From c98090420431e49d5cab8f41b1cb5426a8b87e5a Mon Sep 17 00:00:00 2001 From: Vladislav Bronzov <58587565+VladOS95-cyber@users.noreply.github.com> Date: Wed, 9 Jul 2025 17:04:28 +0200 Subject: [PATCH] Add DeepSeek V2 Model into Transformers (#36400) * add initial structure * doc fixes, add model base logic * update init files * some fixes to config and modular * some improvements for attention * format * remove unused attn * some fixes for moe layer and for decoder * adapt _compute_yarn_parameters for deepseek * format * small fix * fix for decoder forward * add tests, small refactoring * fix dummies * fix init * fix doc * fix config docs * add sequce doc, fix init for gate * fix issues in tests * fix config doc * remove unused args * some fixes and refactoring after review * fix doc for config * small fixes for config args * revert config refactoring * small refactoring * minor fixes after rebase * small fix after merge * fix modular * remove rotaryembd from public init * small test fix * some rotary pos calculation improvement * fix format * some improvements and fixes * fix config * some refactoring * adjust some unit tests * skip test * small fixes and tests adjustment * reapply modular * fix all tests except Integration * fix integration testzs * cleanup BC stuff * rope * fix integrations tests based on a10 * style --------- Co-authored-by: Cyril Vallez Co-authored-by: Cyril Vallez --- docs/source/en/_toctree.yml | 2 + docs/source/en/model_doc/deepseek_v2.md | 49 ++ src/transformers/models/__init__.py | 1 + .../models/auto/configuration_auto.py | 2 + src/transformers/models/auto/modeling_auto.py | 3 + .../models/auto/tokenization_auto.py | 7 + .../models/deepseek_v2/__init__.py | 29 + .../deepseek_v2/configuration_deepseek_v2.py | 239 ++++++ .../deepseek_v2/modeling_deepseek_v2.py | 773 ++++++++++++++++++ .../models/deepseek_v2/modular_deepseek_v2.py | 539 ++++++++++++ tests/models/deepseek_v2/__init__.py | 0 .../deepseek_v2/test_modeling_deepseek_v2.py | 269 ++++++ 12 files changed, 1913 insertions(+) create mode 100644 docs/source/en/model_doc/deepseek_v2.md create mode 100644 src/transformers/models/deepseek_v2/__init__.py create mode 100644 src/transformers/models/deepseek_v2/configuration_deepseek_v2.py create mode 100644 src/transformers/models/deepseek_v2/modeling_deepseek_v2.py create mode 100644 src/transformers/models/deepseek_v2/modular_deepseek_v2.py create mode 100644 tests/models/deepseek_v2/__init__.py create mode 100644 tests/models/deepseek_v2/test_modeling_deepseek_v2.py diff --git a/docs/source/en/_toctree.yml b/docs/source/en/_toctree.yml index 2675a72223..3e2280ca91 100644 --- a/docs/source/en/_toctree.yml +++ b/docs/source/en/_toctree.yml @@ -709,6 +709,8 @@ title: D-FINE - local: model_doc/dab-detr title: DAB-DETR + - local: model_doc/deepseek_v2 + title: DeepSeek-V2 - local: model_doc/deformable_detr title: Deformable DETR - local: model_doc/deit diff --git a/docs/source/en/model_doc/deepseek_v2.md b/docs/source/en/model_doc/deepseek_v2.md new file mode 100644 index 0000000000..ed4876bd67 --- /dev/null +++ b/docs/source/en/model_doc/deepseek_v2.md @@ -0,0 +1,49 @@ + + +# DeepSeek-V2 + +## Overview + +The DeepSeek-V2 model was proposed in [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://arxiv.org/abs/2405.04434) by DeepSeek-AI Team. + +The abstract from the paper is the following: +We present DeepSeek-V2, a strong Mixture-of-Experts (MoE) language model characterized by economical training and efficient inference. It comprises 236B total parameters, of which 21B are activated for each token, and supports a context length of 128K tokens. DeepSeek-V2 adopts innovative architectures including Multi-head Latent Attention (MLA) and DeepSeekMoE. MLA guarantees efficient inference through significantly compressing the Key-Value (KV) cache into a latent vector, while DeepSeekMoE enables training strong models at an economical cost through sparse computation. Compared with DeepSeek 67B, DeepSeek-V2 achieves significantly stronger performance, and meanwhile saves 42.5% of training costs, reduces the KV cache by 93.3%, and boosts the maximum generation throughput to 5.76 times. We pretrain DeepSeek-V2 on a high-quality and multi-source corpus consisting of 8.1T tokens, and further perform Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) to fully unlock its potential. Evaluation results show that, even with only 21B activated parameters, DeepSeek-V2 and its chat versions still achieve top-tier performance among open-source models. + +This model was contributed by [VladOS95-cyber](https://github.com/VladOS95-cyber). +The original code can be found [here](https://huggingface.co/deepseek-ai/DeepSeek-V2). + +### Usage tips +The model uses Multi-head Latent Attention (MLA) and DeepSeekMoE architectures for efficient inference and cost-effective training. It employs an auxiliary-loss-free strategy for load balancing and multi-token prediction training objective. The model can be used for various language tasks after being pre-trained on 14.8 trillion tokens and going through Supervised Fine-Tuning and Reinforcement Learning stages. + +## DeepseekV2Config + +[[autodoc]] DeepseekV2Config + +## DeepseekV2Model + +[[autodoc]] DeepseekV2Model + - forward + +## DeepseekV2ForCausalLM + +[[autodoc]] DeepseekV2ForCausalLM + - forward + +## DeepseekV2ForSequenceClassification + +[[autodoc]] DeepseekV2ForSequenceClassification + - forward \ No newline at end of file diff --git a/src/transformers/models/__init__.py b/src/transformers/models/__init__.py index 98b5a002e6..b0913030e4 100644 --- a/src/transformers/models/__init__.py +++ b/src/transformers/models/__init__.py @@ -82,6 +82,7 @@ if TYPE_CHECKING: from .deberta import * from .deberta_v2 import * from .decision_transformer import * + from .deepseek_v2 import * from .deepseek_v3 import * from .deformable_detr import * from .deit import * diff --git a/src/transformers/models/auto/configuration_auto.py b/src/transformers/models/auto/configuration_auto.py index a3d4cf196f..d097206c12 100644 --- a/src/transformers/models/auto/configuration_auto.py +++ b/src/transformers/models/auto/configuration_auto.py @@ -101,6 +101,7 @@ CONFIG_MAPPING_NAMES = OrderedDict[str, str]( ("deberta", "DebertaConfig"), ("deberta-v2", "DebertaV2Config"), ("decision_transformer", "DecisionTransformerConfig"), + ("deepseek_v2", "DeepseekV2Config"), ("deepseek_v3", "DeepseekV3Config"), ("deformable_detr", "DeformableDetrConfig"), ("deit", "DeiTConfig"), @@ -481,6 +482,7 @@ MODEL_NAMES_MAPPING = OrderedDict[str, str]( ("deberta", "DeBERTa"), ("deberta-v2", "DeBERTa-v2"), ("decision_transformer", "Decision Transformer"), + ("deepseek_v2", "DeepSeek-V2"), ("deepseek_v3", "DeepSeek-V3"), ("deformable_detr", "Deformable DETR"), ("deit", "DeiT"), diff --git a/src/transformers/models/auto/modeling_auto.py b/src/transformers/models/auto/modeling_auto.py index d59e249740..eb46b2d0e4 100644 --- a/src/transformers/models/auto/modeling_auto.py +++ b/src/transformers/models/auto/modeling_auto.py @@ -95,6 +95,7 @@ MODEL_MAPPING_NAMES = OrderedDict( ("deberta", "DebertaModel"), ("deberta-v2", "DebertaV2Model"), ("decision_transformer", "DecisionTransformerModel"), + ("deepseek_v2", "DeepseekV2Model"), ("deepseek_v3", "DeepseekV3Model"), ("deformable_detr", "DeformableDetrModel"), ("deit", "DeiTModel"), @@ -577,6 +578,7 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict( ("ctrl", "CTRLLMHeadModel"), ("data2vec-text", "Data2VecTextForCausalLM"), ("dbrx", "DbrxForCausalLM"), + ("deepseek_v2", "DeepseekV2ForCausalLM"), ("deepseek_v3", "DeepseekV3ForCausalLM"), ("diffllama", "DiffLlamaForCausalLM"), ("doge", "DogeForCausalLM"), @@ -1108,6 +1110,7 @@ MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES = OrderedDict( ("data2vec-text", "Data2VecTextForSequenceClassification"), ("deberta", "DebertaForSequenceClassification"), ("deberta-v2", "DebertaV2ForSequenceClassification"), + ("deepseek_v2", "DeepseekV2ForSequenceClassification"), ("diffllama", "DiffLlamaForSequenceClassification"), ("distilbert", "DistilBertForSequenceClassification"), ("doge", "DogeForSequenceClassification"), diff --git a/src/transformers/models/auto/tokenization_auto.py b/src/transformers/models/auto/tokenization_auto.py index dc1909bbb7..7c3a8b7c53 100644 --- a/src/transformers/models/auto/tokenization_auto.py +++ b/src/transformers/models/auto/tokenization_auto.py @@ -177,6 +177,13 @@ TOKENIZER_MAPPING_NAMES = OrderedDict[str, tuple[Optional[str], Optional[str]]]( "DebertaV2TokenizerFast" if is_tokenizers_available() else None, ), ), + ( + "deepseek_v2", + ( + "LlamaTokenizer" if is_sentencepiece_available() else None, + "LlamaTokenizerFast" if is_tokenizers_available() else None, + ), + ), ( "deepseek_v3", ( diff --git a/src/transformers/models/deepseek_v2/__init__.py b/src/transformers/models/deepseek_v2/__init__.py new file mode 100644 index 0000000000..c9aaf5f0fc --- /dev/null +++ b/src/transformers/models/deepseek_v2/__init__.py @@ -0,0 +1,29 @@ +# Copyright 2025 The HuggingFace 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 typing import TYPE_CHECKING + +from ...utils import _LazyModule +from ...utils.import_utils import define_import_structure + + +if TYPE_CHECKING: + from .configuration_deepseek_v2 import * + from .modeling_deepseek_v2 import * +else: + import sys + + _file = globals()["__file__"] + sys.modules[__name__] = _LazyModule(__name__, _file, define_import_structure(_file), module_spec=__spec__) diff --git a/src/transformers/models/deepseek_v2/configuration_deepseek_v2.py b/src/transformers/models/deepseek_v2/configuration_deepseek_v2.py new file mode 100644 index 0000000000..b0b25625ea --- /dev/null +++ b/src/transformers/models/deepseek_v2/configuration_deepseek_v2.py @@ -0,0 +1,239 @@ +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# This file was automatically generated from src/transformers/models/deepseek_v2/modular_deepseek_v2.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_deepseek_v2.py file directly. One of our CI enforces this. +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# coding=utf-8 +# Copyright 2025 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 ...configuration_utils import PretrainedConfig +from ...modeling_rope_utils import rope_config_validation + + +class DeepseekV2Config(PretrainedConfig): + r""" + This is the configuration class to store the configuration of a [`DeepseekV2Model`]. It is used to instantiate a DeepSeek + model according to the specified arguments, defining the model architecture. Instantiating a configuration with the + defaults will yield a similar configuration to that of DeepSeek-V2-Lite" [deepseek-ai/DeepSeek-V2-Lite"](https://huggingface.co/deepseek-ai/DeepSeek-V2-Lite"). + Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the + documentation from [`PretrainedConfig`] for more information. + + Args: + vocab_size (`int`, *optional*, defaults to 32000): + Vocabulary size of the DeepSeek model. Defines the number of different tokens that can be represented by the + `input_ids` passed when calling [`DeepseekV2Model`]. + hidden_size (`int`, *optional*, defaults to 4096): + Dimension of the hidden representations. + intermediate_size (`int`, *optional*, defaults to 11008): + Dimension of the MLP representations. + num_hidden_layers (`int`, *optional*, defaults to 32): + Number of hidden layers in the Transformer decoder. + num_attention_heads (`int`, *optional*, defaults to 32): + Number of attention heads for each attention layer in the Transformer decoder. + num_key_value_heads (`int`, *optional*): + The number of key-value heads used to implement Grouped Query Attention (GQA). If + `num_key_value_heads=num_attention_heads`, the model will use Multi-Head Attention (MHA). If + `num_key_value_heads=1`, the model will use Multi-Query Attention (MQA). Otherwise, GQA is used. + hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): + The non-linear activation function (function or string) in the decoder. + max_position_embeddings (`int`, *optional*, defaults to 2048): + The maximum sequence length that this model might ever be used with. + initializer_range (`float`, *optional*, defaults to 0.02): + The standard deviation of the truncated normal initializer for initializing all weight matrices. + rms_norm_eps (`float`, *optional*, defaults to 1e-06): + The epsilon value used by the RMS normalization layers. + use_cache (`bool`, *optional*, defaults to `True`): + Whether or not the model should return the last key/value attentions (useful for inference optimization). + pad_token_id (`int`, *optional*): + Padding token ID. + bos_token_id (`int`, *optional*, defaults to 1): + Beginning-of-sequence token ID. + eos_token_id (`int`, *optional*, defaults to 2): + End-of-sequence token ID. + tie_word_embeddings (`bool`, *optional*, defaults to `False`): + Whether to tie input and output embeddings. + rope_theta (`float`, *optional*, defaults to 10000.0): + The base period of the Rotary Position Embeddings (RoPE). + rope_scaling (`Dict`, *optional*): + Configuration for scaling RoPE embeddings. Supports `linear` and `dynamic` scaling strategies. + attention_bias (`bool`, *optional*, defaults to `False`): + Whether to use a bias in the query, key, value, and output projection layers during self-attention. + attention_dropout (`float`, *optional*, defaults to 0.0): + The dropout probability applied to attention weights. + mlp_bias (`bool`, *optional*, defaults to `False`): + Whether to use a bias term in the MLP layers. + aux_loss_alpha (`float`, *optional*, defaults to 0.001): + Weight coefficient for auxiliary loss in Mixture of Experts (MoE) models. + first_k_dense_replace (`int`, *optional*, defaults to 0): + Number of dense layers in the shallow layers before switching to MoE layers. + kv_lora_rank (`int`, *optional*, defaults to 512): + Rank of the LoRA decomposition for key-value projections. + q_lora_rank (`int`, *optional*, defaults to 1536): + Rank of the LoRA decomposition for query projections. + Specifically, it determines the dimensionality to which the query (q) vectors are compressed before being expanded back to their original size. + It reduces computational overhead while maintaining model performance. + n_group (`int`, *optional*): + Number of groups for routed experts. + n_routed_experts (`int`, *optional*, defaults to 64): + Number of routed experts (None indicates a dense model). + n_shared_experts (`int`, *optional*, defaults to 2): + Number of shared experts (None indicates a dense model). + qk_nope_head_dim (`int`, *optional*, defaults to 128): + The head dimension for the QK (query-key) projections when using NOPE (Neural Operator Position Encoding). + qk_rope_head_dim (`int`, *optional*, defaults to 64): + The head dimension for QK projections when using RoPE. + routed_scaling_factor (`float`, *optional*, defaults to 1.0): + Scaling factor for routed experts in MoE models. + seq_aux (`bool`, *optional*, defaults to `True`): + Whether to compute the auxiliary loss for each individual sequence. + topk_group (`int`, *optional*): + Number of selected groups per token for expert selection. + topk_method (`str`, *optional*, defaults to `"greedy"`): + The method used for selecting top-k experts in the routed gate mechanism. + v_head_dim (`int`, *optional*, defaults to 128): + The dimension of value projections in the attention layers. + num_experts_per_tok (`int`, *optional*): + The number of experts selected per token. If `None`, the model behaves as a dense Transformer. + norm_topk_prob (`bool`, *optional*, defaults to `False`): + Whether to normalize the probability distribution over top-k selected experts. + moe_intermediate_size (`int`, *optional*, defaults to 1407): + Dimension of the MoE (Mixture of Experts) representations. + + ```python + >>> from transformers import DeepseekV2Model, DeepseekV2Config + >>> # Initializing a DeepSeek-V2 style configuration + >>> configuration = DeepseekV2Config() + >>> # Accessing the model configuration + >>> model = DeepseekV2Model(configuration) + >>> print(model.config) + ``` + """ + + model_type = "deepseek_v2" + keys_to_ignore_at_inference = ["past_key_values"] + + base_model_tp_plan = { + "layers.*.self_attn.q_proj": "colwise", + "layers.*.self_attn.q_a_proj": "colwise", + "layers.*.self_attn.q_b_proj": "colwise", + "layers.*.self_attn.kv_b_proj": "colwise", + "layers.*.self_attn.o_proj": "rowwise", + "layers.*.mlp.gate_proj": "colwise", + "layers.*.mlp.up_proj": "colwise", + "layers.*.mlp.down_proj": "rowwise", + } + base_model_pp_plan = { + "embed_tokens": (["input_ids"], ["inputs_embeds"]), + "layers": (["hidden_states", "attention_mask"], ["hidden_states"]), + "norm": (["hidden_states"], ["hidden_states"]), + } + + def __init__( + self, + vocab_size=32000, + hidden_size=4096, + intermediate_size=11008, + num_hidden_layers=32, + num_attention_heads=32, + num_key_value_heads=None, + hidden_act="silu", + max_position_embeddings=2048, + initializer_range=0.02, + rms_norm_eps=1e-6, + use_cache=True, + pad_token_id=None, + bos_token_id=1, + eos_token_id=2, + tie_word_embeddings=False, + rope_theta=10000.0, + rope_scaling=None, + attention_bias=False, + attention_dropout=0.0, + mlp_bias=False, + aux_loss_alpha=0.001, + first_k_dense_replace=0, + kv_lora_rank=512, + q_lora_rank=1536, + n_group=None, + n_routed_experts=64, + n_shared_experts=2, + qk_nope_head_dim=128, + qk_rope_head_dim=64, + routed_scaling_factor=1.0, + seq_aux=True, + topk_group=None, + topk_method="greedy", + v_head_dim=128, + num_experts_per_tok=None, + norm_topk_prob=False, + moe_intermediate_size=1407, + **kwargs, + ): + super().__init__( + pad_token_id=pad_token_id, + bos_token_id=bos_token_id, + eos_token_id=eos_token_id, + tie_word_embeddings=tie_word_embeddings, + **kwargs, + ) + self.vocab_size = vocab_size + self.max_position_embeddings = max_position_embeddings + self.hidden_size = hidden_size + self.intermediate_size = intermediate_size + self.num_hidden_layers = num_hidden_layers + self.num_attention_heads = num_attention_heads + + # for backward compatibility + if num_key_value_heads is None: + num_key_value_heads = num_attention_heads + + self.num_key_value_heads = num_key_value_heads + self.hidden_act = hidden_act + self.initializer_range = initializer_range + self.rms_norm_eps = rms_norm_eps + self.use_cache = use_cache + self.rope_theta = rope_theta + self.rope_scaling = rope_scaling + self.attention_bias = attention_bias + self.attention_dropout = attention_dropout + self.mlp_bias = mlp_bias + self.head_dim = qk_rope_head_dim + # Validate the correctness of rotary position embeddings parameters + # BC: if there is a 'type' field, copy it it to 'rope_type'. + if self.rope_scaling is not None and "type" in self.rope_scaling: + self.rope_scaling["rope_type"] = self.rope_scaling["type"] + rope_config_validation(self) + self.aux_loss_alpha = aux_loss_alpha + self.first_k_dense_replace = first_k_dense_replace + self.kv_lora_rank = kv_lora_rank + self.q_lora_rank = q_lora_rank + self.n_group = n_group + self.n_routed_experts = n_routed_experts + self.n_shared_experts = n_shared_experts + self.qk_nope_head_dim = qk_nope_head_dim + self.qk_rope_head_dim = qk_rope_head_dim + self.routed_scaling_factor = routed_scaling_factor + self.seq_aux = seq_aux + self.topk_group = topk_group + self.topk_method = topk_method + self.v_head_dim = v_head_dim + self.num_experts_per_tok = num_experts_per_tok + self.norm_topk_prob = norm_topk_prob + self.moe_intermediate_size = moe_intermediate_size + + +__all__ = ["DeepseekV2Config"] diff --git a/src/transformers/models/deepseek_v2/modeling_deepseek_v2.py b/src/transformers/models/deepseek_v2/modeling_deepseek_v2.py new file mode 100644 index 0000000000..46716316aa --- /dev/null +++ b/src/transformers/models/deepseek_v2/modeling_deepseek_v2.py @@ -0,0 +1,773 @@ +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# This file was automatically generated from src/transformers/models/deepseek_v2/modular_deepseek_v2.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_deepseek_v2.py file directly. One of our CI enforces this. +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# coding=utf-8 +# Copyright 2025 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. + +import warnings +from typing import Callable, Optional, Union + +import torch +import torch.nn.functional as F +from torch import nn + +from ...activations import ACT2FN +from ...cache_utils import Cache, DynamicCache +from ...generation import GenerationMixin +from ...integrations import use_kernel_forward_from_hub +from ...masking_utils import create_causal_mask +from ...modeling_layers import GradientCheckpointingLayer +from ...modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast +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 TransformersKwargs, auto_docstring, can_return_tuple, logging +from ...utils.generic import check_model_inputs +from .configuration_deepseek_v2 import DeepseekV2Config + + +logger = logging.get_logger(__name__) + + +class DeepseekV2MoEGate(nn.Module): + def __init__(self, config: DeepseekV2Config): + super().__init__() + self.config = config + self.top_k = config.num_experts_per_tok + self.num_experts = config.n_routed_experts + self.routed_scaling_factor = config.routed_scaling_factor + self.alpha = config.aux_loss_alpha + self.seq_aux = config.seq_aux + self.topk_method = config.topk_method + self.num_group = config.n_group + self.topk_group = config.topk_group + + # topk selection algorithm + self.norm_topk_prob = config.norm_topk_prob + self.gating_dim = config.hidden_size + self.weight = nn.Parameter(torch.empty((self.num_experts, self.gating_dim))) + + def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: + batch_size, seq_len, hidden_dim = hidden_states.shape + ### compute gating score + hidden_states = hidden_states.view(-1, hidden_dim) + logits = F.linear(hidden_states.type(torch.float32), self.weight.type(torch.float32), None) + scores = logits.softmax(dim=-1, dtype=torch.float32) + + # select top-k experts + # greedy method is used for DeepSeek-V2-Lite + # group_limited_greedy for DeepSeek-V2 and DeepSeek-V2-Chat + if self.topk_method == "greedy": + topk_weight, topk_idx = torch.topk(scores, k=self.top_k, dim=-1, sorted=False) + elif self.topk_method == "group_limited_greedy": + group_scores = scores.view(batch_size * seq_len, self.num_group, -1).max(dim=-1).values # [n, num_group] + group_idx = torch.topk(group_scores, k=self.topk_group, dim=-1, sorted=False)[1] # [n, top_k_group] + group_mask = torch.zeros_like(group_scores) # [n, num_group] + group_mask.scatter_(1, group_idx, 1) # [n, num_group] + score_mask = ( + group_mask.unsqueeze(-1) + .expand(batch_size * seq_len, self.num_group, self.num_experts // self.num_group) + .reshape(batch_size * seq_len, -1) + ) # [n, e] + tmp_scores = scores.masked_fill(~score_mask.bool(), 0.0) # [n, e] + topk_weight, topk_idx = torch.topk(tmp_scores, k=self.top_k, dim=-1, sorted=False) + + topk_weight = topk_weight * self.routed_scaling_factor + ### expert-level computation auxiliary loss + return topk_idx, topk_weight + + +class DeepseekV2MoE(nn.Module): + """ + A mixed expert module containing shared experts. + """ + + def __init__(self, config: DeepseekV2Config): + super().__init__() + self.config = config + self.num_experts_per_tok = config.num_experts_per_tok + + self.experts = nn.ModuleList( + [ + (DeepseekV2MLP(config, intermediate_size=config.moe_intermediate_size)) + for _ in range(config.n_routed_experts) + ] + ) + self.gate = DeepseekV2MoEGate(config) + if config.n_shared_experts is not None: + intermediate_size = config.moe_intermediate_size * config.n_shared_experts + self.shared_experts = DeepseekV2MLP(config=config, intermediate_size=intermediate_size) + self.ep_rank = 0 + self.experts_per_rank = config.n_routed_experts + + def moe(self, hidden_states: torch.Tensor, topk_ids: torch.Tensor, topk_weight: torch.Tensor) -> torch.Tensor: + cnts = topk_ids.new_zeros((topk_ids.shape[0], len(self.experts))) + cnts.scatter_(1, topk_ids, 1) + tokens_per_expert = cnts.sum(dim=0) + indicies = topk_ids.view(-1).argsort() + sorted_tokens = hidden_states[indicies // topk_ids.shape[1]] + + # Process experts + outputs = [] + start_idx = 0 + for i, num_tokens in enumerate(tokens_per_expert): + if num_tokens == 0: + continue + end_idx = start_idx + num_tokens + expert = self.experts[i + self.ep_rank * self.experts_per_rank] + tokens_for_this_expert = sorted_tokens[start_idx:end_idx] + expert_out = expert(tokens_for_this_expert) + outputs.append(expert_out) + start_idx = end_idx + + outs = torch.cat(outputs, dim=0) if outputs else sorted_tokens.new_empty(0) + + # Reorder and combine outputs + new_x = torch.empty_like(outs) + new_x[indicies] = outs + hidden_states = ( + new_x.view(*topk_ids.shape, -1) + .type(topk_weight.dtype) + .mul_(topk_weight.unsqueeze(dim=-1)) + .sum(dim=1) + .type(new_x.dtype) + ) + return hidden_states + + def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: + residuals = hidden_states + orig_shape = hidden_states.shape + topk_indices, topk_weights = self.gate(hidden_states) + hidden_states = hidden_states.view(-1, hidden_states.shape[-1]) + hidden_states = self.moe(hidden_states, topk_indices, topk_weights).view(*orig_shape) + hidden_states = hidden_states + self.shared_experts(residuals) + return hidden_states + + +class DeepseekV2MLP(nn.Module): + def __init__(self, config: DeepseekV2Config, hidden_size=None, intermediate_size=None): + super().__init__() + self.config = config + self.hidden_size = config.hidden_size if hidden_size is None else hidden_size + self.intermediate_size = config.intermediate_size if intermediate_size is None else intermediate_size + self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias) + self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=config.mlp_bias) + self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=config.mlp_bias) + 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 + + +@use_kernel_forward_from_hub("RMSNorm") +class DeepseekV2RMSNorm(nn.Module): + def __init__(self, hidden_size, eps=1e-6): + """ + DeepseekV2RMSNorm 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) + + def extra_repr(self): + return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}" + + +class DeepseekV2RotaryEmbedding(nn.Module): + def __init__(self, config: DeepseekV2Config, device=None): + super().__init__() + # BC: "rope_type" was originally "type" + self.rope_type = ( + config.rope_scaling.get("rope_type", config.rope_scaling.get("type")) + if config.rope_scaling is not None + else "default" + ) + + self.max_seq_len_cached = config.max_position_embeddings + self.original_max_seq_len = config.max_position_embeddings + + self.config = config + self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type] + + inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device) + self.register_buffer("inv_freq", inv_freq, persistent=False) + self.original_inv_freq = self.inv_freq + + @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) + 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 torch.autocast(device_type=device_type, enabled=False): # Force float32 + freqs = (inv_freq_expanded.to(x.device) @ position_ids_expanded).transpose(1, 2) + freqs_cis = torch.polar(torch.ones_like(freqs), freqs) # Convert to complex representation + freqs_cis = freqs_cis * self.attention_scaling + + return freqs_cis + + +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 apply_rotary_emb( + xq: torch.Tensor, + xk: torch.Tensor, + freqs_cis: torch.Tensor, +) -> tuple[torch.Tensor, torch.Tensor]: + xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2)) + xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2)) + + # Broadcast to [1, 1, seq_len, dim // 2] + freqs_cis = freqs_cis.unsqueeze(1).to(xq_.device) + + xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3).type_as(xq) + xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3).type_as(xk) + return xq_out, xk_out + + +class DeepseekV2Attention(nn.Module): + """Multi-headed attention from 'Attention Is All You Need' paper""" + + def __init__(self, config: DeepseekV2Config, layer_idx: Optional[int] = None): + super().__init__() + self.config = config + self.layer_idx = layer_idx + self.attention_dropout = config.attention_dropout + self.hidden_size = config.hidden_size + self.num_heads = config.num_attention_heads + self.head_dim = config.head_dim + self.max_position_embeddings = config.max_position_embeddings + self.rope_theta = config.rope_theta + self.q_lora_rank = config.q_lora_rank + self.qk_rope_head_dim = config.qk_rope_head_dim + self.kv_lora_rank = config.kv_lora_rank + self.v_head_dim = config.v_head_dim + self.qk_nope_head_dim = config.qk_nope_head_dim + self.qk_head_dim = config.qk_nope_head_dim + config.qk_rope_head_dim + self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads + + self.is_causal = True + + if self.q_lora_rank is None: + self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.qk_head_dim, bias=False) + else: + self.q_a_proj = nn.Linear(self.hidden_size, config.q_lora_rank, bias=config.attention_bias) + self.q_a_layernorm = DeepseekV2RMSNorm(config.q_lora_rank) + self.q_b_proj = nn.Linear(config.q_lora_rank, self.num_heads * self.qk_head_dim, bias=False) + + self.kv_a_proj_with_mqa = nn.Linear( + self.hidden_size, + config.kv_lora_rank + config.qk_rope_head_dim, + bias=config.attention_bias, + ) + self.kv_a_layernorm = DeepseekV2RMSNorm(config.kv_lora_rank) + self.kv_b_proj = nn.Linear( + config.kv_lora_rank, + self.num_heads * (self.qk_head_dim - self.qk_rope_head_dim + self.v_head_dim), + bias=False, + ) + + self.o_proj = nn.Linear( + self.num_heads * self.v_head_dim, + self.hidden_size, + bias=config.attention_bias, + ) + + self.scaling = self.qk_head_dim ** (-0.5) + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.Tensor] = None, + past_key_value: Optional[Cache] = None, + cache_position: Optional[torch.LongTensor] = None, + position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None, + position_ids: Optional[torch.Tensor] = None, + **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.`" + ) + batch_size, seq_length = hidden_states.shape[:-1] + query_shape = (batch_size, seq_length, -1, self.qk_head_dim) + key_shape = (batch_size, seq_length, -1, self.qk_nope_head_dim + self.v_head_dim) + + if self.q_lora_rank is None: + q = self.q_proj(hidden_states) + else: + q = self.q_b_proj(self.q_a_layernorm(self.q_a_proj(hidden_states))) + q = q.view(query_shape).transpose(1, 2) + q_nope, q_pe = torch.split(q, [self.qk_nope_head_dim, self.qk_rope_head_dim], dim=-1) + + compressed_kv = self.kv_a_proj_with_mqa(hidden_states) + k_nope, k_pe = torch.split(compressed_kv, [self.kv_lora_rank, self.qk_rope_head_dim], dim=-1) + k_nope = self.kv_b_proj(self.kv_a_layernorm(k_nope)).view(key_shape).transpose(1, 2) + k_nope, value_states = torch.split(k_nope, [self.qk_nope_head_dim, self.v_head_dim], dim=-1) + + k_pe = k_pe.view(batch_size, 1, seq_length, self.qk_rope_head_dim) + q_pe, k_pe = apply_rotary_emb(q_pe, k_pe, position_embeddings.to(q_pe.device)) + + k_pe = k_pe.expand(*k_nope.shape[:-1], -1) + query_states = torch.cat((q_nope, q_pe), dim=-1) + key_states = torch.cat((k_nope, k_pe), dim=-1) + + if past_key_value is not None: + # sin and cos are specific to RoPE models; cache_position needed for the static cache + cache_kwargs = {"cache_position": cache_position} + key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs) + + if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim: + value_states = F.pad(value_states, [0, self.qk_head_dim - self.v_head_dim]) + + 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, + ) + + if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim: + attn_output = attn_output[:, :, :, : self.v_head_dim] + + attn_output = attn_output.reshape(batch_size, seq_length, -1).contiguous() + attn_output = self.o_proj(attn_output) + return attn_output, attn_weights + + +class DeepseekV2DecoderLayer(GradientCheckpointingLayer): + def __init__(self, config: DeepseekV2Config, layer_idx: int): + super().__init__() + self.hidden_size = config.hidden_size + + self.self_attn = DeepseekV2Attention(config=config, layer_idx=layer_idx) + self.mlp = DeepseekV2MoE(config) if layer_idx >= config.first_k_dense_replace else DeepseekV2MLP(config) + + self.input_layernorm = DeepseekV2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.post_attention_layernorm = DeepseekV2RMSNorm(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[Cache] = None, + use_cache: Optional[bool] = False, + cache_position: Optional[torch.LongTensor] = None, + position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None, # necessary, but kept here for BC + **kwargs: Unpack[TransformersKwargs], + ) -> tuple[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_value=past_key_value, + 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 DeepseekV2PreTrainedModel(PreTrainedModel): + config_class = DeepseekV2Config + base_model_prefix = "model" + supports_gradient_checkpointing = True + _no_split_modules = ["DeepseekV2DecoderLayer"] + _skip_keys_device_placement = ["past_key_values"] + _supports_flash_attn_2 = True + _supports_sdpa = True + _supports_flex_attn = True + _supports_cache_class = True + _supports_quantized_cache = True + _supports_static_cache = True + _supports_attention_backend = True + _can_record_outputs = { + "hidden_states": DeepseekV2DecoderLayer, + "attentions": DeepseekV2Attention, + } + + def _init_weights(self, module): + std = self.config.initializer_range + if isinstance(module, nn.Linear): + module.weight.data.normal_(mean=0.0, std=std) + if module.bias is not None: + module.bias.data.zero_() + elif isinstance(module, nn.Embedding): + module.weight.data.normal_(mean=0.0, std=std) + if module.padding_idx is not None: + module.weight.data[module.padding_idx].zero_() + elif isinstance(module, DeepseekV2RMSNorm): + module.weight.data.fill_(1.0) + elif isinstance(module, DeepseekV2MoEGate): + module.weight.data.normal_(mean=0.0, std=self.config.initializer_range) + + +@auto_docstring +class DeepseekV2Model(DeepseekV2PreTrainedModel): + def __init__(self, config: DeepseekV2Config): + super().__init__(config) + self.padding_idx = config.pad_token_id + self.vocab_size = config.vocab_size + + self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx) + self.layers = nn.ModuleList( + [DeepseekV2DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)] + ) + self.norm = DeepseekV2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.rotary_emb = DeepseekV2RotaryEmbedding(config=config) + self.gradient_checkpointing = False + + # Initialize weights and apply final processing + self.post_init() + + def get_input_embeddings(self): + return self.embed_tokens + + def set_input_embeddings(self, value): + self.embed_tokens = value + + @check_model_inputs + @auto_docstring + def forward( + self, + input_ids: Optional[torch.LongTensor] = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[Cache] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + cache_position: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + **kwargs: Unpack[TransformersKwargs], + ) -> BaseModelOutputWithPast: + if (input_ids is None) ^ (inputs_embeds is not None): + raise ValueError("You must specify exactly one of input_ids or inputs_embeds") + + if inputs_embeds is None: + inputs_embeds: torch.Tensor = self.embed_tokens(input_ids) + + if use_cache and past_key_values is None: + past_key_values = DynamicCache() + + if cache_position is None: + past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0 + cache_position: torch.Tensor = torch.arange( + past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device + ) + + if position_ids is None: + position_ids = cache_position.unsqueeze(0) + + causal_mask = create_causal_mask( + config=self.config, + input_embeds=inputs_embeds, + attention_mask=attention_mask, + cache_position=cache_position, + past_key_values=past_key_values, + position_ids=position_ids, + ) + + hidden_states = inputs_embeds + position_embeddings = self.rotary_emb(hidden_states, position_ids) + + for decoder_layer in self.layers[: self.config.num_hidden_layers]: + hidden_states = decoder_layer( + hidden_states, + attention_mask=causal_mask, + position_ids=position_ids, + past_key_value=past_key_values, + cache_position=cache_position, + position_embeddings=position_embeddings, + **kwargs, + ) + + hidden_states = self.norm(hidden_states) + return BaseModelOutputWithPast( + last_hidden_state=hidden_states, + past_key_values=past_key_values, + ) + + +@auto_docstring +class DeepseekV2ForCausalLM(DeepseekV2PreTrainedModel, GenerationMixin): + _tied_weights_keys = ["lm_head.weight"] + _tp_plan = {"lm_head": "colwise_rep"} + _pp_plan = {"lm_head": (["hidden_states"], ["logits"])} + + def __init__(self, config): + super().__init__(config) + self.model = DeepseekV2Model(config) + self.vocab_size = config.vocab_size + self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False) + + # 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 + + @can_return_tuple + @auto_docstring + def forward( + self, + input_ids: Optional[torch.LongTensor] = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[Cache] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + cache_position: Optional[torch.LongTensor] = None, + logits_to_keep: Union[int, torch.Tensor] = 0, + **kwargs: Unpack[TransformersKwargs], + ) -> CausalLMOutputWithPast: + r""" + 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]`. + + Example: + + ```python + >>> from transformers import AutoTokenizer, DeepseekV2ForCausalLM + + >>> model = DeepseekV2ForCausalLM.from_pretrained("meta-deepseek_v2/DeepseekV2-2-7b-hf") + >>> tokenizer = AutoTokenizer.from_pretrained("meta-deepseek_v2/DeepseekV2-2-7b-hf") + + >>> prompt = "Hey, are you conscious? Can you talk to me?" + >>> inputs = tokenizer(prompt, return_tensors="pt") + + >>> # Generate + >>> generate_ids = model.generate(inputs.input_ids, max_length=30) + >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0] + "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you." + ```""" + outputs: BaseModelOutputWithPast = self.model( + input_ids=input_ids, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_values=past_key_values, + inputs_embeds=inputs_embeds, + use_cache=use_cache, + cache_position=cache_position, + **kwargs, + ) + + hidden_states = outputs.last_hidden_state + # Only compute necessary logits, and do not upcast them to float if we are not computing the loss + slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep + logits = self.lm_head(hidden_states[:, slice_indices, :]) + + loss = None + if labels is not None: + loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs) + + return CausalLMOutputWithPast( + loss=loss, + logits=logits, + past_key_values=outputs.past_key_values, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + +@auto_docstring( + custom_intro=""" + The DeepseekV2 Model transformer with a sequence classification head on top (linear layer). + + [`DeepseekV2ForSequenceClassification`] uses the last token in order to do the classification, as other causal models + (e.g. GPT-2) do. + + Since it does classification on the last token, it requires to know the position of the last token. If a + `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If + no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the + padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in + each row of the batch). + """ +) +class DeepseekV2ForSequenceClassification(DeepseekV2PreTrainedModel): + def __init__(self, config): + super().__init__(config) + self.num_labels = config.num_labels + self.model = DeepseekV2Model(config) + self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False) + + # 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 + + @can_return_tuple + @auto_docstring + def forward( + self, + input_ids: Optional[torch.LongTensor] = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[Cache] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + **kwargs: Unpack[TransformersKwargs], + ) -> SequenceClassifierOutputWithPast: + r""" + labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): + Labels for computing the sequence classification/regression loss. Indices should be in `[0, ..., + config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If + `config.num_labels > 1` a classification loss is computed (Cross-Entropy). + """ + + transformer_outputs: BaseModelOutputWithPast = self.model( + input_ids, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_values=past_key_values, + inputs_embeds=inputs_embeds, + use_cache=use_cache, + **kwargs, + ) + hidden_states = transformer_outputs.last_hidden_state + logits = self.score(hidden_states) + + if input_ids is not None: + batch_size = input_ids.shape[0] + else: + batch_size = inputs_embeds.shape[0] + + if self.config.pad_token_id is None and batch_size != 1: + raise ValueError("Cannot handle batch sizes > 1 if no padding token is defined.") + if self.config.pad_token_id is None: + last_non_pad_token = -1 + elif input_ids is not None: + # To handle both left- and right- padding, we take the rightmost token that is not equal to pad_token_id + non_pad_mask = (input_ids != self.config.pad_token_id).to(logits.device, torch.int32) + token_indices = torch.arange(input_ids.shape[-1], device=logits.device, dtype=torch.int32) + last_non_pad_token = (token_indices * non_pad_mask).argmax(-1) + else: + last_non_pad_token = -1 + logger.warning_once( + f"{self.__class__.__name__} will not detect padding tokens in `inputs_embeds`. Results may be " + "unexpected if using padding tokens in conjunction with `inputs_embeds.`" + ) + + pooled_logits = logits[torch.arange(batch_size, device=logits.device), last_non_pad_token] + + loss = None + if labels is not None: + loss = self.loss_function(logits=logits, labels=labels, pooled_logits=pooled_logits, config=self.config) + + return SequenceClassifierOutputWithPast( + loss=loss, + logits=pooled_logits, + past_key_values=transformer_outputs.past_key_values, + hidden_states=transformer_outputs.hidden_states, + attentions=transformer_outputs.attentions, + ) + + +__all__ = [ + "DeepseekV2PreTrainedModel", + "DeepseekV2Model", + "DeepseekV2ForCausalLM", + "DeepseekV2ForSequenceClassification", +] diff --git a/src/transformers/models/deepseek_v2/modular_deepseek_v2.py b/src/transformers/models/deepseek_v2/modular_deepseek_v2.py new file mode 100644 index 0000000000..8244dd8572 --- /dev/null +++ b/src/transformers/models/deepseek_v2/modular_deepseek_v2.py @@ -0,0 +1,539 @@ +# coding=utf-8 +# Copyright 2025 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. + +import warnings +from typing import Callable, Optional + +import torch +import torch.nn.functional as F +from torch import nn + +from ...cache_utils import Cache +from ...modeling_utils import ALL_ATTENTION_FUNCTIONS +from ...utils import ( + logging, +) +from ..llama.configuration_llama import LlamaConfig +from ..llama.modeling_llama import ( + LlamaDecoderLayer, + LlamaForCausalLM, + LlamaForSequenceClassification, + LlamaMLP, + LlamaModel, + LlamaPreTrainedModel, + LlamaRMSNorm, + eager_attention_forward, +) +from ..llama4.modeling_llama4 import Llama4TextRotaryEmbedding + + +logger = logging.get_logger(__name__) + + +class DeepseekV2Config(LlamaConfig): + r""" + This is the configuration class to store the configuration of a [`DeepseekV2Model`]. It is used to instantiate a DeepSeek + model according to the specified arguments, defining the model architecture. Instantiating a configuration with the + defaults will yield a similar configuration to that of DeepSeek-V2-Lite" [deepseek-ai/DeepSeek-V2-Lite"](https://huggingface.co/deepseek-ai/DeepSeek-V2-Lite"). + Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the + documentation from [`PretrainedConfig`] for more information. + + Args: + vocab_size (`int`, *optional*, defaults to 32000): + Vocabulary size of the DeepSeek model. Defines the number of different tokens that can be represented by the + `input_ids` passed when calling [`DeepseekV2Model`]. + hidden_size (`int`, *optional*, defaults to 4096): + Dimension of the hidden representations. + intermediate_size (`int`, *optional*, defaults to 11008): + Dimension of the MLP representations. + num_hidden_layers (`int`, *optional*, defaults to 32): + Number of hidden layers in the Transformer decoder. + num_attention_heads (`int`, *optional*, defaults to 32): + Number of attention heads for each attention layer in the Transformer decoder. + num_key_value_heads (`int`, *optional*): + The number of key-value heads used to implement Grouped Query Attention (GQA). If + `num_key_value_heads=num_attention_heads`, the model will use Multi-Head Attention (MHA). If + `num_key_value_heads=1`, the model will use Multi-Query Attention (MQA). Otherwise, GQA is used. + hidden_act (`str` or `function`, *optional*, defaults to `"silu"`): + The non-linear activation function (function or string) in the decoder. + max_position_embeddings (`int`, *optional*, defaults to 2048): + The maximum sequence length that this model might ever be used with. + initializer_range (`float`, *optional*, defaults to 0.02): + The standard deviation of the truncated normal initializer for initializing all weight matrices. + rms_norm_eps (`float`, *optional*, defaults to 1e-06): + The epsilon value used by the RMS normalization layers. + use_cache (`bool`, *optional*, defaults to `True`): + Whether or not the model should return the last key/value attentions (useful for inference optimization). + pad_token_id (`int`, *optional*): + Padding token ID. + bos_token_id (`int`, *optional*, defaults to 1): + Beginning-of-sequence token ID. + eos_token_id (`int`, *optional*, defaults to 2): + End-of-sequence token ID. + tie_word_embeddings (`bool`, *optional*, defaults to `False`): + Whether to tie input and output embeddings. + rope_theta (`float`, *optional*, defaults to 10000.0): + The base period of the Rotary Position Embeddings (RoPE). + rope_scaling (`Dict`, *optional*): + Configuration for scaling RoPE embeddings. Supports `linear` and `dynamic` scaling strategies. + attention_bias (`bool`, *optional*, defaults to `False`): + Whether to use a bias in the query, key, value, and output projection layers during self-attention. + attention_dropout (`float`, *optional*, defaults to 0.0): + The dropout probability applied to attention weights. + mlp_bias (`bool`, *optional*, defaults to `False`): + Whether to use a bias term in the MLP layers. + aux_loss_alpha (`float`, *optional*, defaults to 0.001): + Weight coefficient for auxiliary loss in Mixture of Experts (MoE) models. + first_k_dense_replace (`int`, *optional*, defaults to 0): + Number of dense layers in the shallow layers before switching to MoE layers. + kv_lora_rank (`int`, *optional*, defaults to 512): + Rank of the LoRA decomposition for key-value projections. + q_lora_rank (`int`, *optional*, defaults to 1536): + Rank of the LoRA decomposition for query projections. + Specifically, it determines the dimensionality to which the query (q) vectors are compressed before being expanded back to their original size. + It reduces computational overhead while maintaining model performance. + n_group (`int`, *optional*): + Number of groups for routed experts. + n_routed_experts (`int`, *optional*, defaults to 64): + Number of routed experts (None indicates a dense model). + n_shared_experts (`int`, *optional*, defaults to 2): + Number of shared experts (None indicates a dense model). + qk_nope_head_dim (`int`, *optional*, defaults to 128): + The head dimension for the QK (query-key) projections when using NOPE (Neural Operator Position Encoding). + qk_rope_head_dim (`int`, *optional*, defaults to 64): + The head dimension for QK projections when using RoPE. + routed_scaling_factor (`float`, *optional*, defaults to 1.0): + Scaling factor for routed experts in MoE models. + seq_aux (`bool`, *optional*, defaults to `True`): + Whether to compute the auxiliary loss for each individual sequence. + topk_group (`int`, *optional*): + Number of selected groups per token for expert selection. + topk_method (`str`, *optional*, defaults to `"greedy"`): + The method used for selecting top-k experts in the routed gate mechanism. + v_head_dim (`int`, *optional*, defaults to 128): + The dimension of value projections in the attention layers. + num_experts_per_tok (`int`, *optional*): + The number of experts selected per token. If `None`, the model behaves as a dense Transformer. + norm_topk_prob (`bool`, *optional*, defaults to `False`): + Whether to normalize the probability distribution over top-k selected experts. + moe_intermediate_size (`int`, *optional*, defaults to 1407): + Dimension of the MoE (Mixture of Experts) representations. + + ```python + >>> from transformers import DeepseekV2Model, DeepseekV2Config + >>> # Initializing a DeepSeek-V2 style configuration + >>> configuration = DeepseekV2Config() + >>> # Accessing the model configuration + >>> model = DeepseekV2Model(configuration) + >>> print(model.config) + ``` + """ + + base_model_tp_plan = { + "layers.*.self_attn.q_proj": "colwise", + "layers.*.self_attn.q_a_proj": "colwise", + "layers.*.self_attn.q_b_proj": "colwise", + "layers.*.self_attn.kv_b_proj": "colwise", + "layers.*.self_attn.o_proj": "rowwise", + "layers.*.mlp.gate_proj": "colwise", + "layers.*.mlp.up_proj": "colwise", + "layers.*.mlp.down_proj": "rowwise", + } + + model_type = "deepseek_v2" + keys_to_ignore_at_inference = ["past_key_values"] + + def __init__( + self, + vocab_size=32000, + hidden_size=4096, + intermediate_size=11008, + num_hidden_layers=32, + num_attention_heads=32, + num_key_value_heads=None, + hidden_act="silu", + max_position_embeddings=2048, + initializer_range=0.02, + rms_norm_eps=1e-6, + use_cache=True, + pad_token_id=None, + bos_token_id=1, + eos_token_id=2, + tie_word_embeddings=False, + rope_theta=10000.0, + rope_scaling=None, + attention_bias=False, + attention_dropout=0.0, + mlp_bias=False, + aux_loss_alpha=0.001, + first_k_dense_replace=0, + kv_lora_rank=512, + q_lora_rank=1536, + n_group=None, + n_routed_experts=64, + n_shared_experts=2, + qk_nope_head_dim=128, + qk_rope_head_dim=64, + routed_scaling_factor=1.0, + seq_aux=True, + topk_group=None, + topk_method="greedy", + v_head_dim=128, + num_experts_per_tok=None, + norm_topk_prob=False, + moe_intermediate_size=1407, + **kwargs, + ): + super().__init__(**kwargs) + + del self.pretraining_tp + self.aux_loss_alpha = aux_loss_alpha + self.first_k_dense_replace = first_k_dense_replace + self.kv_lora_rank = kv_lora_rank + self.q_lora_rank = q_lora_rank + self.n_group = n_group + self.n_routed_experts = n_routed_experts + self.n_shared_experts = n_shared_experts + self.qk_nope_head_dim = qk_nope_head_dim + self.qk_rope_head_dim = qk_rope_head_dim + self.routed_scaling_factor = routed_scaling_factor + self.seq_aux = seq_aux + self.topk_group = topk_group + self.topk_method = topk_method + self.v_head_dim = v_head_dim + self.num_experts_per_tok = num_experts_per_tok + self.norm_topk_prob = norm_topk_prob + self.moe_intermediate_size = moe_intermediate_size + self.head_dim = qk_rope_head_dim + + +def apply_rotary_emb( + xq: torch.Tensor, + xk: torch.Tensor, + freqs_cis: torch.Tensor, +) -> tuple[torch.Tensor, torch.Tensor]: + xq_ = torch.view_as_complex(xq.float().reshape(*xq.shape[:-1], -1, 2)) + xk_ = torch.view_as_complex(xk.float().reshape(*xk.shape[:-1], -1, 2)) + + # Broadcast to [1, 1, seq_len, dim // 2] + freqs_cis = freqs_cis.unsqueeze(1).to(xq_.device) + + xq_out = torch.view_as_real(xq_ * freqs_cis).flatten(3).type_as(xq) + xk_out = torch.view_as_real(xk_ * freqs_cis).flatten(3).type_as(xk) + return xq_out, xk_out + + +class DeepseekV2MoEGate(nn.Module): + def __init__(self, config: DeepseekV2Config): + super().__init__() + self.config = config + self.top_k = config.num_experts_per_tok + self.num_experts = config.n_routed_experts + self.routed_scaling_factor = config.routed_scaling_factor + self.alpha = config.aux_loss_alpha + self.seq_aux = config.seq_aux + self.topk_method = config.topk_method + self.num_group = config.n_group + self.topk_group = config.topk_group + + # topk selection algorithm + self.norm_topk_prob = config.norm_topk_prob + self.gating_dim = config.hidden_size + self.weight = nn.Parameter(torch.empty((self.num_experts, self.gating_dim))) + + def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: + batch_size, seq_len, hidden_dim = hidden_states.shape + ### compute gating score + hidden_states = hidden_states.view(-1, hidden_dim) + logits = F.linear(hidden_states.type(torch.float32), self.weight.type(torch.float32), None) + scores = logits.softmax(dim=-1, dtype=torch.float32) + + # select top-k experts + # greedy method is used for DeepSeek-V2-Lite + # group_limited_greedy for DeepSeek-V2 and DeepSeek-V2-Chat + if self.topk_method == "greedy": + topk_weight, topk_idx = torch.topk(scores, k=self.top_k, dim=-1, sorted=False) + elif self.topk_method == "group_limited_greedy": + group_scores = scores.view(batch_size * seq_len, self.num_group, -1).max(dim=-1).values # [n, num_group] + group_idx = torch.topk(group_scores, k=self.topk_group, dim=-1, sorted=False)[1] # [n, top_k_group] + group_mask = torch.zeros_like(group_scores) # [n, num_group] + group_mask.scatter_(1, group_idx, 1) # [n, num_group] + score_mask = ( + group_mask.unsqueeze(-1) + .expand(batch_size * seq_len, self.num_group, self.num_experts // self.num_group) + .reshape(batch_size * seq_len, -1) + ) # [n, e] + tmp_scores = scores.masked_fill(~score_mask.bool(), 0.0) # [n, e] + topk_weight, topk_idx = torch.topk(tmp_scores, k=self.top_k, dim=-1, sorted=False) + + topk_weight = topk_weight * self.routed_scaling_factor + ### expert-level computation auxiliary loss + return topk_idx, topk_weight + + +class DeepseekV2MoE(nn.Module): + """ + A mixed expert module containing shared experts. + """ + + def __init__(self, config: DeepseekV2Config): + super().__init__() + self.config = config + self.num_experts_per_tok = config.num_experts_per_tok + + self.experts = nn.ModuleList( + [ + (DeepseekV2MLP(config, intermediate_size=config.moe_intermediate_size)) + for _ in range(config.n_routed_experts) + ] + ) + self.gate = DeepseekV2MoEGate(config) + if config.n_shared_experts is not None: + intermediate_size = config.moe_intermediate_size * config.n_shared_experts + self.shared_experts = DeepseekV2MLP(config=config, intermediate_size=intermediate_size) + self.ep_rank = 0 + self.experts_per_rank = config.n_routed_experts + + def moe(self, hidden_states: torch.Tensor, topk_ids: torch.Tensor, topk_weight: torch.Tensor) -> torch.Tensor: + cnts = topk_ids.new_zeros((topk_ids.shape[0], len(self.experts))) + cnts.scatter_(1, topk_ids, 1) + tokens_per_expert = cnts.sum(dim=0) + indicies = topk_ids.view(-1).argsort() + sorted_tokens = hidden_states[indicies // topk_ids.shape[1]] + + # Process experts + outputs = [] + start_idx = 0 + for i, num_tokens in enumerate(tokens_per_expert): + if num_tokens == 0: + continue + end_idx = start_idx + num_tokens + expert = self.experts[i + self.ep_rank * self.experts_per_rank] + tokens_for_this_expert = sorted_tokens[start_idx:end_idx] + expert_out = expert(tokens_for_this_expert) + outputs.append(expert_out) + start_idx = end_idx + + outs = torch.cat(outputs, dim=0) if outputs else sorted_tokens.new_empty(0) + + # Reorder and combine outputs + new_x = torch.empty_like(outs) + new_x[indicies] = outs + hidden_states = ( + new_x.view(*topk_ids.shape, -1) + .type(topk_weight.dtype) + .mul_(topk_weight.unsqueeze(dim=-1)) + .sum(dim=1) + .type(new_x.dtype) + ) + return hidden_states + + def forward(self, hidden_states: torch.Tensor) -> torch.Tensor: + residuals = hidden_states + orig_shape = hidden_states.shape + topk_indices, topk_weights = self.gate(hidden_states) + hidden_states = hidden_states.view(-1, hidden_states.shape[-1]) + hidden_states = self.moe(hidden_states, topk_indices, topk_weights).view(*orig_shape) + hidden_states = hidden_states + self.shared_experts(residuals) + return hidden_states + + +class DeepseekV2MLP(LlamaMLP): + def __init__(self, config: DeepseekV2Config, hidden_size=None, intermediate_size=None): + super().__init__(config) + self.hidden_size = config.hidden_size if hidden_size is None else hidden_size + self.intermediate_size = config.intermediate_size if intermediate_size is None else intermediate_size + + +class DeepseekV2RMSNorm(LlamaRMSNorm): + pass + + +class DeepseekV2RotaryEmbedding(Llama4TextRotaryEmbedding): + def __init__(self, config: DeepseekV2Config, device=None): + super().__init__(config=config, device=device) + # BC: "rope_type" was originally "type" + self.rope_type = ( + config.rope_scaling.get("rope_type", config.rope_scaling.get("type")) + if config.rope_scaling is not None + else "default" + ) + + +class DeepseekV2Attention(nn.Module): + """Multi-headed attention from 'Attention Is All You Need' paper""" + + def __init__(self, config: DeepseekV2Config, layer_idx: Optional[int] = None): + super().__init__() + self.config = config + self.layer_idx = layer_idx + self.attention_dropout = config.attention_dropout + self.hidden_size = config.hidden_size + self.num_heads = config.num_attention_heads + self.head_dim = config.head_dim + self.max_position_embeddings = config.max_position_embeddings + self.rope_theta = config.rope_theta + self.q_lora_rank = config.q_lora_rank + self.qk_rope_head_dim = config.qk_rope_head_dim + self.kv_lora_rank = config.kv_lora_rank + self.v_head_dim = config.v_head_dim + self.qk_nope_head_dim = config.qk_nope_head_dim + self.qk_head_dim = config.qk_nope_head_dim + config.qk_rope_head_dim + self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads + + self.is_causal = True + + if self.q_lora_rank is None: + self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.qk_head_dim, bias=False) + else: + self.q_a_proj = nn.Linear(self.hidden_size, config.q_lora_rank, bias=config.attention_bias) + self.q_a_layernorm = DeepseekV2RMSNorm(config.q_lora_rank) + self.q_b_proj = nn.Linear(config.q_lora_rank, self.num_heads * self.qk_head_dim, bias=False) + + self.kv_a_proj_with_mqa = nn.Linear( + self.hidden_size, + config.kv_lora_rank + config.qk_rope_head_dim, + bias=config.attention_bias, + ) + self.kv_a_layernorm = DeepseekV2RMSNorm(config.kv_lora_rank) + self.kv_b_proj = nn.Linear( + config.kv_lora_rank, + self.num_heads * (self.qk_head_dim - self.qk_rope_head_dim + self.v_head_dim), + bias=False, + ) + + self.o_proj = nn.Linear( + self.num_heads * self.v_head_dim, + self.hidden_size, + bias=config.attention_bias, + ) + + self.scaling = self.qk_head_dim ** (-0.5) + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.Tensor] = None, + past_key_value: Optional[Cache] = None, + cache_position: Optional[torch.LongTensor] = None, + position_embeddings: Optional[tuple[torch.Tensor, torch.Tensor]] = None, + position_ids: Optional[torch.Tensor] = None, + **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.`" + ) + batch_size, seq_length = hidden_states.shape[:-1] + query_shape = (batch_size, seq_length, -1, self.qk_head_dim) + key_shape = (batch_size, seq_length, -1, self.qk_nope_head_dim + self.v_head_dim) + + if self.q_lora_rank is None: + q = self.q_proj(hidden_states) + else: + q = self.q_b_proj(self.q_a_layernorm(self.q_a_proj(hidden_states))) + q = q.view(query_shape).transpose(1, 2) + q_nope, q_pe = torch.split(q, [self.qk_nope_head_dim, self.qk_rope_head_dim], dim=-1) + + compressed_kv = self.kv_a_proj_with_mqa(hidden_states) + k_nope, k_pe = torch.split(compressed_kv, [self.kv_lora_rank, self.qk_rope_head_dim], dim=-1) + k_nope = self.kv_b_proj(self.kv_a_layernorm(k_nope)).view(key_shape).transpose(1, 2) + k_nope, value_states = torch.split(k_nope, [self.qk_nope_head_dim, self.v_head_dim], dim=-1) + + k_pe = k_pe.view(batch_size, 1, seq_length, self.qk_rope_head_dim) + q_pe, k_pe = apply_rotary_emb(q_pe, k_pe, position_embeddings.to(q_pe.device)) + + k_pe = k_pe.expand(*k_nope.shape[:-1], -1) + query_states = torch.cat((q_nope, q_pe), dim=-1) + key_states = torch.cat((k_nope, k_pe), dim=-1) + + if past_key_value is not None: + # sin and cos are specific to RoPE models; cache_position needed for the static cache + cache_kwargs = {"cache_position": cache_position} + key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs) + + if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim: + value_states = F.pad(value_states, [0, self.qk_head_dim - self.v_head_dim]) + + 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, + ) + + if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim: + attn_output = attn_output[:, :, :, : self.v_head_dim] + + attn_output = attn_output.reshape(batch_size, seq_length, -1).contiguous() + attn_output = self.o_proj(attn_output) + return attn_output, attn_weights + + +class DeepseekV2DecoderLayer(LlamaDecoderLayer): + def __init__(self, config: DeepseekV2Config, layer_idx: int): + super().__init__(config, layer_idx) + + self.self_attn = DeepseekV2Attention(config=config, layer_idx=layer_idx) + self.mlp = DeepseekV2MoE(config) if layer_idx >= config.first_k_dense_replace else DeepseekV2MLP(config) + + self.input_layernorm = DeepseekV2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.post_attention_layernorm = DeepseekV2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + + +class DeepseekV2PreTrainedModel(LlamaPreTrainedModel): + def _init_weights(self, module): + std = self.config.initializer_range + if isinstance(module, nn.Linear): + module.weight.data.normal_(mean=0.0, std=std) + if module.bias is not None: + module.bias.data.zero_() + elif isinstance(module, nn.Embedding): + module.weight.data.normal_(mean=0.0, std=std) + if module.padding_idx is not None: + module.weight.data[module.padding_idx].zero_() + elif isinstance(module, DeepseekV2RMSNorm): + module.weight.data.fill_(1.0) + elif isinstance(module, DeepseekV2MoEGate): + module.weight.data.normal_(mean=0.0, std=self.config.initializer_range) + + +class DeepseekV2Model(LlamaModel): + pass + + +class DeepseekV2ForCausalLM(LlamaForCausalLM): + pass + + +class DeepseekV2ForSequenceClassification(LlamaForSequenceClassification): + pass + + +__all__ = [ + "DeepseekV2PreTrainedModel", + "DeepseekV2Model", + "DeepseekV2ForCausalLM", + "DeepseekV2ForSequenceClassification", + "DeepseekV2Config", +] diff --git a/tests/models/deepseek_v2/__init__.py b/tests/models/deepseek_v2/__init__.py new file mode 100644 index 0000000000..e69de29bb2 diff --git a/tests/models/deepseek_v2/test_modeling_deepseek_v2.py b/tests/models/deepseek_v2/test_modeling_deepseek_v2.py new file mode 100644 index 0000000000..02d087cb8b --- /dev/null +++ b/tests/models/deepseek_v2/test_modeling_deepseek_v2.py @@ -0,0 +1,269 @@ +# 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. +"""Testing suite for the PyTorch DeepSeekV2 model.""" + +import unittest + +from transformers import BitsAndBytesConfig, Cache, DeepseekV2Config, is_torch_available +from transformers.testing_utils import require_read_token, require_torch, require_torch_accelerator, slow, torch_device + +from ...causal_lm_tester import CausalLMModelTest, CausalLMModelTester + + +if is_torch_available(): + import torch + + from transformers import AutoTokenizer, DeepseekV2ForCausalLM, DeepseekV2ForSequenceClassification, DeepseekV2Model + from transformers.models.deepseek_v2.modeling_deepseek_v2 import DeepseekV2RotaryEmbedding + + +class DeepseekV2ModelTester(CausalLMModelTester): + if is_torch_available(): + config_class = DeepseekV2Config + base_model_class = DeepseekV2Model + causal_lm_class = DeepseekV2ForCausalLM + sequence_class = DeepseekV2ForSequenceClassification + + def __init__( + self, + parent, + n_routed_experts=8, + kv_lora_rank=32, + q_lora_rank=16, + qk_nope_head_dim=64, + qk_rope_head_dim=64, + ): + super().__init__(parent=parent) + self.n_routed_experts = n_routed_experts + self.kv_lora_rank = kv_lora_rank + self.q_lora_rank = q_lora_rank + self.qk_nope_head_dim = qk_nope_head_dim + self.qk_rope_head_dim = qk_rope_head_dim + + +@require_torch +class DeepseekV2ModelTest(CausalLMModelTest, unittest.TestCase): + all_model_classes = ( + ( + DeepseekV2ForCausalLM, + DeepseekV2ForSequenceClassification, + DeepseekV2Model, + ) + if is_torch_available() + else () + ) + pipeline_model_mapping = ( + { + "feature-extraction": DeepseekV2Model, + "text-classification": DeepseekV2ForSequenceClassification, + "text-generation": DeepseekV2ForCausalLM, + "zero-shot": DeepseekV2ForSequenceClassification, + } + if is_torch_available() + else {} + ) + test_headmasking = False + test_pruning = False + fx_compatible = False + test_torchscript = False + model_tester_class = DeepseekV2ModelTester + rotary_embedding_layer = DeepseekV2RotaryEmbedding + model_split_percents = [0.5, 0.7, 0.8] + + # used in `test_torch_compile_for_training` + _torch_compile_train_cls = DeepseekV2ForCausalLM if is_torch_available() else None + + def test_model_rope_scaling(self): + config, _ = self.model_tester.prepare_config_and_inputs_for_common() + scaling_factor = 10 + short_input_length = 10 + long_input_length = int(config.max_position_embeddings * 1.5) + + # Inputs + x = torch.randn(1, dtype=torch.float32, device=torch_device) # used exlusively to get the dtype and the device + position_ids_short = torch.arange(short_input_length, dtype=torch.long, device=torch_device) + position_ids_short = position_ids_short.unsqueeze(0) + position_ids_long = torch.arange(long_input_length, dtype=torch.long, device=torch_device) + position_ids_long = position_ids_long.unsqueeze(0) + + # Sanity check original RoPE + original_rope = DeepseekV2RotaryEmbedding(config=config).to(torch_device) + original_freqs_cis_short = original_rope(x, position_ids_short) + original_freqs_cis_long = original_rope(x, position_ids_long) + torch.testing.assert_close(original_freqs_cis_short, original_freqs_cis_long[:, :short_input_length, :]) + + # Sanity check linear RoPE scaling + # New position "x" should match original position with index "x/scaling_factor" + config.rope_scaling = {"type": "linear", "factor": scaling_factor} + linear_scaling_rope = DeepseekV2RotaryEmbedding(config=config).to(torch_device) + linear_freqs_cis_short = linear_scaling_rope(x, position_ids_short) + linear_freqs_cis_long = linear_scaling_rope(x, position_ids_long) + torch.testing.assert_close(linear_freqs_cis_short, linear_freqs_cis_long[:, :short_input_length, :]) + + # Sanity check Dynamic NTK RoPE scaling + # Scaling should only be observed after a long input is fed. We can observe that the frequencies increase + # with scaling_factor (or that `inv_freq` decreases) + config.rope_scaling = {"type": "dynamic", "factor": scaling_factor} + ntk_scaling_rope = DeepseekV2RotaryEmbedding(config=config).to(torch_device) + ntk_freqs_cis_short = ntk_scaling_rope(x, position_ids_short) + ntk_freqs_cis_long = ntk_scaling_rope(x, position_ids_long) + torch.testing.assert_close(ntk_freqs_cis_short, original_freqs_cis_short) + with self.assertRaises(AssertionError): + torch.testing.assert_close(ntk_freqs_cis_long, original_freqs_cis_long) + self.assertTrue((ntk_scaling_rope.inv_freq <= original_rope.inv_freq).all()) + + # Sanity check Yarn RoPE scaling + # Scaling should be over the entire input + config.rope_scaling = {"type": "yarn", "factor": scaling_factor} + yarn_scaling_rope = DeepseekV2RotaryEmbedding(config=config).to(torch_device) + yarn_freqs_cis_short = yarn_scaling_rope(x, position_ids_short) + yarn_freqs_cis_long = yarn_scaling_rope(x, position_ids_long) + torch.testing.assert_close(yarn_freqs_cis_short, yarn_freqs_cis_long[:, :short_input_length, :]) + with self.assertRaises(AssertionError): + torch.testing.assert_close(yarn_freqs_cis_short, original_freqs_cis_short) + with self.assertRaises(AssertionError): + torch.testing.assert_close(yarn_freqs_cis_long, original_freqs_cis_long) + + def test_past_key_values_format(self): + """ + Overwriting to pass the expected cache shapes (Deepseek-V3 uses MLA so the cache shapes are non-standard) + """ + config, inputs = self.model_tester.prepare_config_and_inputs_for_common() + batch_size, seq_length = inputs["input_ids"].shape + # difference: last dim + k_embed_dim = config.qk_nope_head_dim + config.qk_rope_head_dim + v_embed_dim = config.v_head_dim + self_attention_key_cache_shape = (batch_size, config.num_key_value_heads, seq_length, k_embed_dim) + self_attention_value_cache_shape = (batch_size, config.num_key_value_heads, seq_length, v_embed_dim) + # build the full cache shapes + num_hidden_layers = config.num_hidden_layers + all_cache_shapes = [ + [self_attention_key_cache_shape, self_attention_value_cache_shape] for _ in range(num_hidden_layers) + ] + super().test_past_key_values_format(custom_all_cache_shapes=all_cache_shapes) + + def _check_past_key_values_for_generate(self, batch_size, decoder_past_key_values, cache_length, config): + """Needs to be overriden as deepseek has special MLA cache format (though we don't really use the MLA)""" + self.assertIsInstance(decoder_past_key_values, Cache) + + # (batch, head, seq_length, head_features) + expected_common_shape = ( + batch_size, + config.num_key_value_heads if hasattr(config, "num_key_value_heads") else config.num_attention_heads, + cache_length, + ) + expected_key_shape = expected_common_shape + (config.qk_nope_head_dim + config.qk_rope_head_dim,) + expected_value_shape = expected_common_shape + (config.v_head_dim,) + + if isinstance(decoder_past_key_values, Cache): + self.assertListEqual( + [key_tensor.shape for key_tensor in decoder_past_key_values.key_cache], + [expected_key_shape] * len(decoder_past_key_values.key_cache), + ) + self.assertListEqual( + [value_tensor.shape for value_tensor in decoder_past_key_values.value_cache], + [expected_value_shape] * len(decoder_past_key_values.value_cache), + ) + + @unittest.skip("Deepseek-V2 uses MLA which has a special head dim and is not compatible with StaticCache shape") + def test_generate_compilation_all_outputs(self): + pass + + @unittest.skip("Deepseek-V2 uses MLA which has a special head dim and is not compatible with StaticCache shape") + def test_generate_compile_model_forward(self): + pass + + @unittest.skip("Deepseek-V2 uses MLA which has a special head dim and is not compatible with StaticCache shape") + def test_generate_from_inputs_embeds_with_static_cache(self): + pass + + @unittest.skip("Deepseek-V2 uses MLA which has a special head dim and is not compatible with StaticCache shape") + def test_generate_with_static_cache(self): + pass + + @unittest.skip("Dynamic control flow in MoE") + def test_torch_compile_for_training(self): + pass + + +@slow +@require_read_token +@require_torch_accelerator +class DeepseekV2IntegrationTest(unittest.TestCase): + def test_deepseek_v2_lite(self): + EXPECTED_TEXT = ['An attention function can be described as mapping a query and a set of key-value pairs to an output, where the query, keys, values, and output are all vectors.\n\nAttention functions are used in a variety of applications, including natural language processing, computer vision, and reinforcement learning.\n\nThe attention function is a function that takes a query and a set of key-value pairs as input and outputs a vector'] # fmt: skip + + tokenizer = AutoTokenizer.from_pretrained("deepseek-ai/DeepSeek-V2-Lite") + model = DeepseekV2ForCausalLM.from_pretrained( + "deepseek-ai/DeepSeek-V2-Lite", + device_map=torch_device, + torch_dtype=torch.bfloat16, + quantization_config=BitsAndBytesConfig(load_in_8bit=True), + ) + + input_text = [ + "An attention function can be described as mapping a query and a set of key-value pairs to an output, where the query, keys, values, and output are all vectors." # fmt: skip + ] + model_inputs = tokenizer(input_text, return_tensors="pt").to(model.device) + + generated_ids = model.generate(**model_inputs, max_new_tokens=50, do_sample=False) + generated_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True) + self.assertEqual(generated_text, EXPECTED_TEXT) + + def test_logits_eager(self): + input_ids = [1, 306, 4658, 278, 6593, 310, 2834, 338] + + model = DeepseekV2ForCausalLM.from_pretrained( + "deepseek-ai/DeepSeek-V2-Lite", + device_map=torch_device, + torch_dtype=torch.bfloat16, + quantization_config=BitsAndBytesConfig(load_in_8bit=True), + attn_implementation="eager", + ) + + with torch.no_grad(): + out = model(torch.tensor([input_ids]).to(torch_device)) + + EXPECTED_MEAN = torch.tensor([[-6.1232, -5.0952, -4.4493, -2.6536, -2.0608, -2.3991, -3.8013, -2.8681]], device=torch_device) # fmt: skip + torch.testing.assert_close(out.logits.float().mean(-1), EXPECTED_MEAN, atol=1e-3, rtol=1e-3) + + EXPECTED_SLICE = torch.tensor([-1.2500, -0.9961, -0.0194, -3.1562, 1.2812, -2.7656, -0.8438, -3.0469, -2.7812, -0.6328, -0.4160, -1.9688, -2.4219, -1.0391, -3.8906], device=torch_device) # fmt: skip + torch.testing.assert_close(out.logits[0, 0, :15].float(), EXPECTED_SLICE, atol=1e-3, rtol=1e-3) + + def test_batch_fa2(self): + EXPECTED_TEXT = [ + "Simply put, the theory of relativity states that \nthe laws of physics are the same for all observers, regardless of their \nrelative motion.\nThe theory of relativity is a theory of space, time, and gravity.\nThe theory of", # fmt: skip + "My favorite all time favorite condiment is ketchup. I love ketchup. I love ketchup on my hot dogs, hamburgers, french fries, and even on my eggs. I love ketchup. I love ketchup so much that I", # fmt: skip + ] + + prompts = [ + "Simply put, the theory of relativity states that ", + "My favorite all time favorite condiment is ketchup.", + ] + tokenizer = AutoTokenizer.from_pretrained( + "deepseek-ai/DeepSeek-V2-Lite", pad_token="", padding_side="right" + ) + + model = DeepseekV2ForCausalLM.from_pretrained( + "deepseek-ai/DeepSeek-V2-Lite", + device_map=torch_device, + torch_dtype=torch.bfloat16, + quantization_config=BitsAndBytesConfig(load_in_8bit=True), + ) + inputs = tokenizer(prompts, return_tensors="pt", padding=True).to(model.device) + + generated_ids = model.generate(**inputs, max_new_tokens=40, do_sample=False) + generated_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True) + self.assertEqual(EXPECTED_TEXT, generated_text)