diff --git a/docs/source/en/index.md b/docs/source/en/index.md index bf1fd008692..ac026067ac2 100644 --- a/docs/source/en/index.md +++ b/docs/source/en/index.md @@ -145,6 +145,7 @@ Flax), PyTorch, and/or TensorFlow. | [Funnel Transformer](model_doc/funnel) | ✅ | ✅ | ❌ | | [Fuyu](model_doc/fuyu) | ✅ | ❌ | ❌ | | [Gemma](model_doc/gemma) | ✅ | ❌ | ✅ | +| [Gemma2](model_doc/gemma2) | ✅ | ❌ | ❌ | | [GIT](model_doc/git) | ✅ | ❌ | ❌ | | [GLPN](model_doc/glpn) | ✅ | ❌ | ❌ | | [GPT Neo](model_doc/gpt_neo) | ✅ | ❌ | ✅ | diff --git a/docs/source/en/model_doc/gemma2.md b/docs/source/en/model_doc/gemma2.md new file mode 100644 index 00000000000..fa16dfbc4ba --- /dev/null +++ b/docs/source/en/model_doc/gemma2.md @@ -0,0 +1,58 @@ + + + +# Gemma2 + +## Overview + +The Gemma2 model was proposed in [Gemma2: Open Models Based on Gemini Technology and Research](https://blog.google/technology/developers/Gemma2-open-models/) by Gemma2 Team, Google. +Gemma2 models are trained on 6T tokens, and released with 2 versions, 2b and 7b. + +The abstract from the paper is the following: + +*This work introduces Gemma2, a new family of open language models demonstrating strong performance across academic benchmarks for language understanding, reasoning, and safety. We release two sizes of models (2 billion and 7 billion parameters), and provide both pretrained and fine-tuned checkpoints. Gemma2 outperforms similarly sized open models on 11 out of 18 text-based tasks, and we present comprehensive evaluations of safety and responsibility aspects of the models, alongside a detailed description of our model development. We believe the responsible release of LLMs is critical for improving the safety of frontier models, and for enabling the next wave of LLM innovations* + +Tips: + +- The original checkpoints can be converted using the conversion script `src/transformers/models/Gemma2/convert_Gemma2_weights_to_hf.py` + +This model was contributed by [Arthur Zucker](https://huggingface.co/ArthurZ), [Pedro Cuenca](https://huggingface.co/pcuenq) and [Tom Arsen](). + + +## Gemma2Config + +[[autodoc]] Gemma2Config + +## Gemma2Model + +[[autodoc]] Gemma2Model + - forward + +## Gemma2ForCausalLM + +[[autodoc]] Gemma2ForCausalLM + - forward + +## Gemma2ForSequenceClassification + +[[autodoc]] Gemma2ForSequenceClassification + - forward + +## Gemma2ForTokenClassification + +[[autodoc]] Gemma2ForTokenClassification + - forward diff --git a/src/transformers/__init__.py b/src/transformers/__init__.py index 922e19915d8..7b39fd479ed 100755 --- a/src/transformers/__init__.py +++ b/src/transformers/__init__.py @@ -435,6 +435,7 @@ _import_structure = { ], "models.fuyu": ["FuyuConfig"], "models.gemma": ["GemmaConfig"], + "models.gemma2": ["Gemma2Config"], "models.git": [ "GitConfig", "GitProcessor", @@ -2181,6 +2182,15 @@ else: "GemmaPreTrainedModel", ] ) + _import_structure["models.gemma2"].extend( + [ + "Gemma2ForCausalLM", + "Gemma2ForSequenceClassification", + "Gemma2ForTokenClassification", + "Gemma2Model", + "Gemma2PreTrainedModel", + ] + ) _import_structure["models.git"].extend( [ "GitForCausalLM", @@ -5062,6 +5072,7 @@ if TYPE_CHECKING: ) from .models.fuyu import FuyuConfig from .models.gemma import GemmaConfig + from .models.gemma2 import Gemma2Config from .models.git import ( GitConfig, GitProcessor, @@ -6694,6 +6705,13 @@ if TYPE_CHECKING: GemmaModel, GemmaPreTrainedModel, ) + from .models.gemma2 import ( + Gemma2ForCausalLM, + Gemma2ForSequenceClassification, + Gemma2ForTokenClassification, + Gemma2Model, + Gemma2PreTrainedModel, + ) from .models.git import ( GitForCausalLM, GitModel, diff --git a/src/transformers/cache_utils.py b/src/transformers/cache_utils.py index 532b921a369..b167cd1d117 100644 --- a/src/transformers/cache_utils.py +++ b/src/transformers/cache_utils.py @@ -970,3 +970,125 @@ class SlidingWindowCache(StaticCache): # in theory there is no limit because the sliding window size is fixed # no matter how long the sentence is return None + + +class HybridCache(Cache): + def __init__(self, config: PretrainedConfig, max_batch_size, max_cache_len, device="cpu", dtype=None) -> None: + if not hasattr(config, "sliding_window") or config.sliding_window is None: + raise ValueError( + "Setting `cache_implementation` to 'sliding_window' requires the model config supporting " + "sliding window attention, please check if there is a `sliding_window` field in the model " + "config and it's not set to None." + ) + self.max_cache_len = max_cache_len + self.max_batch_size = max_batch_size + # Some model define a custom `head_dim` != config.hidden_size // config.num_attention_heads + self.head_dim = ( + config.head_dim if hasattr(config, "head_dim") else config.hidden_size // config.num_attention_heads + ) + + self.dtype = dtype if dtype is not None else torch.float32 + self.num_key_value_heads = ( + config.num_attention_heads if config.num_key_value_heads is None else config.num_key_value_heads + ) + self.is_sliding = torch.tensor( + [i % 2 for i in range(config.num_hidden_layers)], dtype=torch.bool, device=device + ) + self.key_cache: List[torch.Tensor] = [] + self.value_cache: List[torch.Tensor] = [] + global_cache_shape = (max_batch_size, self.num_key_value_heads, max_cache_len, self.head_dim) + sliding_cache_shape = ( + max_batch_size, + self.num_key_value_heads, + min(config.sliding_window, max_cache_len), + self.head_dim, + ) + for i in range(config.num_hidden_layers): + # Note: `mark_static_address` is used to tag the cache as an fixed data pointer, preventing cuda graph + # breaks when updating the cache. + cache_shape = global_cache_shape if not self.is_sliding[i] else sliding_cache_shape + new_layer_key_cache = torch.zeros(cache_shape, dtype=self.dtype, device=device) + new_layer_value_cache = torch.zeros(cache_shape, dtype=self.dtype, device=device) + torch._dynamo.mark_static_address(new_layer_key_cache) + torch._dynamo.mark_static_address(new_layer_value_cache) + self.key_cache.append(new_layer_key_cache) + self.value_cache.append(new_layer_value_cache) + + def _sliding_update(self, cache_position, layer_idx, key_states, value_states, k_out, v_out, max_cache_len): + if cache_position.shape[0] > max_cache_len: + k_out = key_states[:, :, -max_cache_len:, :] + v_out = value_states[:, :, -max_cache_len:, :] + # Assumption: caches are all zeros at this point, `+=` is equivalent to `=` but compile-friendly + self.key_cache[layer_idx] += k_out + self.value_cache[layer_idx] += v_out + # we should return the whole states instead of k_out, v_out to take the whole prompt + # into consideration when building kv cache instead of just throwing away tokens outside of the window + return key_states, value_states + + slicing = torch.ones(max_cache_len, dtype=torch.long, device=value_states.device).cumsum(0) + cache_position = cache_position.clamp(0, max_cache_len - 1) + to_shift = cache_position >= max_cache_len - 1 + indices = (slicing + to_shift[-1].int() - 1) % max_cache_len + k_out = k_out[:, :, indices] + v_out = v_out[:, :, indices] + + k_out[:, :, cache_position] = key_states + v_out[:, :, cache_position] = value_states + # `_.zero()` followed by `+=` is equivalent `=`, but compile-friendly (without graph breaks due to assignment) + self.key_cache[layer_idx].zero_() + self.value_cache[layer_idx].zero_() + + self.key_cache[layer_idx] += k_out + self.value_cache[layer_idx] += v_out + return k_out, v_out + + def _static_update(self, cache_position, layer_idx, key_states, value_states, k_out, v_out, max_cache_len): + k_out[:, :, cache_position] = key_states + v_out[:, :, cache_position] = value_states + + self.key_cache[layer_idx] = k_out + self.value_cache[layer_idx] = v_out + return k_out, v_out + + def update( + self, + key_states: torch.Tensor, + value_states: torch.Tensor, + layer_idx: int, + cache_kwargs: Optional[Dict[str, Any]] = None, + sliding_window: Optional[int] = None, + ) -> Tuple[torch.Tensor]: + cache_position = cache_kwargs.get("cache_position") + self.key_cache[layer_idx] = self.key_cache[layer_idx].to(device=key_states.device) + self.value_cache[layer_idx] = self.value_cache[layer_idx].to(device=value_states.device) + k_out = self.key_cache[layer_idx] + v_out = self.value_cache[layer_idx] + if sliding_window: + update_fn = self._sliding_update + else: + update_fn = self._static_update + + return update_fn( + cache_position, + layer_idx, + key_states, + value_states, + k_out, + v_out, + k_out.shape[2], + ) + + def get_max_length(self) -> Optional[int]: + # in theory there is no limit because the sliding window size is fixed + # no matter how long the sentence is + return self.max_cache_len + + def get_seq_length(self, layer_idx: Optional[int] = 0) -> int: + return None + + def reset(self): + """Resets the cache values while preserving the objects""" + for layer_idx in range(len(self.key_cache)): + # In-place ops prevent breaking the static address + self.key_cache[layer_idx].zero_() + self.value_cache[layer_idx].zero_() diff --git a/src/transformers/generation/configuration_utils.py b/src/transformers/generation/configuration_utils.py index 8bb5e091d6d..8ba17a6a350 100644 --- a/src/transformers/generation/configuration_utils.py +++ b/src/transformers/generation/configuration_utils.py @@ -400,7 +400,7 @@ class GenerationConfig(PushToHubMixin): # Cache implementation self.cache_implementation = kwargs.pop("cache_implementation", None) self.cache_config = kwargs.pop("cache_config", None) - if self.cache_implementation is not None: + if self.cache_implementation is not None and self.cache_implementation in NEEDS_CACHE_CONFIG: cache_config_class = NEEDS_CACHE_CONFIG[self.cache_implementation] if self.cache_config is None: self.cache_config = cache_config_class() diff --git a/src/transformers/generation/utils.py b/src/transformers/generation/utils.py index 3abd604cdb8..2686f3af7af 100644 --- a/src/transformers/generation/utils.py +++ b/src/transformers/generation/utils.py @@ -28,6 +28,7 @@ from ..cache_utils import ( Cache, DynamicCache, HQQQuantizedCache, + HybridCache, QuantizedCacheConfig, QuantoQuantizedCache, SlidingWindowCache, @@ -112,7 +113,7 @@ logger = logging.get_logger(__name__) if is_accelerate_available(): from accelerate.hooks import AlignDevicesHook, add_hook_to_module -NEED_SETUP_CACHE_CLASSES_MAPPING = {"static": StaticCache, "sliding_window": SlidingWindowCache} +NEED_SETUP_CACHE_CLASSES_MAPPING = {"static": StaticCache, "sliding_window": SlidingWindowCache, "hybrid": HybridCache} QUANT_BACKEND_CLASSES_MAPPING = {"quanto": QuantoQuantizedCache, "HQQ": HQQQuantizedCache} @@ -1395,10 +1396,12 @@ class GenerationMixin: past_length = 0 if model_kwargs.get("past_key_values") is not None: - if isinstance(model_kwargs["past_key_values"], Cache): - past_length = model_kwargs["past_key_values"].get_seq_length() - else: - past_length = model_kwargs["past_key_values"][0][0].shape[2] + cache = model_kwargs["past_key_values"] + if not isinstance(cache, Cache): + past_length = cache[0][0].shape[2] + elif hasattr(cache, "get_seq_length"): + past_length = cache.get_seq_length() + if "inputs_embeds" in model_kwargs: cur_len = model_kwargs["inputs_embeds"].shape[1] else: @@ -1739,7 +1742,9 @@ class GenerationMixin: "issue: https://github.com/huggingface/transformers/issues/28981" ) model_kwargs["past_key_values"] = self._get_cache( - generation_config.cache_implementation, batch_size, generation_config.max_length + generation_config.cache_implementation, + getattr(generation_config, "num_beams", 1) * batch_size, + generation_config.max_length, ) elif generation_config.cache_implementation == "quantized": if not self._supports_quantized_cache: diff --git a/src/transformers/models/__init__.py b/src/transformers/models/__init__.py index 87586686a02..f4c33491472 100644 --- a/src/transformers/models/__init__.py +++ b/src/transformers/models/__init__.py @@ -92,6 +92,7 @@ from . import ( funnel, fuyu, gemma, + gemma2, git, glpn, gpt2, diff --git a/src/transformers/models/auto/configuration_auto.py b/src/transformers/models/auto/configuration_auto.py index dab9244dd01..7f52b3dc280 100755 --- a/src/transformers/models/auto/configuration_auto.py +++ b/src/transformers/models/auto/configuration_auto.py @@ -108,6 +108,7 @@ CONFIG_MAPPING_NAMES = OrderedDict( ("funnel", "FunnelConfig"), ("fuyu", "FuyuConfig"), ("gemma", "GemmaConfig"), + ("gemma2", "Gemma2Config"), ("git", "GitConfig"), ("glpn", "GLPNConfig"), ("gpt-sw3", "GPT2Config"), @@ -385,6 +386,7 @@ MODEL_NAMES_MAPPING = OrderedDict( ("funnel", "Funnel Transformer"), ("fuyu", "Fuyu"), ("gemma", "Gemma"), + ("gemma2", "Gemma2"), ("git", "GIT"), ("glpn", "GLPN"), ("gpt-sw3", "GPT-Sw3"), diff --git a/src/transformers/models/auto/modeling_auto.py b/src/transformers/models/auto/modeling_auto.py index 7190d75a873..f674b777fca 100755 --- a/src/transformers/models/auto/modeling_auto.py +++ b/src/transformers/models/auto/modeling_auto.py @@ -105,6 +105,7 @@ MODEL_MAPPING_NAMES = OrderedDict( ("fsmt", "FSMTModel"), ("funnel", ("FunnelModel", "FunnelBaseModel")), ("gemma", "GemmaModel"), + ("gemma2", "Gemma2Model"), ("git", "GitModel"), ("glpn", "GLPNModel"), ("gpt-sw3", "GPT2Model"), @@ -454,6 +455,7 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict( ("falcon", "FalconForCausalLM"), ("fuyu", "FuyuForCausalLM"), ("gemma", "GemmaForCausalLM"), + ("gemma2", "Gemma2ForCausalLM"), ("git", "GitForCausalLM"), ("gpt-sw3", "GPT2LMHeadModel"), ("gpt2", "GPT2LMHeadModel"), @@ -863,6 +865,7 @@ MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING_NAMES = OrderedDict( ("fnet", "FNetForSequenceClassification"), ("funnel", "FunnelForSequenceClassification"), ("gemma", "GemmaForSequenceClassification"), + ("gemma2", "Gemma2ForSequenceClassification"), ("gpt-sw3", "GPT2ForSequenceClassification"), ("gpt2", "GPT2ForSequenceClassification"), ("gpt_bigcode", "GPTBigCodeForSequenceClassification"), @@ -1044,6 +1047,7 @@ MODEL_FOR_TOKEN_CLASSIFICATION_MAPPING_NAMES = OrderedDict( ("fnet", "FNetForTokenClassification"), ("funnel", "FunnelForTokenClassification"), ("gemma", "GemmaForTokenClassification"), + ("gemma2", "Gemma2ForTokenClassification"), ("gpt-sw3", "GPT2ForTokenClassification"), ("gpt2", "GPT2ForTokenClassification"), ("gpt_bigcode", "GPTBigCodeForTokenClassification"), diff --git a/src/transformers/models/auto/tokenization_auto.py b/src/transformers/models/auto/tokenization_auto.py index 7dd805ae7f1..dddab5379f5 100644 --- a/src/transformers/models/auto/tokenization_auto.py +++ b/src/transformers/models/auto/tokenization_auto.py @@ -188,6 +188,13 @@ else: "GemmaTokenizerFast" if is_tokenizers_available() else None, ), ), + ( + "gemma2", + ( + "GemmaTokenizer" if is_sentencepiece_available() else None, + "GemmaTokenizerFast" if is_tokenizers_available() else None, + ), + ), ("git", ("BertTokenizer", "BertTokenizerFast" if is_tokenizers_available() else None)), ("gpt-sw3", ("GPTSw3Tokenizer" if is_sentencepiece_available() else None, None)), ("gpt2", ("GPT2Tokenizer", "GPT2TokenizerFast" if is_tokenizers_available() else None)), diff --git a/src/transformers/models/gemma/diff_gemma.py b/src/transformers/models/gemma/diff_gemma.py index 1165b05483f..d1df9d8cfb0 100644 --- a/src/transformers/models/gemma/diff_gemma.py +++ b/src/transformers/models/gemma/diff_gemma.py @@ -257,6 +257,7 @@ class GemmaAttention(nn.Module): self.max_position_embeddings = config.max_position_embeddings self.rope_theta = config.rope_theta self.is_causal = True + self.scaling = 1 / math.sqrt(config.head_dim) if self.hidden_size % self.num_heads != 0: raise ValueError( @@ -305,7 +306,7 @@ class GemmaAttention(nn.Module): key_states = repeat_kv(key_states, self.num_key_value_groups) value_states = repeat_kv(value_states, self.num_key_value_groups) - attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim) + attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) * self.scaling if attention_mask is not None: # no matter the length, we just slice it causal_mask = attention_mask[:, :, :, : key_states.shape[-2]] diff --git a/src/transformers/models/gemma/modeling_gemma.py b/src/transformers/models/gemma/modeling_gemma.py index ed6d61793bc..c0da2530fe2 100644 --- a/src/transformers/models/gemma/modeling_gemma.py +++ b/src/transformers/models/gemma/modeling_gemma.py @@ -240,6 +240,7 @@ class GemmaAttention(nn.Module): self.max_position_embeddings = config.max_position_embeddings self.rope_theta = config.rope_theta self.is_causal = True + self.scaling = 1 / math.sqrt(config.head_dim) if self.hidden_size % self.num_heads != 0: raise ValueError( @@ -288,7 +289,7 @@ class GemmaAttention(nn.Module): key_states = repeat_kv(key_states, self.num_key_value_groups) value_states = repeat_kv(value_states, self.num_key_value_groups) - attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim) + attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) * self.scaling if attention_mask is not None: # no matter the length, we just slice it causal_mask = attention_mask[:, :, :, : key_states.shape[-2]] @@ -898,6 +899,13 @@ class GemmaModel(GemmaPreTrainedModel): # See https://github.com/huggingface/transformers/pull/29402 normalizer = torch.tensor(self.config.hidden_size**0.5, dtype=hidden_states.dtype) hidden_states = hidden_states * normalizer + if use_cache and not isinstance(past_key_values, Cache): # kept for BC (non `Cache` `past_key_values` inputs) + return_legacy_cache = True + past_key_values = DynamicCache.from_legacy_cache(past_key_values) + logger.warning_once( + "We detected that you are passing `past_key_values` as a tuple and this is deprecated and will be removed in v4.43. " + "Please use an appropriate `Cache` class (https://huggingface.co/docs/transformers/v4.41.3/en/internal/generation_utils#transformers.Cache)" + ) # decoder layers all_hidden_states = () if output_hidden_states else None @@ -1397,7 +1405,7 @@ class GemmaForTokenClassification(GemmaPreTrainedModel): @add_start_docstrings_to_model_forward(GEMMA_INPUTS_DOCSTRING) def forward( self, - input_ids: torch.LongTensor = None, + input_ids: Optional[torch.LongTensor] = None, attention_mask: Optional[torch.Tensor] = None, position_ids: Optional[torch.LongTensor] = None, past_key_values: Optional[List[torch.FloatTensor]] = None, @@ -1407,7 +1415,7 @@ class GemmaForTokenClassification(GemmaPreTrainedModel): output_attentions: Optional[bool] = None, output_hidden_states: Optional[bool] = None, return_dict: Optional[bool] = None, - ) -> Union[Tuple, SequenceClassifierOutputWithPast]: + ) -> Union[Tuple, TokenClassifierOutput]: r""" labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*): Labels for computing the sequence classification/regression loss. Indices should be in `[0, ..., diff --git a/src/transformers/models/gemma2/__init__.py b/src/transformers/models/gemma2/__init__.py new file mode 100644 index 00000000000..0d0aa148be5 --- /dev/null +++ b/src/transformers/models/gemma2/__init__.py @@ -0,0 +1,61 @@ +# Copyright 2024 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 typing import TYPE_CHECKING + +from ...utils import ( + OptionalDependencyNotAvailable, + _LazyModule, + is_torch_available, +) + + +_import_structure = { + "configuration_gemma2": ["Gemma2Config"], +} + +try: + if not is_torch_available(): + raise OptionalDependencyNotAvailable() +except OptionalDependencyNotAvailable: + pass +else: + _import_structure["modeling_gemma2"] = [ + "Gemma2ForCausalLM", + "Gemma2Model", + "Gemma2PreTrainedModel", + "Gemma2ForSequenceClassification", + "Gemma2ForTokenClassification", + ] + +if TYPE_CHECKING: + from .configuration_gemma import Gemma2Config + + try: + if not is_torch_available(): + raise OptionalDependencyNotAvailable() + except OptionalDependencyNotAvailable: + pass + else: + from .modeling_gemma import ( + Gemma2ForCausalLM, + Gemma2ForSequenceClassification, + Gemma2ForTokenClassification, + Gemma2Model, + Gemma2PreTrainedModel, + ) + +else: + import sys + + sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__) diff --git a/src/transformers/models/gemma2/configuration_gemma2.py b/src/transformers/models/gemma2/configuration_gemma2.py new file mode 100644 index 00000000000..47207d7ca12 --- /dev/null +++ b/src/transformers/models/gemma2/configuration_gemma2.py @@ -0,0 +1,149 @@ +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# This file was automatically generated from . +# Do NOT edit this file manually as any edits will be overwritten by the generation of +# the file from the diff. If any change should be done, please apply the change to the +# diff.py file directly. +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# coding=utf-8 +# Copyright 2024 Google Inc. 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 transformers import PretrainedConfig + + +class Gemma2Config(PretrainedConfig): + r""" + This is the configuration class to store the configuration of a [`Gemma2Model`]. It is used to instantiate an Gemma2 + model according to the specified arguments, defining the model architecture. Instantiating a configuration with the + defaults will yield a similar configuration to that of the Gemma2-7B. + e.g. [google/gemma2-7b](https://huggingface.co/google/gemma2-7b) + 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 256000): + Vocabulary size of the Gemma2 model. Defines the number of different tokens that can be represented by the + `inputs_ids` passed when calling [`Gemma2Model`] + hidden_size (`int`, *optional*, defaults to 3072): + Dimension of the hidden representations. + intermediate_size (`int`, *optional*, defaults to 24576): + Dimension of the MLP representations. + num_hidden_layers (`int`, *optional*, defaults to 28): + Number of hidden layers in the Transformer decoder. + num_attention_heads (`int`, *optional*, defaults to 16): + Number of attention heads for each attention layer in the Transformer decoder. + num_key_value_heads (`int`, *optional*, defaults to 16): + This is the number of key_value heads that should be used to implement Grouped Query Attention. 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. When + converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed + by meanpooling all the original heads within that group. For more details checkout [this + paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to + `num_attention_heads`. + head_dim (`int`, *optional*, defaults to 256): + The attention head dimension. + hidden_activation (`str` or `function`, *optional*, defaults to `"gelu_pytorch_tanh"`): + The non-linear activation function (function or string) in the decoder. + max_position_embeddings (`int`, *optional*, defaults to 8192): + 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 used by the rms normalization layers. + use_cache (`bool`, *optional*, defaults to `True`): + Whether or not the model should return the last key/values attentions (not used by all models). Only + relevant if `config.is_decoder=True`. + pad_token_id (`int`, *optional*, defaults to 0): + Padding token id. + eos_token_id (`int`, *optional*, defaults to 1): + End of stream token id. + bos_token_id (`int`, *optional*, defaults to 2): + Beginning of stream token id. + tie_word_embeddings (`bool`, *optional*, defaults to `True`): + Whether to tie weight embeddings + rope_theta (`float`, *optional*, defaults to 10000.0): + The base period of the RoPE embeddings. + attention_bias (`bool`, defaults to `False`, *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 ratio for the attention probabilities. + final_logit_softcapping (`float`, *optional*, defaults to 30.0): scaling factor when applying tanh softcapping on the logits. + query_pre_attn_scalar (`float`, *optional*, defaults to 224): scaling factor used on the attention scores + sliding_window (`int`, *optional*, defaults to 4096): in Gemma2, every other layer uses sliding window attention. This is the + size of the sliding window. + ```python + >>> from transformers import Gemma2Model, Gemma2Config + >>> # Initializing a Gemma2 gemma2-9b style configuration + >>> configuration = Gemma2Config() + >>> # Initializing a model from the gemma2-9b style configuration + >>> model = Gemma2Model(configuration) + >>> # Accessing the model configuration + >>> configuration = model.config + ```""" + + model_type = "gemma2" + keys_to_ignore_at_inference = ["past_key_values"] + + def __init__( + self, + vocab_size=256000, + hidden_size=3072, + intermediate_size=24576, + num_hidden_layers=28, + num_attention_heads=16, + num_key_value_heads=16, + head_dim=256, + hidden_activation="gelu_pytorch_tanh", + max_position_embeddings=8192, + initializer_range=0.02, + rms_norm_eps=1e-6, + use_cache=True, + pad_token_id=0, + eos_token_id=1, + bos_token_id=2, + tie_word_embeddings=True, + rope_theta=10000.0, + attention_bias=False, + attention_dropout=0.0, + final_logit_softcapping=30.0, + query_pre_attn_scalar=224, + sliding_window=4096, + **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 + self.head_dim = head_dim + self.num_key_value_heads = num_key_value_heads + self.hidden_activation = hidden_activation + self.initializer_range = initializer_range + self.rms_norm_eps = rms_norm_eps + self.use_cache = use_cache + self.rope_theta = rope_theta + self.attention_bias = attention_bias + self.attention_dropout = attention_dropout + + 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.final_logit_softcapping = final_logit_softcapping + self.query_pre_attn_scalar = query_pre_attn_scalar + self.sliding_window = sliding_window + self.cache_implementation = "hybrid" diff --git a/src/transformers/models/gemma2/convert_gemma2_weights_to_hf.py b/src/transformers/models/gemma2/convert_gemma2_weights_to_hf.py new file mode 100644 index 00000000000..1ad7d23c3c3 --- /dev/null +++ b/src/transformers/models/gemma2/convert_gemma2_weights_to_hf.py @@ -0,0 +1,239 @@ +# Copyright 2024 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. +import argparse +import os +import warnings + +import torch +from accelerate import init_empty_weights + +from transformers import Gemma2Config, Gemma2ForCausalLM, GemmaTokenizer + + +try: + from transformers import GemmaTokenizerFast +except ImportError as e: + warnings.warn(e) + warnings.warn( + "The converted tokenizer will be the `slow` tokenizer. To use the fast, update your `tokenizers` library and re-run the tokenizer conversion" + ) + GemmaTokenizerFast = None + +""" +Sample usage: + +``` +python src/transformers/models/gemma2/convert_gemma2_weights_to_hf.py \ + --input_dir /path/to/downloaded/gemma/weights --model_size 9B --output_dir /output/path +``` + +Thereafter, models can be loaded via: + +```py +from transformers import Gemma2ForCausalLM, GemmaTokenizerFast + +model = Gemma2ForCausalLM.from_pretrained("/output/path") +tokenizer = GemmaTokenizerFast.from_pretrained("/output/path") +``` + +Important note: you need to be able to host the whole model in RAM to execute this script (even if the biggest versions +come in several checkpoints they each contain a part of each weight of the model, so we need to load them all in RAM). +""" + +gemma_9b_config = Gemma2Config( + num_hidden_layers=42, + num_attention_heads=16, + num_key_value_heads=8, + hidden_size=3584, + intermediate_size=14336, + final_logit_softcapping=30.0, + attn_logit_softcapping=50.0, + head_dim=256, + sliding_window=4096, + query_pre_attn_scalar=224, +) + +gemma_27b_config = Gemma2Config( + num_hidden_layers=46, + num_attention_heads=32, + num_key_value_heads=16, + hidden_size=4608, + intermediate_size=36864, + final_logit_softcapping=30.0, + attn_logit_softcapping=50.0, + head_dim=128, + sliding_window=4096, + query_pre_attn_scalar=144, +) + +CONFIG_MAPPING = {"9B": gemma_9b_config, "27B": gemma_27b_config} +LAYER_NAME_MAPPING = {"embedder.weight": "model.embed_tokens.weight"} + + +def write_model(save_path, input_base_path, config, safe_serialization=True, push_to_hub=False, dtype=torch.float32): + num_attn_heads = config.num_attention_heads + hidden_size = config.hidden_size + num_kv_heads = config.num_key_value_heads + head_dim = config.head_dim + + print(f"Fetching all parameters from the checkpoint at '{input_base_path}'") + + if os.path.isdir(input_base_path): + print("Model seems sharded") + + model_state_dict = {} + files = [file for file in os.listdir(input_base_path) if file.endswith(".bin")] + + for file in files: + print(file) + loaded_state_dict = torch.load(os.path.join(input_base_path, file), map_location="cpu") + model_state_dict.update(loaded_state_dict) + else: + print("Model does not seem to be sharded") + model_state_dict = torch.load(input_base_path, map_location="cpu")["model_state_dict"] + model_state_dict.pop("freqs_cis") + + state_dict = {} + for k, v in model_state_dict.items(): + if "qkv_proj" in k: + if num_kv_heads == 1: + v = v.reshape(num_attn_heads + num_kv_heads * 2, head_dim, hidden_size) + q_proj = v[:num_attn_heads, ...] + k_proj = v[num_attn_heads : num_attn_heads + num_kv_heads, ...].repeat(num_kv_heads, 1, 1) + v_proj = v[-num_kv_heads:, ...].repeat(num_kv_heads, 1, 1) + + state_dict[k.replace("qkv_proj", "q_proj")] = q_proj.reshape( + num_attn_heads * head_dim, hidden_size + ).clone() + state_dict[k.replace("qkv_proj", "k_proj")] = k_proj.reshape( + num_kv_heads * head_dim, hidden_size + ).clone() + state_dict[k.replace("qkv_proj", "v_proj")] = v_proj[0].clone() + else: + q_proj, k_proj, v_proj = torch.split( + v, [num_attn_heads * head_dim, num_kv_heads * head_dim, num_kv_heads * head_dim], 0 + ) + state_dict[k.replace("qkv_proj", "q_proj")] = q_proj.reshape( + num_attn_heads * head_dim, hidden_size + ).clone() + state_dict[k.replace("qkv_proj", "k_proj")] = k_proj.reshape( + num_kv_heads * head_dim, hidden_size + ).clone() + state_dict[k.replace("qkv_proj", "v_proj")] = v_proj.reshape( + num_kv_heads * head_dim, hidden_size + ).clone() + + elif k == "embedder.weight": + state_dict[LAYER_NAME_MAPPING[k]] = v + state_dict["lm_head.weight"] = v + else: + state_dict[k] = v + + torch.set_default_dtype(dtype) + + print("Loading the checkpoint in a Gemma2 model.") + with init_empty_weights(): + model = Gemma2ForCausalLM(config) + model.load_state_dict(state_dict, assign=True, strict=False) + + model.config.torch_dtype = torch.float32 + del model.config._name_or_path + print("Saving in the Transformers format.") + + if push_to_hub: + print(f"pushing the model to {save_path}") + model.push_to_hub(save_path, safe_serialization=safe_serialization, private=True) + else: + model.save_pretrained(save_path, safe_serialization=safe_serialization) + + +def write_tokenizer(input_tokenizer_path, save_path, push_to_hub=False): + # Initialize the tokenizer based on the `spm` model + tokenizer_class = GemmaTokenizer if GemmaTokenizerFast is None else GemmaTokenizerFast + print(f"Saving a {tokenizer_class.__name__} to {save_path}.") + tokenizer = tokenizer_class(input_tokenizer_path) + if push_to_hub: + tokenizer.push_to_hub(save_path) + else: + tokenizer.save_pretrained(save_path) + + +def main(): + parser = argparse.ArgumentParser() + parser.add_argument( + "--input_checkpoint", + help="Absolute path to the target Gemma2 weights.", + required=True, + ) + parser.add_argument( + "--tokenizer_checkpoint", + help="Location of Gemma2 tokenizer model", + ) + parser.add_argument( + "--model_size", + default="9B", + choices=["9B", "27B", "tokenizer_only"], + help="'f' models correspond to the finetuned versions, and are specific to the Gemma22 official release. For more details on Gemma2, checkout the original repo: https://huggingface.co/google/gemma-7b", + ) + parser.add_argument( + "--output_dir", + default="google/gemma-9b", + help="Location to write HF model and tokenizer", + ) + parser.add_argument( + "--pickle_serialization", + help="Whether or not to save using `safetensors`.", + action="store_true", + default=False, + ) + parser.add_argument( + "--convert_tokenizer", + help="Whether or not to convert the tokenizer as well.", + action="store_true", + default=False, + ) + parser.add_argument( + "--push_to_hub", + help="Whether or not to push the model to the hub at `output_dir` instead of saving it locally.", + action="store_true", + default=False, + ) + parser.add_argument( + "--dtype", + default="float32", + help="Target dtype of the converted model", + ) + args = parser.parse_args() + + if args.convert_tokenizer: + if args.tokenizer_checkpoint is None: + raise ValueError("Path to the tokenizer is required when passing --convert_tokenizer") + + spm_path = os.path.join(args.tokenizer_checkpoint) + write_tokenizer(spm_path, args.output_dir, args.push_to_hub) + if not args.model_size == "tokenizer_only": + config = CONFIG_MAPPING[args.model_size] + dtype = getattr(torch, args.dtype) + write_model( + config=config, + input_base_path=args.input_checkpoint, + save_path=args.output_dir, + safe_serialization=not args.pickle_serialization, + push_to_hub=args.push_to_hub, + dtype=dtype, + ) + + +if __name__ == "__main__": + main() diff --git a/src/transformers/models/gemma2/diff_gemma2.py b/src/transformers/models/gemma2/diff_gemma2.py new file mode 100644 index 00000000000..443be0cf87f --- /dev/null +++ b/src/transformers/models/gemma2/diff_gemma2.py @@ -0,0 +1,781 @@ +# coding=utf-8 +# Copyright 2024 Google Inc. 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 inspect +from typing import List, Optional, Tuple, Union + +import torch +import torch.nn.functional as F +import torch.utils.checkpoint +from torch.nn import CrossEntropyLoss + +from transformers.models.gemma.configuration_gemma import GemmaConfig +from transformers.models.gemma.modeling_gemma import ( + GemmaAttention, + GemmaDecoderLayer, + GemmaForCausalLM, + GemmaForSequenceClassification, + GemmaForTokenClassification, + GemmaModel, + GemmaRMSNorm, + apply_rotary_pos_emb, + repeat_kv, +) + +from ...cache_utils import Cache +from ...modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast +from ...utils import is_flash_attn_2_available, is_flash_attn_greater_or_equal_2_10, logging + + +if is_flash_attn_2_available(): + from flash_attn import flash_attn_func, flash_attn_varlen_func + from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input # noqa + + _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters) + + +logger = logging.get_logger(__name__) + + +def _get_unpad_data(attention_mask): + seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32) + indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten() + max_seqlen_in_batch = seqlens_in_batch.max().item() + cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0)) + return ( + indices, + cu_seqlens, + max_seqlen_in_batch, + ) + + +logger = logging.get_logger(__name__) + + +class Gemma2Config(GemmaConfig): + cache_implementation = "hybrid" # TODO this is not properly ported, but cls attr is better + + def __init__( + self, + query_pre_attn_scalar=224, + sliding_window=4096, + final_logit_softcapping=30.0, + **super_kwargs, + ): + super().__init__(self, **super_kwargs) + self.query_pre_attn_scalar = query_pre_attn_scalar + self.sliding_window = sliding_window + self.cache_implementation = "hybrid" + self.final_logit_softcapping = final_logit_softcapping + + +class Gemma2RMSNorm(GemmaRMSNorm): + pass + + +class Gemma2Attention(GemmaAttention): + """Multi-headed attention from 'Attention Is All You Need' paper""" + + def __init__(self, config: Gemma2Config, layer_idx: Optional[int] = None): + self.scaling = config.query_pre_attn_scalar**-0.5 + + super().__init__(config, layer_idx) + + +class Gemma2FlashAttention2(Gemma2Attention): + """ + Gemma2 flash attention module. This module inherits from `Gemma2Attention` as the weights of the module stays + untouched. The only required change would be on the forward pass where it needs to correctly call the public API of + flash attention and deal with padding tokens in case the input contains any of them. + """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1. + # flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0. + # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left). + self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10() + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.LongTensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_value: Optional[Cache] = None, + output_attentions: bool = False, + use_cache: bool = False, + cache_position: Optional[torch.LongTensor] = None, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: + output_attentions = False + + bsz, q_len, _ = hidden_states.size() + + query_states = self.q_proj(hidden_states) + key_states = self.k_proj(hidden_states) + value_states = self.v_proj(hidden_states) + + # Flash attention requires the input to have the shape + # batch_size x seq_length x head_dim x hidden_dim + # therefore we just need to keep the original shape + query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) + key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + + cos, sin = self.rotary_emb(value_states, position_ids) + query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin) + + if past_key_value is not None: + # sin and cos are specific to RoPE models; cache_position needed for the static cache + cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position} + key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs) + + # TODO: These transpose are quite inefficient but Flash Attention requires the layout [batch_size, sequence_length, num_heads, head_dim]. We would need to refactor the KV cache + # to be able to avoid many of these transpose/reshape/view. + query_states = query_states.transpose(1, 2) + key_states = key_states.transpose(1, 2) + value_states = value_states.transpose(1, 2) + + dropout_rate = self.attention_dropout if self.training else 0.0 + + # In PEFT, usually we cast the layer norms in float32 for training stability reasons + # therefore the input hidden states gets silently casted in float32. Hence, we need + # cast them back in the correct dtype just to be sure everything works as expected. + # This might slowdown training & inference so it is recommended to not cast the LayerNorms + # in fp32. (Gemma2RMSNorm handles it correctly) + + input_dtype = query_states.dtype + if input_dtype == torch.float32: + if torch.is_autocast_enabled(): + target_dtype = torch.get_autocast_gpu_dtype() + # Handle the case where the model is quantized + elif hasattr(self.config, "_pre_quantization_dtype"): + target_dtype = self.config._pre_quantization_dtype + else: + target_dtype = self.q_proj.weight.dtype + + logger.warning_once( + f"The input hidden states seems to be silently casted in float32, this might be related to" + f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in" + f" {target_dtype}." + ) + + query_states = query_states.to(target_dtype) + key_states = key_states.to(target_dtype) + value_states = value_states.to(target_dtype) + + ########### ONLY DIFFERENCE IS WE USE SLIDING AND PASS THE SOFTMAX SCALING + attn_output = self._flash_attention_forward( + query_states, + key_states, + value_states, + attention_mask, + q_len, + dropout=dropout_rate, + softmax_scale=self.scaling, + ) + + attn_output = attn_output.reshape(bsz, q_len, -1).contiguous() + attn_output = self.o_proj(attn_output) + + if not output_attentions: + attn_weights = None + + return attn_output, attn_weights, past_key_value + + def _flash_attention_forward( + self, + query_states, + key_states, + value_states, + attention_mask, + query_length, + dropout=0.0, + softmax_scale=None, + cache_position=0, + ): + """ + Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token + first unpad the input, then computes the attention scores and pad the final attention scores. + + Args: + query_states (`torch.Tensor`): + Input query states to be passed to Flash Attention API + key_states (`torch.Tensor`): + Input key states to be passed to Flash Attention API + value_states (`torch.Tensor`): + Input value states to be passed to Flash Attention API + attention_mask (`torch.Tensor`): + The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the + position of padding tokens and 1 for the position of non-padding tokens. + dropout (`float`): + Attention dropout + softmax_scale (`float`, *optional*): + The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim) + """ + if not self._flash_attn_uses_top_left_mask: + causal = self.is_causal + else: + # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in Gemma2FlashAttention2 __init__. + causal = self.is_causal and query_length != 1 + + use_sliding_windows = ( + _flash_supports_window_size and self.sliding_window is not None and cache_position > self.sliding_window + ) + flash_kwargs = {"window_size": (self.sliding_window, self.sliding_window)} if use_sliding_windows else {} + # Contains at least one padding token in the sequence + if attention_mask is not None: + batch_size = query_states.shape[0] + query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input( + query_states, key_states, value_states, attention_mask, query_length + ) + + cu_seqlens_q, cu_seqlens_k = cu_seq_lens + max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens + + attn_output_unpad = flash_attn_varlen_func( + query_states, + key_states, + value_states, + cu_seqlens_q=cu_seqlens_q, + cu_seqlens_k=cu_seqlens_k, + max_seqlen_q=max_seqlen_in_batch_q, + max_seqlen_k=max_seqlen_in_batch_k, + dropout_p=dropout, + softmax_scale=softmax_scale, + causal=causal, + **flash_kwargs, + ) + + attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length) + else: + attn_output = flash_attn_func( + query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=causal + ) + + return attn_output + + def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length): + indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask) + batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape + + key_layer = index_first_axis( + key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k + ) + value_layer = index_first_axis( + value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k + ) + if query_length == kv_seq_len: + query_layer = index_first_axis( + query_layer.reshape(batch_size * kv_seq_len, self.num_heads, head_dim), indices_k + ) + cu_seqlens_q = cu_seqlens_k + max_seqlen_in_batch_q = max_seqlen_in_batch_k + indices_q = indices_k + elif query_length == 1: + max_seqlen_in_batch_q = 1 + cu_seqlens_q = torch.arange( + batch_size + 1, dtype=torch.int32, device=query_layer.device + ) # There is a memcpy here, that is very bad. + indices_q = cu_seqlens_q[:-1] + query_layer = query_layer.squeeze(1) + else: + # The -q_len: slice assumes left padding. + attention_mask = attention_mask[:, -query_length:] + query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask) + + return ( + query_layer, + key_layer, + value_layer, + indices_q, + (cu_seqlens_q, cu_seqlens_k), + (max_seqlen_in_batch_q, max_seqlen_in_batch_k), + ) + + +class Gemma2SdpaAttention(Gemma2Attention): + """ + Gemma2 attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from + `Gemma2Attention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to + SDPA API. + """ + + # Adapted from Gemma2Attention.forward + 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, + output_attentions: bool = False, + use_cache: bool = False, + cache_position: Optional[torch.LongTensor] = None, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: + if output_attentions: + # TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented. + logger.warning_once( + "Gemma2Model is using Gemma2SdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, " + 'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.' + ) + return super().forward( + hidden_states=hidden_states, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_value=past_key_value, + output_attentions=output_attentions, + use_cache=use_cache, + cache_position=cache_position, + ) + + bsz, q_len, _ = hidden_states.size() + + query_states = self.q_proj(hidden_states) + key_states = self.k_proj(hidden_states) + value_states = self.v_proj(hidden_states) + + query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) + key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + + cos, sin = self.rotary_emb(value_states, position_ids) + query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin) + + if past_key_value is not None: + # sin and cos are specific to RoPE models; cache_position needed for the static cache + cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position} + key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs) + + key_states = repeat_kv(key_states, self.num_key_value_groups) + value_states = repeat_kv(value_states, self.num_key_value_groups) + + causal_mask = attention_mask + if attention_mask is not None: + causal_mask = causal_mask[:, :, :, : key_states.shape[-2]] + + # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask, + # Reference: https://github.com/pytorch/pytorch/issues/112577. + if query_states.device.type == "cuda" and causal_mask is not None: + query_states = query_states.contiguous() + key_states = key_states.contiguous() + value_states = value_states.contiguous() + + # We dispatch to SDPA's Flash Attention or Efficient kernels via this `is_causal` if statement instead of an inline conditional assignment + # in SDPA to support both torch.compile's dynamic shapes and full graph options. An inline conditional prevents dynamic shapes from compiling. + is_causal = True if causal_mask is None and q_len > 1 else False + + attn_output = torch.nn.functional.scaled_dot_product_attention( + query_states, + key_states, + value_states, + attn_mask=causal_mask, + dropout_p=self.attention_dropout if self.training else 0.0, + is_causal=is_causal, + scale=self.scaling, + ) + + attn_output = attn_output.transpose(1, 2).contiguous() + attn_output = attn_output.view(bsz, q_len, -1) + + attn_output = self.o_proj(attn_output) + + return attn_output, None, past_key_value + + +class Gemma2DecoderLayer(GemmaDecoderLayer): + def __init__(self, config: Gemma2Config, layer_idx: int): + super().__init__(config, layer_idx) + + self.is_sliding = bool(layer_idx % 2) + self.pre_feedforward_layernorm = Gemma2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.post_feedforward_layernorm = Gemma2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.sliding_window = config.sliding_window + + 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, + output_attentions: Optional[bool] = False, + use_cache: Optional[bool] = False, + cache_position: Optional[torch.LongTensor] = None, + ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]: + if self.is_sliding and attention_mask is not None: # efficient SDPA and no padding + attention_mask = attention_mask * torch.tril( + torch.ones_like(attention_mask), diagonal=(self.sliding_window - cache_position[-1]) + ) + if cache_position[0] > 0: + attention_mask = attention_mask[:, -self.sliding_window :] + + residual = hidden_states + + hidden_states = self.input_layernorm(hidden_states) + + # Self Attention + hidden_states, self_attn_weights, present_key_value = self.self_attn( + hidden_states=hidden_states, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_value=past_key_value, + output_attentions=output_attentions, + use_cache=use_cache, + cache_position=cache_position, + ) + hidden_states = self.post_attention_layernorm(hidden_states) + hidden_states = residual + hidden_states + + residual = hidden_states + hidden_states = self.pre_feedforward_layernorm(hidden_states) + hidden_states = self.mlp(hidden_states) + hidden_states = self.post_feedforward_layernorm(hidden_states) + hidden_states = residual + hidden_states + + outputs = (hidden_states,) + + if output_attentions: + outputs += (self_attn_weights,) + + if use_cache: + outputs += (present_key_value,) + + return outputs + + +class Gemma2Model(GemmaModel): + def forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + cache_position: Optional[torch.LongTensor] = None, + ) -> Union[Tuple, BaseModelOutputWithPast]: + output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + use_cache = use_cache if use_cache is not None else self.config.use_cache + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + if (input_ids is None) ^ (inputs_embeds is not None): + raise ValueError( + "You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one" + ) + + if self.gradient_checkpointing and self.training and use_cache: + logger.warning_once( + "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`." + ) + use_cache = False + + if inputs_embeds is None: + inputs_embeds = self.embed_tokens(input_ids) + + if cache_position is None: + cache_position = torch.arange(0, inputs_embeds.shape[1], device=inputs_embeds.device) + + if position_ids is None: + position_ids = cache_position.unsqueeze(0) + + causal_mask = self._update_causal_mask( + attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions + ) + + # embed positions + hidden_states = inputs_embeds + + # normalized + # Gemma2 downcasts the below to float16, causing sqrt(3072)=55.4256 to become 55.5 + # See https://github.com/huggingface/transformers/pull/29402 + normalizer = torch.tensor(self.config.hidden_size**0.5, dtype=hidden_states.dtype) + hidden_states = hidden_states * normalizer + + # decoder layers + all_hidden_states = () if output_hidden_states else None + all_self_attns = () if output_attentions else None + + for decoder_layer in self.layers: + if output_hidden_states: + all_hidden_states += (hidden_states,) + + if self.gradient_checkpointing and self.training: + layer_outputs = self._gradient_checkpointing_func( + decoder_layer.__call__, + hidden_states, + causal_mask, + position_ids, + past_key_values, + output_attentions, + use_cache, + cache_position, + ) + else: + layer_outputs = decoder_layer( + hidden_states, + attention_mask=causal_mask, + position_ids=position_ids, + past_key_value=past_key_values, + output_attentions=output_attentions, + use_cache=use_cache, + cache_position=cache_position, + ) + + hidden_states = layer_outputs[0] + + if output_attentions: + all_self_attns += (layer_outputs[1],) + + hidden_states = self.norm(hidden_states) + + if output_hidden_states: + all_hidden_states += (hidden_states,) + + next_cache = past_key_values if use_cache else None + + if not return_dict: + return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None) + return BaseModelOutputWithPast( + last_hidden_state=hidden_states, + past_key_values=next_cache, + hidden_states=all_hidden_states, + attentions=all_self_attns, + ) + + @torch.no_grad() + def _update_causal_mask( + self, + attention_mask: torch.Tensor, + input_tensor: torch.Tensor, + cache_position: torch.Tensor, + past_key_values: Cache, + output_attentions: bool, + ): + if self.config._attn_implementation == "flash_attention_2": + if attention_mask is not None and 0.0 in attention_mask: + return attention_mask + return None + + dtype, device = input_tensor.dtype, input_tensor.device + min_dtype = torch.finfo(dtype).min + sequence_length = input_tensor.shape[1] + if past_key_values is not None: + target_length = past_key_values.get_max_length() + else: + target_length = attention_mask.shape[-1] + + if attention_mask is not None and attention_mask.dim() == 4: + # in this case we assume that the mask comes already in inverted form and requires no inversion or slicing + if attention_mask.max() != 0: + raise ValueError("Custom 4D attention mask should be passed in inverted form with max==0`") + causal_mask = attention_mask + else: + causal_mask = torch.full( + (sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=device + ) + if sequence_length != 1: + causal_mask = torch.triu(causal_mask, diagonal=1) + causal_mask *= torch.arange(target_length, device=device) > cache_position.reshape(-1, 1) + causal_mask = causal_mask[None, None, :, :].expand(input_tensor.shape[0], 1, -1, -1) + if attention_mask is not None: + causal_mask = causal_mask.clone() # copy to contiguous memory for in-place edit + mask_length = attention_mask.shape[-1] + padding_mask = causal_mask[:, :, :, :mask_length] + attention_mask[:, None, None, :] + padding_mask = padding_mask == 0 + causal_mask[:, :, :, :mask_length] = causal_mask[:, :, :, :mask_length].masked_fill( + padding_mask, min_dtype + ) + return causal_mask + + +class Gemma2ForCausalLM(GemmaForCausalLM): + def forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + cache_position: Optional[torch.LongTensor] = None, + ) -> Union[Tuple, CausalLMOutputWithPast]: + r""" + Args: + labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Labels for computing the masked language modeling loss. Indices should either be in `[0, ..., + config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored + (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. + + Returns: + + Example: + + ```python + >>> from transformers import AutoTokenizer, GemmaForCausalLM + + >>> model = GemmaForCausalLM.from_pretrained("google/gemma-2-9b") + >>> tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-9b") + + >>> prompt = "What is your favorite condiment?" + >>> 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] + "What is your favorite condiment?" + ```""" + output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn) + outputs = self.model( + input_ids=input_ids, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_values=past_key_values, + inputs_embeds=inputs_embeds, + use_cache=use_cache, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + cache_position=cache_position, + ) + + hidden_states = outputs[0] + logits = self.lm_head(hidden_states) + if self.config.final_logit_softcapping is not None: + logits = logits / self.config.final_logit_softcapping + logits = torch.tanh(logits) + logits = logits * self.config.final_logit_softcapping + + logits = logits.float() + loss = None + if labels is not None: + # Shift so that tokens < n predict n + shift_logits = logits[..., :-1, :].contiguous() + shift_labels = labels[..., 1:].contiguous() + # Flatten the tokens + loss_fct = CrossEntropyLoss() + shift_logits = shift_logits.view(-1, self.config.vocab_size) + shift_labels = shift_labels.view(-1) + # Enable model parallelism + shift_labels = shift_labels.to(shift_logits.device) + loss = loss_fct(shift_logits, shift_labels) + + if not return_dict: + output = (logits,) + outputs[1:] + return (loss,) + output if loss is not None else output + + return CausalLMOutputWithPast( + loss=loss, + logits=logits, + past_key_values=outputs.past_key_values, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + def prepare_inputs_for_generation( + self, + input_ids, + past_key_values=None, + attention_mask=None, + inputs_embeds=None, + cache_position=None, + use_cache=True, + **kwargs, + ): + past_length = 0 + if past_key_values is not None: + # Past key values are always initialized with a `Cache` object -> no need for if-else anymore + past_length = cache_position[0] if cache_position is not None else torch.tensor(0, device=input_ids.device) + max_cache_length = ( + torch.tensor(past_key_values.get_max_length(), device=input_ids.device) + if past_key_values.get_max_length() is not None + else None + ) + cache_length = past_length if max_cache_length is None else torch.min(max_cache_length, past_length) + + # Keep only the unprocessed tokens: + # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where + # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as input) + if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]: + input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :] + # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard + # input_ids based on the past_length. + elif past_length < input_ids.shape[1]: + input_ids = input_ids[:, past_length:] + # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens. + + # If we are about to go beyond the maximum cache length, we need to crop the input attention mask. + if ( + max_cache_length is not None + and attention_mask is not None + and cache_length + input_ids.shape[1] > max_cache_length + ): + attention_mask = attention_mask[:, -max_cache_length:] + + position_ids = kwargs.get("position_ids", None) + if attention_mask is not None and position_ids is None: + # create position_ids on the fly for batch generation + position_ids = attention_mask.long().cumsum(-1) - 1 + position_ids.masked_fill_(attention_mask == 0, 1) + if past_key_values: + position_ids = position_ids[:, -input_ids.shape[1] :] + + # if `inputs_embeds` are passed, we only want to use them in the 1st generation step + if inputs_embeds is not None and past_length == 0: + model_inputs = {"inputs_embeds": inputs_embeds} + else: + # The `contiguous()` here is necessary to have a static stride during decoding. torchdynamo otherwise + # recompiles graphs as the stride of the inputs is a guard. Ref: https://github.com/huggingface/transformers/pull/29114 + # TODO: use `next_tokens` directly instead. + model_inputs = {"input_ids": input_ids.contiguous()} + + input_length = position_ids.shape[-1] if position_ids is not None else input_ids.shape[-1] + if cache_position is None: + cache_position = torch.arange(past_length, past_length + input_length, device=input_ids.device) + elif use_cache: + cache_position = cache_position[-input_length:] + + model_inputs.update( + { + "position_ids": position_ids, + "cache_position": cache_position, + "past_key_values": past_key_values, + "use_cache": use_cache, + "attention_mask": attention_mask, + } + ) + return model_inputs + + +class Gemma2ForSequenceClassification(GemmaForSequenceClassification): + pass + + +class Gemma2ForTokenClassification(GemmaForTokenClassification): + pass diff --git a/src/transformers/models/gemma2/modeling_gemma2.py b/src/transformers/models/gemma2/modeling_gemma2.py new file mode 100644 index 00000000000..2f4768e59f4 --- /dev/null +++ b/src/transformers/models/gemma2/modeling_gemma2.py @@ -0,0 +1,1392 @@ +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# This file was automatically generated from . +# Do NOT edit this file manually as any edits will be overwritten by the generation of +# the file from the diff. If any change should be done, please apply the change to the +# diff.py file directly. +# 🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨 +# coding=utf-8 +# Copyright 2024 Google Inc. 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 inspect +from typing import List, Optional, Tuple, Union + +import torch +import torch.nn.functional as F +import torch.utils.checkpoint +from torch import nn +from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss + +from ...activations import ACT2FN +from ...cache_utils import Cache +from ...modeling_outputs import ( + BaseModelOutputWithPast, + CausalLMOutputWithPast, + SequenceClassifierOutputWithPast, + TokenClassifierOutput, +) +from ...modeling_utils import PreTrainedModel +from ...utils import ( + add_start_docstrings, + add_start_docstrings_to_model_forward, + is_flash_attn_2_available, + is_flash_attn_greater_or_equal_2_10, + logging, + replace_return_docstrings, +) +from .configuration_gemma2 import Gemma2Config + + +if is_flash_attn_2_available(): + from flash_attn import flash_attn_func, flash_attn_varlen_func + from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input # noqa + + _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters) + + +logger = logging.get_logger(__name__) + + +def _get_unpad_data(attention_mask): + seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32) + indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten() + max_seqlen_in_batch = seqlens_in_batch.max().item() + cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0)) + return ( + indices, + cu_seqlens, + max_seqlen_in_batch, + ) + + +class Gemma2RMSNorm(nn.Module): + def __init__(self, dim: int, eps: float = 1e-6): + super().__init__() + self.eps = eps + self.weight = nn.Parameter(torch.zeros(dim)) + + def _norm(self, x): + return x * torch.rsqrt(x.pow(2).mean(-1, keepdim=True) + self.eps) + + def forward(self, x): + output = self._norm(x.float()) + # Llama does x.to(float16) * w whilst Gemma2 is (x * w).to(float16) + # See https://github.com/huggingface/transformers/pull/29402 + output = output * (1.0 + self.weight.float()) + return output.type_as(x) + + +class Gemma2RotaryEmbedding(nn.Module): + def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None): + super().__init__() + + self.dim = dim + self.max_position_embeddings = max_position_embeddings + self.base = base + inv_freq = 1.0 / (self.base ** (torch.arange(0, self.dim, 2, dtype=torch.int64).float() / self.dim)) + self.register_buffer("inv_freq", tensor=inv_freq, persistent=False) + + @torch.no_grad() + def forward(self, x, position_ids, seq_len=None): + # x: [bs, num_attention_heads, seq_len, head_size] + self.inv_freq.to(x.device) + inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1) + position_ids_expanded = position_ids[:, None, :].float() + # Force float32 since bfloat16 loses precision on long contexts + # See https://github.com/huggingface/transformers/pull/29285 + device_type = x.device.type + device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu" + with torch.autocast(device_type=device_type, enabled=False): + freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2) + emb = torch.cat((freqs, freqs), dim=-1) + cos = emb.cos() + sin = emb.sin() + return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype) + + +def rotate_half(x): + """Rotates half the hidden dims of the input.""" + x1 = x[..., : x.shape[-1] // 2] + x2 = x[..., x.shape[-1] // 2 :] + return torch.cat((-x2, x1), dim=-1) + + +def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1): + """Applies Rotary Position Embedding to the query and key tensors. + + Args: + q (`torch.Tensor`): The query tensor. + k (`torch.Tensor`): The key tensor. + cos (`torch.Tensor`): The cosine part of the rotary embedding. + sin (`torch.Tensor`): The sine part of the rotary embedding. + position_ids (`torch.Tensor`, *optional*): + Deprecated and unused. + unsqueeze_dim (`int`, *optional*, defaults to 1): + The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and + sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note + that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and + k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes + cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have + the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2. + Returns: + `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding. + """ + cos = cos.unsqueeze(unsqueeze_dim) + sin = sin.unsqueeze(unsqueeze_dim) + q_embed = (q * cos) + (rotate_half(q) * sin) + k_embed = (k * cos) + (rotate_half(k) * sin) + return q_embed, k_embed + + +class Gemma2MLP(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_activation] + + def forward(self, x): + return self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x)) + + +def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor: + """ + This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch, + num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim) + """ + batch, num_key_value_heads, slen, head_dim = hidden_states.shape + if n_rep == 1: + return hidden_states + hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim) + return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim) + + +class Gemma2Attention(nn.Module): + """Multi-headed attention from 'Attention Is All You Need' paper""" + + def __init__(self, config: Gemma2Config, layer_idx: Optional[int] = None): + super().__init__() + self.config = config + self.layer_idx = layer_idx + if layer_idx is None: + logger.warning_once( + f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will " + "lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` " + "when creating this class." + ) + + 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.num_key_value_heads = config.num_key_value_heads + self.num_key_value_groups = self.num_heads // self.num_key_value_heads + self.max_position_embeddings = config.max_position_embeddings + self.rope_theta = config.rope_theta + self.is_causal = True + self.scaling = config.query_pre_attn_scalar**-0.5 + + if self.hidden_size % self.num_heads != 0: + raise ValueError( + f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}" + f" and `num_heads`: {self.num_heads})." + ) + + self.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=config.attention_bias) + self.k_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias) + self.v_proj = nn.Linear(self.hidden_size, self.num_key_value_heads * self.head_dim, bias=config.attention_bias) + self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=config.attention_bias) + self.rotary_emb = Gemma2RotaryEmbedding( + self.head_dim, + max_position_embeddings=self.max_position_embeddings, + base=self.rope_theta, + ) + self.sliding_window = config.sliding_window if layer_idx % 2 else None + + 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, + output_attentions: bool = False, + use_cache: bool = False, + cache_position: Optional[torch.LongTensor] = None, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: + bsz, q_len, _ = hidden_states.size() + + query_states = self.q_proj(hidden_states) + key_states = self.k_proj(hidden_states) + value_states = self.v_proj(hidden_states) + + query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) + key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + + cos, sin = self.rotary_emb(value_states, position_ids) + query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin) + + if past_key_value is not None: + # sin and cos are specific to RoPE models; cache_position needed for the static cache + cache_kwargs = { + "sin": sin, + "cos": cos, + "sliding_window": self.sliding_window, + "cache_position": cache_position, + } + key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs) + + key_states = repeat_kv(key_states, self.num_key_value_groups) + value_states = repeat_kv(value_states, self.num_key_value_groups) + + attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) * self.scaling + + if attention_mask is not None: # no matter the length, we just slice it + causal_mask = attention_mask[:, :, :, : key_states.shape[-2]] + attn_weights = attn_weights + causal_mask + + # upcast attention to fp32 + attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype) + attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout, training=self.training) + attn_output = torch.matmul(attn_weights, value_states) + + if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim): + raise ValueError( + f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is" + f" {attn_output.size()}" + ) + + attn_output = attn_output.transpose(1, 2).contiguous() + + attn_output = attn_output.view(bsz, q_len, -1) + attn_output = self.o_proj(attn_output) + + if not output_attentions: + attn_weights = None + + return attn_output, attn_weights, past_key_value + + +class Gemma2FlashAttention2(Gemma2Attention): + """ + Gemma2 flash attention module. This module inherits from `Gemma2Attention` as the weights of the module stays + untouched. The only required change would be on the forward pass where it needs to correctly call the public API of + flash attention and deal with padding tokens in case the input contains any of them. + """ + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1. + # flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0. + # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left). + self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10() + + def forward( + self, + hidden_states: torch.Tensor, + attention_mask: Optional[torch.LongTensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_value: Optional[Cache] = None, + output_attentions: bool = False, + use_cache: bool = False, + cache_position: Optional[torch.LongTensor] = None, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: + output_attentions = False + + bsz, q_len, _ = hidden_states.size() + + query_states = self.q_proj(hidden_states) + key_states = self.k_proj(hidden_states) + value_states = self.v_proj(hidden_states) + + # Flash attention requires the input to have the shape + # batch_size x seq_length x head_dim x hidden_dim + # therefore we just need to keep the original shape + query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) + key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + + cos, sin = self.rotary_emb(value_states, position_ids) + query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin) + + if past_key_value is not None: + # sin and cos are specific to RoPE models; cache_position needed for the static cache + cache_kwargs = { + "sin": sin, + "cos": cos, + "sliding_window": self.sliding_window, + "cache_position": cache_position, + } + key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs) + + # TODO: These transpose are quite inefficient but Flash Attention requires the layout [batch_size, sequence_length, num_heads, head_dim]. We would need to refactor the KV cache + # to be able to avoid many of these transpose/reshape/view. + query_states = query_states.transpose(1, 2) + key_states = key_states.transpose(1, 2) + value_states = value_states.transpose(1, 2) + + dropout_rate = self.attention_dropout if self.training else 0.0 + + # In PEFT, usually we cast the layer norms in float32 for training stability reasons + # therefore the input hidden states gets silently casted in float32. Hence, we need + # cast them back in the correct dtype just to be sure everything works as expected. + # This might slowdown training & inference so it is recommended to not cast the LayerNorms + # in fp32. (Gemma2RMSNorm handles it correctly) + + input_dtype = query_states.dtype + if input_dtype == torch.float32: + if torch.is_autocast_enabled(): + target_dtype = torch.get_autocast_gpu_dtype() + # Handle the case where the model is quantized + elif hasattr(self.config, "_pre_quantization_dtype"): + target_dtype = self.config._pre_quantization_dtype + else: + target_dtype = self.q_proj.weight.dtype + + logger.warning_once( + f"The input hidden states seems to be silently casted in float32, this might be related to" + f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in" + f" {target_dtype}." + ) + + query_states = query_states.to(target_dtype) + key_states = key_states.to(target_dtype) + value_states = value_states.to(target_dtype) + + attn_output = self._flash_attention_forward( + query_states, + key_states, + value_states, + attention_mask, + q_len, + dropout=dropout_rate, + softmax_scale=self.scaling, + ) + + attn_output = attn_output.reshape(bsz, q_len, -1).contiguous() + attn_output = self.o_proj(attn_output) + + if not output_attentions: + attn_weights = None + + return attn_output, attn_weights, past_key_value + + def _flash_attention_forward( + self, + query_states, + key_states, + value_states, + attention_mask, + query_length, + dropout=0.0, + softmax_scale=None, + cache_position=0, + ): + """ + Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token + first unpad the input, then computes the attention scores and pad the final attention scores. + + Args: + query_states (`torch.Tensor`): + Input query states to be passed to Flash Attention API + key_states (`torch.Tensor`): + Input key states to be passed to Flash Attention API + value_states (`torch.Tensor`): + Input value states to be passed to Flash Attention API + attention_mask (`torch.Tensor`): + The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the + position of padding tokens and 1 for the position of non-padding tokens. + dropout (`float`): + Attention dropout + softmax_scale (`float`, *optional*): + The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim) + """ + if not self._flash_attn_uses_top_left_mask: + causal = self.is_causal + else: + # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in Gemma2FlashAttention2 __init__. + causal = self.is_causal and query_length != 1 + + # TODO this is not compile compatible + use_sliding_windows = ( + _flash_supports_window_size and self.sliding_window is not None and cache_position > self.sliding_window + ) + flash_kwargs = {"window_size": (self.sliding_window, self.sliding_window)} if use_sliding_windows else {} + # Contains at least one padding token in the sequence + if attention_mask is not None: + batch_size = query_states.shape[0] + query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = self._upad_input( + query_states, key_states, value_states, attention_mask, query_length + ) + + cu_seqlens_q, cu_seqlens_k = cu_seq_lens + max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens + + attn_output_unpad = flash_attn_varlen_func( + query_states, + key_states, + value_states, + cu_seqlens_q=cu_seqlens_q, + cu_seqlens_k=cu_seqlens_k, + max_seqlen_q=max_seqlen_in_batch_q, + max_seqlen_k=max_seqlen_in_batch_k, + dropout_p=dropout, + softmax_scale=softmax_scale, + causal=causal, + **flash_kwargs, + ) + + attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length) + else: + attn_output = flash_attn_func( + query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=causal + ) + + return attn_output + + def _upad_input(self, query_layer, key_layer, value_layer, attention_mask, query_length): + indices_k, cu_seqlens_k, max_seqlen_in_batch_k = _get_unpad_data(attention_mask) + batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape + + key_layer = index_first_axis( + key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k + ) + value_layer = index_first_axis( + value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k + ) + if query_length == kv_seq_len: + query_layer = index_first_axis( + query_layer.reshape(batch_size * kv_seq_len, self.num_heads, head_dim), indices_k + ) + cu_seqlens_q = cu_seqlens_k + max_seqlen_in_batch_q = max_seqlen_in_batch_k + indices_q = indices_k + elif query_length == 1: + max_seqlen_in_batch_q = 1 + cu_seqlens_q = torch.arange( + batch_size + 1, dtype=torch.int32, device=query_layer.device + ) # There is a memcpy here, that is very bad. + indices_q = cu_seqlens_q[:-1] + query_layer = query_layer.squeeze(1) + else: + # The -q_len: slice assumes left padding. + attention_mask = attention_mask[:, -query_length:] + query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask) + + return ( + query_layer, + key_layer, + value_layer, + indices_q, + (cu_seqlens_q, cu_seqlens_k), + (max_seqlen_in_batch_q, max_seqlen_in_batch_k), + ) + + +class Gemma2SdpaAttention(Gemma2Attention): + """ + Gemma2 attention module using torch.nn.functional.scaled_dot_product_attention. This module inherits from + `Gemma2Attention` as the weights of the module stays untouched. The only changes are on the forward pass to adapt to + SDPA API. + """ + + # Adapted from Gemma2Attention.forward + 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, + output_attentions: bool = False, + use_cache: bool = False, + cache_position: Optional[torch.LongTensor] = None, + ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]: + if output_attentions: + # TODO: Improve this warning with e.g. `model.config.attn_implementation = "manual"` once this is implemented. + logger.warning_once( + "Gemma2Model is using Gemma2SdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, " + 'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.' + ) + return super().forward( + hidden_states=hidden_states, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_value=past_key_value, + output_attentions=output_attentions, + use_cache=use_cache, + cache_position=cache_position, + ) + + bsz, q_len, _ = hidden_states.size() + + query_states = self.q_proj(hidden_states) + key_states = self.k_proj(hidden_states) + value_states = self.v_proj(hidden_states) + + query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2) + key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2) + + cos, sin = self.rotary_emb(value_states, position_ids) + query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin) + + if past_key_value is not None: + # sin and cos are specific to RoPE models; cache_position needed for the static cache + cache_kwargs = { + "sin": sin, + "cos": cos, + "sliding_window": self.sliding_window, + "cache_position": cache_position, + } + key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs) + + key_states = repeat_kv(key_states, self.num_key_value_groups) + value_states = repeat_kv(value_states, self.num_key_value_groups) + + causal_mask = attention_mask + if attention_mask is not None: + causal_mask = causal_mask[:, :, :, : key_states.shape[-2]] + + # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask, + # Reference: https://github.com/pytorch/pytorch/issues/112577. + if query_states.device.type == "cuda" and causal_mask is not None: + query_states = query_states.contiguous() + key_states = key_states.contiguous() + value_states = value_states.contiguous() + + # We dispatch to SDPA's Flash Attention or Efficient kernels via this `is_causal` if statement instead of an inline conditional assignment + # in SDPA to support both torch.compile's dynamic shapes and full graph options. An inline conditional prevents dynamic shapes from compiling. + is_causal = True if causal_mask is None and q_len > 1 else False + + attn_output = torch.nn.functional.scaled_dot_product_attention( + query_states, + key_states, + value_states, + attn_mask=causal_mask, + dropout_p=self.attention_dropout if self.training else 0.0, + is_causal=is_causal, + scale=self.scaling, + ) + + attn_output = attn_output.transpose(1, 2).contiguous() + attn_output = attn_output.view(bsz, q_len, -1) + + attn_output = self.o_proj(attn_output) + + return attn_output, None, past_key_value + + +GEMMA2_ATTENTION_CLASSES = { + "eager": Gemma2Attention, + "flash_attention_2": Gemma2FlashAttention2, + "sdpa": Gemma2SdpaAttention, +} + + +class Gemma2DecoderLayer(nn.Module): + def __init__(self, config: Gemma2Config, layer_idx: int): + super().__init__() + self.hidden_size = config.hidden_size + + self.self_attn = GEMMA2_ATTENTION_CLASSES[config._attn_implementation](config=config, layer_idx=layer_idx) + + self.mlp = Gemma2MLP(config) + self.input_layernorm = Gemma2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.post_attention_layernorm = Gemma2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + + self.is_sliding = bool(layer_idx % 2) + self.pre_feedforward_layernorm = Gemma2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.post_feedforward_layernorm = Gemma2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + self.sliding_window = config.sliding_window + + 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, + output_attentions: Optional[bool] = False, + use_cache: Optional[bool] = False, + cache_position: Optional[torch.LongTensor] = None, + ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]: + if self.is_sliding and attention_mask is not None: # efficient SDPA and no padding + attention_mask = attention_mask * torch.tril( + torch.ones_like(attention_mask), diagonal=-self.sliding_window + ) + if attention_mask.shape[1] <= 1: # when decoding + attention_mask = attention_mask[:, -self.sliding_window :] + + residual = hidden_states + + hidden_states = self.input_layernorm(hidden_states) + + # Self Attention + hidden_states, self_attn_weights, present_key_value = self.self_attn( + hidden_states=hidden_states, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_value=past_key_value, + output_attentions=output_attentions, + use_cache=use_cache, + cache_position=cache_position, + ) + hidden_states = self.post_attention_layernorm(hidden_states) + hidden_states = residual + hidden_states + + residual = hidden_states + hidden_states = self.pre_feedforward_layernorm(hidden_states) + hidden_states = self.mlp(hidden_states) + hidden_states = self.post_feedforward_layernorm(hidden_states) + hidden_states = residual + hidden_states + + outputs = (hidden_states,) + + if output_attentions: + outputs += (self_attn_weights,) + + if use_cache: + outputs += (present_key_value,) + + return outputs + + +GEMMA2_START_DOCSTRING = r""" + This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the + library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads + etc.) + + This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass. + Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage + and behavior. + + Parameters: + config ([`Gemma2Config`]): + Model configuration class with all the parameters of the model. Initializing with a config file does not + load the weights associated with the model, only the configuration. Check out the + [`~PreTrainedModel.from_pretrained`] method to load the model weights. +""" + + +@add_start_docstrings( + "The bare Gemma2 Model outputting raw hidden-states without any specific head on top.", + GEMMA2_START_DOCSTRING, +) +class Gemma2PreTrainedModel(PreTrainedModel): + config_class = Gemma2Config + base_model_prefix = "model" + supports_gradient_checkpointing = True + _no_split_modules = ["Gemma2DecoderLayer"] + _skip_keys_device_placement = ["past_key_values"] + _supports_flash_attn_2 = True + _supports_sdpa = True + _supports_cache_class = False + _supports_quantized_cache = False + _supports_static_cache = True + _is_stateful = True + + 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_() + + +_CONFIG_FOR_DOC = "Gemma2Config" + + +GEMMA2_INPUTS_DOCSTRING = r""" + Args: + input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): + Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide + it. + + Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and + [`PreTrainedTokenizer.__call__`] for details. + + [What are input IDs?](../glossary#input-ids) + attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*): + Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`: + + - 1 for tokens that are **not masked**, + - 0 for tokens that are **masked**. + + [What are attention masks?](../glossary#attention-mask) + + Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and + [`PreTrainedTokenizer.__call__`] for details. + + If `past_key_values` is used, optionally only the last `input_ids` have to be input (see + `past_key_values`). + + If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`] + and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more + information on the default strategy. + + - 1 indicates the head is **not masked**, + - 0 indicates the head is **masked**. + position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0, + config.n_positions - 1]`. + + [What are position IDs?](../glossary#position-ids) + past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*): + Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention + blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values` + returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`. + + Two formats are allowed: + - a [`~cache_utils.Cache`] instance; + - Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of + shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy + cache format. + + The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the + legacy cache format will be returned. + + If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't + have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids` + of shape `(batch_size, sequence_length)`. + inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*): + Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This + is useful if you want more control over how to convert `input_ids` indices into associated vectors than the + model's internal embedding lookup matrix. + use_cache (`bool`, *optional*): + If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see + `past_key_values`). + output_attentions (`bool`, *optional*): + Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned + tensors for more detail. + output_hidden_states (`bool`, *optional*): + Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for + more detail. + return_dict (`bool`, *optional*): + Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple. + cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*): + Indices depicting the position of the input sequence tokens in the sequence. Contrarily to `position_ids`, + this tensor is not affected by padding. It is used to update the cache in the correct position and to infer + the complete sequence length. +""" + + +@add_start_docstrings( + "The bare Gemma2 Model outputting raw hidden-states without any specific head on top.", + GEMMA2_START_DOCSTRING, +) +class Gemma2Model(Gemma2PreTrainedModel): + """ + Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`Gemma2DecoderLayer`] + + Args: + config: Gemma2Config + """ + + def __init__(self, config: Gemma2Config): + 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( + [Gemma2DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)] + ) + self.norm = Gemma2RMSNorm(config.hidden_size, eps=config.rms_norm_eps) + 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 + + @add_start_docstrings_to_model_forward(GEMMA2_INPUTS_DOCSTRING) + def forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + cache_position: Optional[torch.LongTensor] = None, + ) -> Union[Tuple, BaseModelOutputWithPast]: + output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + use_cache = use_cache if use_cache is not None else self.config.use_cache + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + if (input_ids is None) ^ (inputs_embeds is not None): + raise ValueError( + "You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one" + ) + + if self.gradient_checkpointing and self.training and use_cache: + logger.warning_once( + "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`." + ) + use_cache = False + + if inputs_embeds is None: + inputs_embeds = self.embed_tokens(input_ids) + + if cache_position is None: + cache_position = torch.arange(0, inputs_embeds.shape[1], device=inputs_embeds.device) + + if position_ids is None: + position_ids = cache_position.unsqueeze(0) + + causal_mask = self._update_causal_mask( + attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions + ) + + # embed positions + hidden_states = inputs_embeds + + # normalized + # Gemma2 downcasts the below to float16, causing sqrt(3072)=55.4256 to become 55.5 + # See https://github.com/huggingface/transformers/pull/29402 + normalizer = torch.tensor(self.config.hidden_size**0.5, dtype=hidden_states.dtype) + hidden_states = hidden_states * normalizer + + all_hidden_states = () if output_hidden_states else None + all_self_attns = () if output_attentions else None + + for decoder_layer in self.layers: + if output_hidden_states: + all_hidden_states += (hidden_states,) + + if self.gradient_checkpointing and self.training: + layer_outputs = self._gradient_checkpointing_func( + decoder_layer.__call__, + hidden_states, + causal_mask, + position_ids, + past_key_values, + output_attentions, + use_cache, + cache_position, + ) + else: + layer_outputs = decoder_layer( + hidden_states, + attention_mask=causal_mask, + position_ids=position_ids, + past_key_value=past_key_values, + output_attentions=output_attentions, + use_cache=use_cache, + cache_position=cache_position, + ) + + hidden_states = layer_outputs[0] + + if output_attentions: + all_self_attns += (layer_outputs[1],) + + hidden_states = self.norm(hidden_states) + + # add hidden states from the last decoder layer + if output_hidden_states: + all_hidden_states += (hidden_states,) + + next_cache = past_key_values if use_cache else None + + if not return_dict: + return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None) + return BaseModelOutputWithPast( + last_hidden_state=hidden_states, + past_key_values=next_cache, + hidden_states=all_hidden_states, + attentions=all_self_attns, + ) + + def _update_causal_mask( + self, + attention_mask: torch.Tensor, + input_tensor: torch.Tensor, + cache_position: torch.Tensor, + past_key_values: Cache, + output_attentions: bool, + ): + if self.config._attn_implementation == "flash_attention_2": + if attention_mask is not None and 0.0 in attention_mask: + return attention_mask + return None + + dtype, device = input_tensor.dtype, input_tensor.device + min_dtype = torch.finfo(dtype).min + sequence_length = input_tensor.shape[1] + if past_key_values is not None: + target_length = past_key_values.get_max_length() + else: + target_length = attention_mask.shape[-1] if attention_mask is not None else input_tensor.shape[1] + + if attention_mask is not None and attention_mask.dim() == 4: + # in this case we assume that the mask comes already in inverted form and requires no inversion or slicing + if attention_mask.max() != 0: + raise ValueError("Custom 4D attention mask should be passed in inverted form with max==0`") + causal_mask = attention_mask + else: + causal_mask = torch.full( + (sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=device + ) + if sequence_length != 1: + causal_mask = torch.triu(causal_mask, diagonal=1) + causal_mask *= torch.arange(target_length, device=device) > cache_position.reshape(-1, 1) + causal_mask = causal_mask[None, None, :, :].expand(input_tensor.shape[0], 1, -1, -1) + if attention_mask is not None: + causal_mask = causal_mask.clone() # copy to contiguous memory for in-place edit + mask_length = attention_mask.shape[-1] + padding_mask = causal_mask[:, :, :, :mask_length] + attention_mask[:, None, None, :] + padding_mask = padding_mask == 0 + causal_mask[:, :, :, :mask_length] = causal_mask[:, :, :, :mask_length].masked_fill( + padding_mask, min_dtype + ) + return causal_mask + + +class Gemma2ForCausalLM(Gemma2PreTrainedModel): + _tied_weights_keys = ["lm_head.weight"] + + def __init__(self, config): + super().__init__(config) + self.model = Gemma2Model(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 + + @add_start_docstrings_to_model_forward(GEMMA2_INPUTS_DOCSTRING) + @replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC) + def forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + cache_position: Optional[torch.LongTensor] = None, + ) -> Union[Tuple, CausalLMOutputWithPast]: + r""" + Args: + labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*): + Labels for computing the masked language modeling loss. Indices should either be in `[0, ..., + config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored + (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`. + + Returns: + + Example: + + ```python + >>> from transformers import AutoTokenizer, GemmaForCausalLM + + >>> model = GemmaForCausalLM.from_pretrained("google/gemma-2-9b") + >>> tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-9b") + + >>> prompt = "What is your favorite condiment?" + >>> 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] + "What is your favorite condiment?" + ```""" + if self.training and self.config._attn_implementation != "eager": + logger.warning_once( + "It is strongly recommended to train Gemma2 models with the `eager` attention implementation " + f"instead of `{self.config._attn_implementation}`. Use `eager` with `AutoModelForCausalLM.from_pretrained('', attn_implementation='eager')`." + ) + output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions + output_hidden_states = ( + output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states + ) + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn) + outputs = self.model( + input_ids=input_ids, + attention_mask=attention_mask, + position_ids=position_ids, + past_key_values=past_key_values, + inputs_embeds=inputs_embeds, + use_cache=use_cache, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + cache_position=cache_position, + ) + + hidden_states = outputs[0] + logits = self.lm_head(hidden_states) + if self.config.final_logit_softcapping is not None: + logits = logits / self.config.final_logit_softcapping + logits = torch.tanh(logits) + logits = logits * self.config.final_logit_softcapping + + logits = logits.float() + loss = None + if labels is not None: + # Shift so that tokens < n predict n + shift_logits = logits[..., :-1, :].contiguous() + shift_labels = labels[..., 1:].contiguous() + # Flatten the tokens + loss_fct = CrossEntropyLoss() + shift_logits = shift_logits.view(-1, self.config.vocab_size) + shift_labels = shift_labels.view(-1) + # Enable model parallelism + shift_labels = shift_labels.to(shift_logits.device) + loss = loss_fct(shift_logits, shift_labels) + + if not return_dict: + output = (logits,) + outputs[1:] + return (loss,) + output if loss is not None else output + + return CausalLMOutputWithPast( + loss=loss, + logits=logits, + past_key_values=outputs.past_key_values, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) + + def prepare_inputs_for_generation( + self, + input_ids, + past_key_values=None, + attention_mask=None, + inputs_embeds=None, + cache_position=None, + use_cache=True, + **kwargs, + ): + past_length = 0 + if past_key_values is not None: + # Past key values are always initialized with a `Cache` object -> no need for if-else anymore + past_length = cache_position[0] if cache_position is not None else torch.tensor(0, device=input_ids.device) + max_cache_length = ( + torch.tensor(past_key_values.get_max_length(), device=input_ids.device) + if past_key_values.get_max_length() is not None + else None + ) + cache_length = past_length if max_cache_length is None else torch.min(max_cache_length, past_length) + + # Keep only the unprocessed tokens: + # 1 - If the length of the attention_mask exceeds the length of input_ids, then we are in a setting where + # some of the inputs are exclusively passed as part of the cache (e.g. when passing input_embeds as input) + if attention_mask is not None and attention_mask.shape[1] > input_ids.shape[1]: + input_ids = input_ids[:, -(attention_mask.shape[1] - past_length) :] + # 2 - If the past_length is smaller than input_ids', then input_ids holds all input tokens. We can discard + # input_ids based on the past_length. + elif past_length < input_ids.shape[1]: + input_ids = input_ids[:, past_length:] + # 3 - Otherwise (past_length >= input_ids.shape[1]), let's assume input_ids only has unprocessed tokens. + + # If we are about to go beyond the maximum cache length, we need to crop the input attention mask. + if ( + max_cache_length is not None + and attention_mask is not None + and cache_length + input_ids.shape[1] > max_cache_length + ): + attention_mask = attention_mask[:, -max_cache_length:] + + position_ids = kwargs.get("position_ids", None) + if attention_mask is not None and position_ids is None: + # create position_ids on the fly for batch generation + position_ids = attention_mask.long().cumsum(-1) - 1 + position_ids.masked_fill_(attention_mask == 0, 1) + if past_key_values: + position_ids = position_ids[:, -input_ids.shape[1] :] + + # if `inputs_embeds` are passed, we only want to use them in the 1st generation step + if inputs_embeds is not None and past_length == 0: + model_inputs = {"inputs_embeds": inputs_embeds} + else: + # The `contiguous()` here is necessary to have a static stride during decoding. torchdynamo otherwise + # recompiles graphs as the stride of the inputs is a guard. Ref: https://github.com/huggingface/transformers/pull/29114 + # TODO: use `next_tokens` directly instead. + model_inputs = {"input_ids": input_ids.contiguous()} + + input_length = position_ids.shape[-1] if position_ids is not None else input_ids.shape[-1] + if cache_position is None: + cache_position = torch.arange(past_length, past_length + input_length, device=input_ids.device) + elif use_cache: + cache_position = cache_position[-input_length:] + + model_inputs.update( + { + "position_ids": position_ids, + "cache_position": cache_position, + "past_key_values": past_key_values, + "use_cache": use_cache, + "attention_mask": attention_mask, + } + ) + return model_inputs + + @staticmethod + def _reorder_cache(past_key_values, beam_idx): + reordered_past = () + for layer_past in past_key_values: + reordered_past += ( + tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past), + ) + return reordered_past + + +@add_start_docstrings( + """ + The Gemma2 Model transformer with a sequence classification head on top (linear layer). + + [`Gemma2ForSequenceClassification`] 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). + """, + GEMMA2_START_DOCSTRING, +) +class Gemma2ForSequenceClassification(Gemma2PreTrainedModel): + def __init__(self, config): + super().__init__(config) + self.num_labels = config.num_labels + self.model = Gemma2Model(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 + + @add_start_docstrings_to_model_forward(GEMMA2_INPUTS_DOCSTRING) + def forward( + self, + input_ids: torch.LongTensor = None, + attention_mask: Optional[torch.Tensor] = None, + position_ids: Optional[torch.LongTensor] = None, + past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple, 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). + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + transformer_outputs = 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, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + hidden_states = transformer_outputs[0] + 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: + sequence_lengths = -1 + else: + if input_ids is not None: + # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility + sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1 + sequence_lengths = sequence_lengths % input_ids.shape[-1] + sequence_lengths = sequence_lengths.to(logits.device) + else: + sequence_lengths = -1 + + pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths] + + loss = None + if labels is not None: + labels = labels.to(logits.device) + if self.config.problem_type is None: + if self.num_labels == 1: + self.config.problem_type = "regression" + elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int): + self.config.problem_type = "single_label_classification" + else: + self.config.problem_type = "multi_label_classification" + + if self.config.problem_type == "regression": + loss_fct = MSELoss() + if self.num_labels == 1: + loss = loss_fct(pooled_logits.squeeze(), labels.squeeze()) + else: + loss = loss_fct(pooled_logits, labels) + elif self.config.problem_type == "single_label_classification": + loss_fct = CrossEntropyLoss() + loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1)) + elif self.config.problem_type == "multi_label_classification": + loss_fct = BCEWithLogitsLoss() + loss = loss_fct(pooled_logits, labels) + if not return_dict: + output = (pooled_logits,) + transformer_outputs[1:] + return ((loss,) + output) if loss is not None else output + + 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, + ) + + +@add_start_docstrings( + """ + The Gemma2 Model transformer with a token classification head on top (a linear layer on top of the hidden-states + output) e.g. for Named-Entity-Recognition (NER) tasks. + """, + GEMMA2_START_DOCSTRING, +) +class Gemma2ForTokenClassification(Gemma2PreTrainedModel): + def __init__(self, config): + super().__init__(config) + self.num_labels = config.num_labels + self.model = Gemma2Model(config) + if getattr(config, "classifier_dropout", None) is not None: + classifier_dropout = config.classifier_dropout + elif getattr(config, "hidden_dropout", None) is not None: + classifier_dropout = config.hidden_dropout + else: + classifier_dropout = 0.1 + self.dropout = nn.Dropout(classifier_dropout) + self.score = nn.Linear(config.hidden_size, config.num_labels) + + # 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 + + @add_start_docstrings_to_model_forward(GEMMA2_INPUTS_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[List[torch.FloatTensor]] = None, + inputs_embeds: Optional[torch.FloatTensor] = None, + labels: Optional[torch.LongTensor] = None, + use_cache: Optional[bool] = None, + output_attentions: Optional[bool] = None, + output_hidden_states: Optional[bool] = None, + return_dict: Optional[bool] = None, + ) -> Union[Tuple, TokenClassifierOutput]: + 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). + """ + return_dict = return_dict if return_dict is not None else self.config.use_return_dict + + outputs = 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, + output_attentions=output_attentions, + output_hidden_states=output_hidden_states, + return_dict=return_dict, + ) + sequence_output = outputs[0] + sequence_output = self.dropout(sequence_output) + logits = self.score(sequence_output) + + loss = None + if labels is not None: + loss_fct = CrossEntropyLoss() + loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1)) + + if not return_dict: + output = (logits,) + outputs[2:] + return ((loss,) + output) if loss is not None else output + + return TokenClassifierOutput( + loss=loss, + logits=logits, + hidden_states=outputs.hidden_states, + attentions=outputs.attentions, + ) diff --git a/src/transformers/models/mistral/modeling_mistral.py b/src/transformers/models/mistral/modeling_mistral.py index 07942e87e68..475dda72c59 100644 --- a/src/transformers/models/mistral/modeling_mistral.py +++ b/src/transformers/models/mistral/modeling_mistral.py @@ -227,7 +227,6 @@ class MistralAttention(nn.Module): base=self.rope_theta, ) - # Copied from transformers.models.gemma.modeling_gemma.GemmaAttention.forward with Gemma->Mistral def forward( self, hidden_states: torch.Tensor, diff --git a/src/transformers/utils/dummy_pt_objects.py b/src/transformers/utils/dummy_pt_objects.py index 9132d161820..c9267debc5d 100644 --- a/src/transformers/utils/dummy_pt_objects.py +++ b/src/transformers/utils/dummy_pt_objects.py @@ -4197,6 +4197,41 @@ class GemmaPreTrainedModel(metaclass=DummyObject): requires_backends(self, ["torch"]) +class Gemma2ForCausalLM(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class Gemma2ForSequenceClassification(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class Gemma2ForTokenClassification(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class Gemma2Model(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + +class Gemma2PreTrainedModel(metaclass=DummyObject): + _backends = ["torch"] + + def __init__(self, *args, **kwargs): + requires_backends(self, ["torch"]) + + class GitForCausalLM(metaclass=DummyObject): _backends = ["torch"] diff --git a/tests/models/gemma/test_modeling_gemma.py b/tests/models/gemma/test_modeling_gemma.py index bdf01277437..c7fb55f682e 100644 --- a/tests/models/gemma/test_modeling_gemma.py +++ b/tests/models/gemma/test_modeling_gemma.py @@ -47,11 +47,18 @@ if is_torch_available(): GemmaForSequenceClassification, GemmaForTokenClassification, GemmaModel, - GemmaTokenizer, ) +@require_torch class GemmaModelTester: + config_class = GemmaConfig + if is_torch_available(): + model_class = GemmaModel + for_causal_lm_class = GemmaForCausalLM + for_sequence_class = GemmaForSequenceClassification + for_token_class = GemmaForTokenClassification + def __init__( self, parent, @@ -129,9 +136,8 @@ class GemmaModelTester: return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels - # Ignore copy def get_config(self): - return GemmaConfig( + return self.config_class( vocab_size=self.vocab_size, hidden_size=self.hidden_size, num_hidden_layers=self.num_hidden_layers, @@ -149,18 +155,16 @@ class GemmaModelTester: head_dim=self.head_dim, ) - # Copied from tests.models.llama.test_modeling_llama.LlamaModelTester.create_and_check_model with Llama->Gemma def create_and_check_model( self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels ): - model = GemmaModel(config=config) + model = self.model_class(config=config) model.to(torch_device) model.eval() result = model(input_ids, attention_mask=input_mask) result = model(input_ids) self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size)) - # Copied from tests.models.llama.test_modeling_llama.LlamaModelTester.create_and_check_model_as_decoder with Llama->Gemma def create_and_check_model_as_decoder( self, config, @@ -174,7 +178,7 @@ class GemmaModelTester: encoder_attention_mask, ): config.add_cross_attention = True - model = GemmaModel(config) + model = self.model_class(config) model.to(torch_device) model.eval() result = model( @@ -191,7 +195,6 @@ class GemmaModelTester: result = model(input_ids, attention_mask=input_mask) self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size)) - # Copied from tests.models.llama.test_modeling_llama.LlamaModelTester.create_and_check_for_causal_lm with Llama->Gemma def create_and_check_for_causal_lm( self, config, @@ -204,13 +207,12 @@ class GemmaModelTester: encoder_hidden_states, encoder_attention_mask, ): - model = GemmaForCausalLM(config=config) + model = self.for_causal_lm_class(config=config) model.to(torch_device) model.eval() result = model(input_ids, attention_mask=input_mask, labels=token_labels) self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size)) - # Copied from tests.models.llama.test_modeling_llama.LlamaModelTester.create_and_check_decoder_model_past_large_inputs with Llama->Gemma def create_and_check_decoder_model_past_large_inputs( self, config, @@ -225,7 +227,7 @@ class GemmaModelTester: ): config.is_decoder = True config.add_cross_attention = True - model = GemmaForCausalLM(config=config) + model = self.for_causal_lm_class(config=config) model.to(torch_device) model.eval() @@ -348,7 +350,7 @@ class GemmaModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixi input_ids = input_dict["input_ids"] attention_mask = input_ids.ne(1).to(torch_device) sequence_labels = ids_tensor([self.model_tester.batch_size], self.model_tester.type_sequence_label_size) - model = GemmaForSequenceClassification(config) + model = self.model_tester.for_sequence_class(config) model.to(torch_device) model.eval() result = model(input_ids, attention_mask=attention_mask, labels=sequence_labels) @@ -361,7 +363,7 @@ class GemmaModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixi input_ids = input_dict["input_ids"] attention_mask = input_ids.ne(1).to(torch_device) sequence_labels = ids_tensor([self.model_tester.batch_size], self.model_tester.type_sequence_label_size) - model = GemmaForSequenceClassification(config) + model = self.model_tester.for_sequence_class(config) model.to(torch_device) model.eval() result = model(input_ids, attention_mask=attention_mask, labels=sequence_labels) @@ -376,20 +378,19 @@ class GemmaModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixi sequence_labels = ids_tensor( [self.model_tester.batch_size, config.num_labels], self.model_tester.type_sequence_label_size ).to(torch.float) - model = GemmaForSequenceClassification(config) + model = self.model_tester.for_sequence_class(config) model.to(torch_device) model.eval() result = model(input_ids, attention_mask=attention_mask, labels=sequence_labels) self.assertEqual(result.logits.shape, (self.model_tester.batch_size, self.model_tester.num_labels)) - # Copied from tests.models.llama.test_modeling_llama.LlamaModelTest.test_llama_token_classification_model with Llama->Gemma,llama->Gemma def test_Gemma_token_classification_model(self): config, input_dict = self.model_tester.prepare_config_and_inputs_for_common() config.num_labels = 3 input_ids = input_dict["input_ids"] attention_mask = input_ids.ne(1).to(torch_device) token_labels = ids_tensor([self.model_tester.batch_size, self.model_tester.seq_length], config.num_labels) - model = GemmaForTokenClassification(config=config) + model = self.model_tester.for_token_class(config=config) model.to(torch_device) model.eval() result = model(input_ids, attention_mask=attention_mask, labels=token_labels) @@ -539,47 +540,9 @@ class GemmaIntegrationTest(unittest.TestCase): # 8 is for A100 / A10 and 7 for T4 cls.cuda_compute_capability_major_version = torch.cuda.get_device_capability()[0] - @require_read_token - def test_model_2b_fp32(self): - model_id = "google/gemma-2b" - EXPECTED_TEXTS = [ - "Hello I am doing a project on the 1990s and I need to know what the most popular music", - "Hi today I am going to share with you a very easy and simple recipe of Kaju Kat", - ] - - model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True).to(torch_device) - - tokenizer = AutoTokenizer.from_pretrained(model_id) - inputs = tokenizer(self.input_text, return_tensors="pt", padding=True).to(torch_device) - - output = model.generate(**inputs, max_new_tokens=20, do_sample=False) - output_text = tokenizer.batch_decode(output, skip_special_tokens=True) - - self.assertEqual(output_text, EXPECTED_TEXTS) - @require_read_token def test_model_2b_fp16(self): - model_id = "google/gemma-2b" - EXPECTED_TEXTS = [ - "Hello I am doing a project on the 1990s and I need to know what the most popular music", - "Hi today I am going to share with you a very easy and simple recipe of Kaju Kat", - ] - - model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True, torch_dtype=torch.float16).to( - torch_device - ) - - tokenizer = AutoTokenizer.from_pretrained(model_id) - inputs = tokenizer(self.input_text, return_tensors="pt", padding=True).to(torch_device) - - output = model.generate(**inputs, max_new_tokens=20, do_sample=False) - output_text = tokenizer.batch_decode(output, skip_special_tokens=True) - - self.assertEqual(output_text, EXPECTED_TEXTS) - - @require_read_token - def test_model_2b_fp16_static_cache(self): - model_id = "google/gemma-2b" + model_id = "google/gemma-2-9b" EXPECTED_TEXTS = [ "Hello I am doing a project on the 1990s and I need to know what the most popular music", "Hi today I am going to share with you a very easy and simple recipe of Kaju Kat", @@ -903,7 +866,7 @@ class GemmaIntegrationTest(unittest.TestCase): } prompts = ["Hello I am doing", "Hi today"] - tokenizer = GemmaTokenizer.from_pretrained("google/gemma-2b", pad_token="", padding_side="right") + tokenizer = AutoTokenizer.from_pretrained("google/gemma-2b", pad_token="", padding_side="right") model = GemmaForCausalLM.from_pretrained("google/gemma-2b", device_map="sequential", torch_dtype=torch.float16) inputs = tokenizer(prompts, return_tensors="pt", padding=True).to(model.device) diff --git a/tests/models/gemma2/__init__.py b/tests/models/gemma2/__init__.py new file mode 100644 index 00000000000..e69de29bb2d diff --git a/tests/models/gemma2/test_modeling_gemma2.py b/tests/models/gemma2/test_modeling_gemma2.py new file mode 100644 index 00000000000..6a6c5688d5d --- /dev/null +++ b/tests/models/gemma2/test_modeling_gemma2.py @@ -0,0 +1,141 @@ +# coding=utf-8 +# Copyright 2024 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 Gemma2 model.""" + +import unittest + +from transformers import AutoModelForCausalLM, AutoTokenizer, Gemma2Config, is_torch_available +from transformers.testing_utils import ( + require_read_token, + require_torch, + require_torch_gpu, + slow, + torch_device, +) + +from ...models.gemma.test_modeling_gemma import GemmaModelTest, GemmaModelTester +from ...test_configuration_common import ConfigTester + + +if is_torch_available(): + import torch + + from transformers import ( + Gemma2ForCausalLM, + Gemma2ForSequenceClassification, + Gemma2ForTokenClassification, + Gemma2Model, + ) + + +class Gemma2ModelTester(GemmaModelTester): + config_class = Gemma2Config + model_class = Gemma2Model + for_causal_lm_class = Gemma2ForCausalLM + for_sequence_class = Gemma2ForSequenceClassification + for_token_class = Gemma2ForTokenClassification + + +@require_torch +class Gemma2ModelTest(GemmaModelTest, unittest.TestCase): + all_model_classes = ( + (Gemma2Model, Gemma2ForCausalLM, Gemma2ForSequenceClassification, Gemma2ForTokenClassification) + if is_torch_available() + else () + ) + all_generative_model_classes = () + pipeline_model_mapping = ( + { + "feature-extraction": Gemma2Model, + "text-classification": Gemma2ForSequenceClassification, + "token-classification": Gemma2ForTokenClassification, + "text-generation": Gemma2ForCausalLM, + "zero-shot": Gemma2ForSequenceClassification, + } + if is_torch_available() + else {} + ) + test_headmasking = False + test_pruning = False + _is_stateful = True + model_split_percents = [0.5, 0.6] + _torch_compile_test_ckpt = "google/gemma-2-9b" + + def setUp(self): + self.model_tester = Gemma2ModelTester(self) + self.config_tester = ConfigTester(self, config_class=Gemma2Config, hidden_size=37) + + @unittest.skip("Eager and SDPA do not produce the same outputs, thus this test fails") + def test_model_outputs_equivalence(self, **kwargs): + pass + + @unittest.skip("Gemma2's outputs are expected to be different") + def test_eager_matches_sdpa_inference(self): + pass + + +@slow +@require_torch_gpu +class Gemma2IntegrationTest(unittest.TestCase): + input_text = ["Hello I am doing", "Hi today"] + # This variable is used to determine which CUDA device are we using for our runners (A10 or T4) + # Depending on the hardware we get different logits / generations + cuda_compute_capability_major_version = None + + @classmethod + def setUpClass(cls): + if is_torch_available() and torch.cuda.is_available(): + # 8 is for A100 / A10 and 7 for T4 + cls.cuda_compute_capability_major_version = torch.cuda.get_device_capability()[0] + + @require_read_token + def test_model_2b_bf16(self): + model_id = "google/gemma-2-9b" + EXPECTED_TEXTS = [ + "Hello I am doing a project for a class and I am trying to use the ", + "Hi today. So, I'm going to show you how to do a problem from the textbook. So", + ] + + model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16).to( + torch_device + ) + + tokenizer = AutoTokenizer.from_pretrained(model_id) + inputs = tokenizer(self.input_text, return_tensors="pt", padding=True).to(torch_device) + + output = model.generate(**inputs, max_new_tokens=20, do_sample=False) + output_text = tokenizer.batch_decode(output, skip_special_tokens=True) + + self.assertEqual(output_text, EXPECTED_TEXTS) + + @require_read_token + def test_model_2b_fp16(self): + model_id = "google/gemma-2-9b" + EXPECTED_TEXTS = [ + "Hello I am doing a project on the effect of the temperature on the rate of a reaction. I am using a ", + "Hi today I'm going to be talking about the 1000-4000-", + ] + + model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True, torch_dtype=torch.float16).to( + torch_device + ) + + tokenizer = AutoTokenizer.from_pretrained(model_id) + inputs = tokenizer(self.input_text, return_tensors="pt", padding=True).to(torch_device) + + output = model.generate(**inputs, max_new_tokens=20, do_sample=False) + output_text = tokenizer.batch_decode(output, skip_special_tokens=True) + + self.assertEqual(output_text, EXPECTED_TEXTS) diff --git a/tests/test_modeling_common.py b/tests/test_modeling_common.py index 11c34462ba5..4561c93c21d 100755 --- a/tests/test_modeling_common.py +++ b/tests/test_modeling_common.py @@ -505,7 +505,7 @@ class ModelTesterMixin: # Check that the parameters are equal. for p1, p2 in zip(model_low_usage.parameters(), model_non_low_usage.parameters()): - self.assertEquals(p1.data.ne(p2.data).sum(), 0) + self.assertEqual(p1.data.ne(p2.data).sum(), 0) # Check that the state dict keys are equal. self.assertEqual(set(model_low_usage.state_dict().keys()), set(model_non_low_usage.state_dict().keys())) diff --git a/utils/check_config_attributes.py b/utils/check_config_attributes.py index e6edcf517a0..91113717610 100644 --- a/utils/check_config_attributes.py +++ b/utils/check_config_attributes.py @@ -41,6 +41,7 @@ SPECIAL_CASES_TO_ALLOW = { "expert_layer_offset", "expert_layer_period", ], + "Gemma2Config": ["tie_word_embeddings"], # used to compute the property `self.chunk_length` "EncodecConfig": ["overlap"], # used to compute the property `self.layers_block_type`