transformers 5.0.0rc0__py3-none-any.whl → 5.0.0rc2__py3-none-any.whl

transformers/models/pe_video/modeling_pe_video.py

@@ -0,0 +1,636 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/pe_video/modular_pe_video.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_pe_video.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+# coding=utf-8
+# Copyright 2025 the HuggingFace Inc. team. All rights reserved.
+#
+# 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.
+from collections.abc import Callable
+from dataclasses import dataclass
+from typing import Any, Optional
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+from ... import initialization as init
+from ...activations import ACT2FN
+from ...cache_utils import Cache
+from ...integrations import use_kernel_forward_from_hub, use_kernelized_func
+from ...modeling_attn_mask_utils import _prepare_4d_attention_mask
+from ...modeling_layers import GradientCheckpointingLayer
+from ...modeling_outputs import BaseModelOutputWithPooling, MaskedLMOutput
+from ...modeling_rope_utils import ROPE_INIT_FUNCTIONS, dynamic_rope_update
+from ...modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from ...processing_utils import Unpack
+from ...utils import ModelOutput, TransformersKwargs, auto_docstring, can_return_tuple
+from ...utils.generic import check_model_inputs, maybe_autocast
+from ..auto import AutoModel, AutoModelForImageClassification
+from .configuration_pe_video import PeVideoConfig, PeVideoEncoderConfig
+
+
+# TODO: not sure about the typing for text_model_output
+@dataclass
+# @auto_docstring
+class PeVideoOutput(ModelOutput):
+    loss: Optional[torch.FloatTensor] = None
+    logits_video_text: Optional[torch.FloatTensor] = None
+    text_video_embeds: Optional[torch.FloatTensor] = None
+    video_embeds: Optional[torch.FloatTensor] = None
+    text_outputs: BaseModelOutputWithPooling = None
+    video_outputs: BaseModelOutputWithPooling = None
+
+    def to_tuple(self) -> tuple[Any]:
+        return tuple(
+            self[k] if k not in ["text_outputs", "video_outputs"] else getattr(self, k).to_tuple() for k in self.keys()
+        )
+
+
+class PeVideoContrastiveHead(nn.Module):
+    def __init__(
+        self,
+        in_dim: int,
+        out_dim: int,
+    ) -> None:
+        super().__init__()
+        self.layer_norm = nn.LayerNorm(normalized_shape=in_dim, eps=1e-6)
+        self.proj = nn.Linear(in_dim, out_dim, bias=False)
+
+    def forward(self, x: torch.Tensor) -> torch.FloatTensor:
+        return self.proj(self.layer_norm(x))
+
+
+class PeVideoMaskedGroupNorm(nn.GroupNorm):
+    def forward(self, x, padding_mask=None):
+        if padding_mask is None:
+            return super().forward(x)
+
+        batch_size, hidden_size, seq_len = x.shape
+        group_size = hidden_size // self.num_groups
+        grouped_shape = (batch_size, -1, group_size, seq_len)
+
+        x_grouped = x.view(grouped_shape)
+        padding_mask_grouped = padding_mask.reshape(grouped_shape).bool()
+
+        mean = torch.masked.mean(x_grouped, mask=padding_mask_grouped, dim=(2, 3), keepdim=True)
+        var = torch.masked.var(x_grouped, mask=padding_mask_grouped, dim=(2, 3), keepdim=True, unbiased=False)
+
+        x_norm = (x_grouped - mean) / torch.sqrt(var + self.eps)
+        x_norm = x_norm.view(x.shape)
+
+        if self.affine:
+            x_norm = x_norm * self.weight.view(1, -1, 1) + self.bias.view(1, -1, 1)
+
+        return x_norm * padding_mask
+
+
+class PeVideoConvBlock1d(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.groupnorm = PeVideoMaskedGroupNorm(num_groups=1, num_channels=config.hidden_size)
+        self.activation = nn.SiLU()
+        self.project = nn.Conv1d(
+            in_channels=config.hidden_size,
+            out_channels=config.hidden_size,
+            kernel_size=3,
+            padding="same",
+        )
+
+    def forward(self, x, padding_mask=None):
+        x = self.groupnorm(x, padding_mask=padding_mask)
+        x = self.activation(x)
+        return self.project(x)
+
+
+class PeVideoResnetBlock1d(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.block1 = PeVideoConvBlock1d(config)
+        self.block2 = PeVideoConvBlock1d(config)
+
+    def forward(self, hidden_states, padding_mask=None):
+        """
+        Args:
+            hidden_states: (batch_size, seq_len, hidden_size)
+            padding_mask: (batch_size, seq_len)
+        Returns:
+            hidden_states: (batch_size, seq_len, hidden_size)
+        """
+        # transpose for convolutions
+        # (batch_size, seq_len, hidden_size) -> (batch_size, hidden_size, seq_len)
+        hidden_states = hidden_states.transpose(1, 2)
+
+        if padding_mask is not None:
+            padding_mask = padding_mask.unsqueeze(1).expand_as(hidden_states)
+
+        residual = hidden_states
+        hidden_states = self.block1(hidden_states, padding_mask=padding_mask)
+        hidden_states = self.block2(hidden_states, padding_mask=padding_mask)
+        hidden_states = residual + hidden_states
+
+        return hidden_states.transpose(1, 2)
+
+
+class PeVideoEncoderPatchEmbedder(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.resnet_block = PeVideoResnetBlock1d(config)
+        self.class_embedding = nn.Parameter(torch.randn(1, 1, config.hidden_size))
+
+    def forward(self, inputs_embeds, padding_mask=None):
+        # Embedding step: prepend class token and run the ResNet block.
+        hidden_states = torch.cat(
+            [self.class_embedding.expand(inputs_embeds.size(0), -1, -1), inputs_embeds],
+            dim=1,
+        )
+
+        if padding_mask is not None:
+            # TODO: any reason why we take padding_mask[0] and not just 1?
+            padding_mask = torch.cat([padding_mask[:, [0]], padding_mask], dim=1)
+
+        hidden_states = self.resnet_block(hidden_states, padding_mask=padding_mask)
+        return hidden_states, padding_mask
+
+
+class PeVideoEncoderEmbedder(nn.Module):
+    def __init__(self, config: PeVideoEncoderConfig):
+        super().__init__()
+        self.vision_model = AutoModelForImageClassification.from_config(config.vision_config)
+        self.proj = nn.Linear(config.vision_config.num_labels, config.hidden_size, bias=False)
+        self.data_proj = nn.Linear(config.hidden_size, config.hidden_size)
+
+    def forward(
+        self,
+        pixel_values_videos: torch.Tensor,
+        padding_mask: Optional[torch.Tensor] = None,
+    ) -> tuple[torch.Tensor, Optional[torch.Tensor]]:
+        input_shape = pixel_values_videos.shape
+
+        pixel_values_videos = pixel_values_videos.view(-1, *input_shape[2:])
+        vision_encoder_outputs = self.vision_model(pixel_values_videos)
+
+        logits = vision_encoder_outputs.logits.view(*input_shape[:2], -1)
+        logits = F.normalize(logits, dim=-1)
+
+        vision_features = self.proj(logits)
+        inputs_embeds = self.data_proj(vision_features)
+
+        return inputs_embeds, padding_mask
+
+
+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)
+
+
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    **kwargs: Unpack[TransformersKwargs],
+):
+    key_states = repeat_kv(key, module.num_key_value_groups)
+    value_states = repeat_kv(value, module.num_key_value_groups)
+
+    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
+    if attention_mask is not None:
+        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
+    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
+    attn_output = torch.matmul(attn_weights, value_states)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    return attn_output, attn_weights
+
+
+def stack_freqs(cos: torch.Tensor, sin: torch.Tensor):
+    dim = cos.size(-1)
+    cos = cos.narrow(-1, 0, dim // 2)
+    sin = sin.narrow(-1, 0, dim // 2)
+    freqs_cis = torch.stack((cos, -sin, sin, cos), dim=-1).view(*cos.size(), 2, 2)
+    return freqs_cis
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, unsqueeze_dim=1):
+    freqs_cis = stack_freqs(cos, sin)
+    freqs_cis = freqs_cis.unsqueeze(unsqueeze_dim)
+    q_ = q.reshape(*q.shape[:-1], -1, 1, 2)
+    k_ = k.reshape(*k.shape[:-1], -1, 1, 2)
+    return (q_ * freqs_cis).sum(5).flatten(3), (k_ * freqs_cis).sum(5).flatten(3)
+
+
+@use_kernel_forward_from_hub("RMSNorm")
+class PeVideoEncoderRMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps: float = 1e-6) -> None:
+        """
+        PeVideoEncoderRMSNorm is equivalent to T5LayerNorm
+        """
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        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)
+
+    def extra_repr(self):
+        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
+
+
+@use_kernelized_func(apply_rotary_pos_emb)
+class PeVideoEncoderAttention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config, layer_idx):
+        super().__init__()
+        self.layer_type = config.layer_types[layer_idx] if hasattr(config, "layer_types") else None
+        self.config = config
+        self.layer_idx = layer_idx
+        self.head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.scaling = self.head_dim**-0.5
+        self.attention_dropout = config.attention_dropout
+        self.is_causal = False
+
+        self.q_proj = nn.Linear(
+            config.hidden_size, config.num_attention_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.k_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.v_proj = nn.Linear(
+            config.hidden_size, config.num_key_value_heads * self.head_dim, bias=config.attention_bias
+        )
+        self.o_proj = nn.Linear(
+            config.num_attention_heads * self.head_dim, config.hidden_size, bias=config.attention_bias
+        )
+        self.q_norm = PeVideoEncoderRMSNorm(
+            self.head_dim, eps=config.rms_norm_eps
+        )  # unlike olmo, only on the head dim!
+        self.k_norm = PeVideoEncoderRMSNorm(
+            self.head_dim, eps=config.rms_norm_eps
+        )  # thus post q_norm does not need reshape
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> tuple[torch.Tensor, torch.Tensor]:
+        input_shape = hidden_states.shape[:-1]
+        hidden_shape = (*input_shape, -1, self.head_dim)
+
+        query_states = self.q_norm(self.q_proj(hidden_states).view(hidden_shape)).transpose(1, 2)
+        key_states = self.k_norm(self.k_proj(hidden_states).view(hidden_shape)).transpose(1, 2)
+        value_states = self.v_proj(hidden_states).view(hidden_shape).transpose(1, 2)
+
+        cos, sin = position_embeddings
+        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            **kwargs,
+        )
+
+        attn_output = attn_output.reshape(*input_shape, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class PeVideoEncoderMLP(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):
+        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+        return down_proj
+
+
+class PeVideoEncoderLayer(GradientCheckpointingLayer):
+    def __init__(self, config, layer_idx):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = PeVideoEncoderAttention(config=config, layer_idx=layer_idx)
+
+        self.mlp = PeVideoEncoderMLP(config)
+        self.input_layernorm = PeVideoEncoderRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = PeVideoEncoderRMSNorm(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_values: Optional[Cache] = None,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None,
+        **kwargs: Unpack[TransformersKwargs],
+    ) -> torch.Tensor:
+        residual = hidden_states
+        hidden_states = self.input_layernorm(hidden_states)
+        # Self Attention
+        hidden_states, _ = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+        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
+        return hidden_states
+
+
+@auto_docstring
+class PeVideoPreTrainedModel(PreTrainedModel):
+    config: PeVideoConfig
+    base_model_prefix = "video_model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["PeVideoEncoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+
+    _can_compile_fullgraph = True
+    _supports_attention_backend = True
+    _can_record_outputs = {
+        "hidden_states": PeVideoEncoderLayer,
+        "attentions": PeVideoEncoderAttention,
+    }
+    main_input_name = "pixel_values_videos"
+
+    def _init_weights(self, module):
+        super()._init_weights(module)
+
+        if hasattr(self.config, "initializer_range"):
+            std = self.config.initializer_range
+        else:
+            # 0.02 is the standard default value across the library
+            std = getattr(self.config.get_text_config(), "initializer_range", 0.02)
+
+        if isinstance(module, PeVideoEncoderPatchEmbedder):
+            embed_dim = module.class_embedding.shape[-1]
+            init.normal_(module.class_embedding, mean=0.0, std=embed_dim**-0.5 * std)
+
+
+class PeVideoEncoderRotaryEmbedding(nn.Module):
+    inv_freq: torch.Tensor  # fix linting for `register_buffer`
+
+    def __init__(self, config: PeVideoEncoderConfig, device=None):
+        super().__init__()
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        self.config = config
+
+        self.rope_type = self.config.rope_parameters["rope_type"]
+        rope_init_fn: Callable = self.compute_default_rope_parameters
+        if self.rope_type != "default":
+            rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+        inv_freq, self.attention_scaling = rope_init_fn(self.config, device)
+
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.register_buffer("original_inv_freq", inv_freq.clone(), persistent=False)
+
+    @staticmethod
+    def compute_default_rope_parameters(
+        config: Optional[PeVideoEncoderConfig] = None,
+        device: Optional["torch.device"] = None,
+        seq_len: Optional[int] = None,
+    ) -> tuple["torch.Tensor", float]:
+        """
+        Computes the inverse frequencies according to the original RoPE implementation
+        Args:
+            config ([`~transformers.PreTrainedConfig`]):
+                The model configuration.
+            device (`torch.device`):
+                The device to use for initialization of the inverse frequencies.
+            seq_len (`int`, *optional*):
+                The current sequence length. Unused for this type of RoPE.
+        Returns:
+            Tuple of (`torch.Tensor`, `float`), containing the inverse frequencies for the RoPE embeddings and the
+            post-processing scaling factor applied to the computed cos/sin (unused in this type of RoPE).
+        """
+        base = config.rope_parameters["rope_theta"]
+        dim = getattr(config, "head_dim", None) or config.hidden_size // config.num_attention_heads
+
+        attention_factor = 1.0  # Unused in this type of RoPE
+
+        # Compute the inverse frequencies
+        inv_freq = 1.0 / (
+            base ** (torch.arange(0, dim, 2, dtype=torch.int64).to(device=device, dtype=torch.float) / dim)
+        )
+        return inv_freq, attention_factor
+
+    @torch.no_grad()
+    @dynamic_rope_update  # power user: used with advanced RoPE types (e.g. dynamic rope)
+    def forward(self, x, position_ids):
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1).to(x.device)
+        position_ids_expanded = position_ids[:, None, :].float()
+
+        device_type = x.device.type if isinstance(x.device.type, str) and x.device.type != "mps" else "cpu"
+        with maybe_autocast(device_type=device_type, enabled=False):  # Force float32
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos() * self.attention_scaling
+            sin = emb.sin() * self.attention_scaling
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+@auto_docstring(
+    custom_intro="""
+    The PeVideo Encoder model.
+    """
+)
+class PeVideoEncoder(PeVideoPreTrainedModel):
+    config: PeVideoEncoderConfig
+    main_input_name = "pixel_values_videos"
+    base_model_prefix = "video_model.video_encoder"
+
+    def __init__(self, config: PeVideoEncoderConfig):
+        super().__init__(config)
+        self.embedder = PeVideoEncoderEmbedder(config)
+        self.patch_embedder = PeVideoEncoderPatchEmbedder(config)
+        self.layers = nn.ModuleList(
+            [PeVideoEncoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = PeVideoEncoderRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = PeVideoEncoderRotaryEmbedding(config=config)
+        self.output = nn.Linear(config.hidden_size, config.hidden_size, bias=False)
+        self.gradient_checkpointing = False
+
+        self.post_init()
+
+    @can_return_tuple
+    @check_model_inputs
+    def forward(
+        self,
+        pixel_values_videos: torch.Tensor,
+        padding_mask_videos: Optional[torch.Tensor] = None,
+        **kwargs,
+    ) -> BaseModelOutputWithPooling:
+        inputs_embeds, padding_mask = self.embedder(pixel_values_videos, padding_mask=padding_mask_videos)
+        inputs_embeds, attention_mask = self.patch_embedder(inputs_embeds, padding_mask=padding_mask)
+
+        if attention_mask is not None:
+            attention_mask = _prepare_4d_attention_mask(attention_mask, inputs_embeds.dtype)
+
+        position_ids = torch.arange(inputs_embeds.shape[1], device=inputs_embeds.device).unsqueeze(0)
+        position_embeddings = self.rotary_emb(inputs_embeds, position_ids)
+
+        hidden_states = inputs_embeds
+        for encoder_layer in self.layers[: self.config.num_hidden_layers]:
+            hidden_states = encoder_layer(
+                hidden_states,
+                attention_mask=attention_mask,
+                position_embeddings=position_embeddings,
+                **kwargs,
+            )
+
+        hidden_states = self.norm(hidden_states)
+        hidden_states = self.output(hidden_states)
+
+        return BaseModelOutputWithPooling(
+            last_hidden_state=hidden_states[:, 1:],
+            pooler_output=hidden_states[:, 0],
+        )
+
+
+class PeVideoModel(PeVideoPreTrainedModel):
+    main_input_name = "input_ids"
+
+    def __init__(self, config: PeVideoConfig):
+        super().__init__(config)
+        self.text_model = AutoModel.from_config(config.text_config)
+        self.video_encoder = PeVideoEncoder(config.video_config)
+
+        self.text_video_head = PeVideoContrastiveHead(config.text_config.hidden_size, config.text_config.hidden_size)
+        self.video_head = PeVideoContrastiveHead(config.video_config.hidden_size, config.text_config.hidden_size)
+
+        self.text_video_logit_scale = nn.Parameter(torch.zeros(1))
+        self.text_video_logit_bias = nn.Parameter(torch.zeros(1))
+
+        self.post_init()
+
+    def get_text_features(self, input_ids, attention_mask=None):
+        # TODO: should it be named feature or embeds
+        text_outputs: MaskedLMOutput = self.text_model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            return_dict=True,
+        )
+
+        text_features = text_outputs.last_hidden_state
+        text_features = self.text_video_head(text_features)
+        return text_features
+
+    def get_video_features(self, pixel_values_videos, padding_mask_videos=None):
+        # TODO: should it be named feature or embeds
+        video_outputs: BaseModelOutputWithPooling = self.video_encoder(
+            pixel_values_videos=pixel_values_videos,
+            padding_mask_videos=padding_mask_videos,
+            return_dict=True,
+        )
+        video_features = self.video_head(video_outputs.pooler_output)
+        return video_features
+
+    @can_return_tuple
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        pixel_values_videos: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        padding_mask_videos: Optional[torch.Tensor] = None,
+        return_loss: Optional[bool] = None,
+        **kwargs,
+    ) -> PeVideoOutput:
+        video_outputs: BaseModelOutputWithPooling = self.video_encoder(
+            pixel_values_videos=pixel_values_videos, padding_mask_videos=padding_mask_videos, **kwargs
+        )
+        kwargs["output_hidden_states"] = True
+        text_outputs: MaskedLMOutput = self.text_model(input_ids=input_ids, attention_mask=attention_mask, **kwargs)
+
+        video_embeds = video_outputs.pooler_output
+        video_embeds = self.video_head(video_embeds)
+
+        text_video_embeds = text_outputs.hidden_states[-1][:, 0]
+        text_video_embeds = self.text_video_head(text_video_embeds)
+
+        logits_video_text = video_embeds @ text_video_embeds.T
+        logits_video_text = logits_video_text * self.text_video_logit_scale + self.text_video_logit_bias
+
+        loss = None
+        if return_loss:
+            labels = torch.eye(logits_video_text.shape[0], device=logits_video_text.device)
+            loss = -F.logsigmoid(labels * logits_video_text).sum() / logits_video_text.shape[0]
+
+        return PeVideoOutput(
+            logits_video_text=logits_video_text,
+            text_video_embeds=text_video_embeds,
+            video_embeds=video_embeds,
+            text_outputs=text_outputs,
+            video_outputs=video_outputs,
+            loss=loss,
+        )
+
+
+__all__ = ["PeVideoEncoder", "PeVideoModel"]