1:].to(logits.device) shift_logits = logits[..., :-1, :][shift_attention_mask != 0].contiguous() shift_labels = labels[..., 1:][shift_attention_mask != 0].contiguous() else: shift_logits = logits[..., :-1, :].contiguous() shift_labels = labels[..., 1:].contiguous() # Flatten the tokens loss_fct = CrossEntropyLoss(ignore_index=self.image_token_id) loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1)) if not return_dict: output = (logits,) + outputs[1:] return (loss,) + output if loss is not None else output return VMistralCausalLMOutputWithPast( loss=loss, logits=logits, past_key_values=outputs.past_key_values, hidden_states=outputs.hidden_states, attentions=outputs.attentions, image_hidden_states=outputs.image_hidden_states, ) def prepare_inputs_for_generation(self, input_ids, past=None, **kwargs): image_hidden_states = kwargs.pop("image_hidden_states", None) if image_hidden_states is not None: kwargs["pixel_values"] = None inputs = prepare_inputs_for_generation(input_ids, past=past, **kwargs) unwanted_kwargs = ["token_type_ids"] for kwarg in unwanted_kwargs: inputs.pop(kwarg, None) return inputs @staticmethod def _expand_inputs_for_generation( *args, **model_kwargs, ): return expand_inputs_for_generation(*args, **model_kwargs) @staticmethod def _update_model_kwargs_for_generation(outputs, model_kwargs, is_encoder_decoder): return update_model_kwargs_for_generation(outputs, model_kwargs) @staticmethod def _reorder_cache(past, beam_idx): reordered_past = () for layer_past in past: reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),) return reordered_past