Add OLMo November 2024 (huggingface#34551)

* Add model skeletion with transformers-cli add-new-model-like

* Convert config to modular, add rms_norm_eps, delete clip_qkv

* Convert model to modular, add RMSNorm

* Add flash attention with qk norm and no qkv clipping

* Add decoder layer with RMSNorm after attention/feedforward layers

* Add base and causal model

* Add converter improvements from OLMo repo

* Update weight loading in OLMo to HF converter

* Set correct default for rms_norm_eps

* Set correct pipeline_model_mapping in test

* Run make fixup

* Fix model type

* Re-run modular conversion

* Manually set config docs to fix build errors

* Convert olmo-1124 to olmo_1124 to fix flash attention docs errors

* Start updating tests

* Update tests

* Copy upstream test_eager_matches_sdpa_inference_1_bfloat16 changes to olmo_1124

* Rename input_layernorm and post_attention_layernorm to reflect their ops better

* Use correct tokenizer

* Remove test unsupported by GPT2 tokenizer

* Create GenerationConfig outside of from_pretrained call

* Use simpler init file structure

* Add explicit __all__ to support simplified init

* Make safetensor serialization the default

* Update OLMo November 2024 docs

docs/source/en/_toctree.yml

@@ -514,6 +514,8 @@
         title: Nyströmformer
       - local: model_doc/olmo
         title: OLMo
+      - local: model_doc/olmo_1124
+        title: OLMo November 2024
       - local: model_doc/olmoe
         title: OLMoE
       - local: model_doc/open-llama

docs/source/en/index.md

@@ -241,6 +241,7 @@ Flax), PyTorch, and/or TensorFlow.
 |                        [Nougat](model_doc/nougat)                        |       ✅        |         ✅         |      ✅      |
 |                 [Nyströmformer](model_doc/nystromformer)                 |       ✅        |         ❌         |      ❌      |
 |                          [OLMo](model_doc/olmo)                          |       ✅        |         ❌         |      ❌      |
+|                [OLMo November 2024](model_doc/olmo_1124)                 |       ✅        |         ❌         |      ❌      |
 |                         [OLMoE](model_doc/olmoe)                         |       ✅        |         ❌         |      ❌      |
 |                   [OmDet-Turbo](model_doc/omdet-turbo)                   |       ✅        |         ❌         |      ❌      |
 |                     [OneFormer](model_doc/oneformer)                     |       ✅        |         ❌         |      ❌      |

docs/source/en/model_doc/olmo_1124.md

@@ -0,0 +1,46 @@
+<!--Copyright 2024 The HuggingFace Team. All rights reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with
+the License. You may obtain a copy of the License at
+
+http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on
+an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the
+specific language governing permissions and limitations under the License.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+# OLMo November 2024
+
+## Overview
+
+The OLMo November 2024 model is a successor of the OLMo model, which was proposed in
+[OLMo: Accelerating the Science of Language Models](https://arxiv.org/abs/2402.00838).
+
+ The architectural changes from the original OLMo model to this model are:
+
+- RMSNorm is used instead of standard layer norm.
+- Norm is applied to attention queries and keys.
+- Norm is applied after attention/feedforward layers rather than before.
+
+This model was contributed by [shanearora](https://huggingface.co/shanearora).
+The original code can be found [here](https://github.com/allenai/OLMo/tree/main/olmo).
+
+
+## Olmo1124Config
+
+[[autodoc]] Olmo1124Config
+
+## Olmo1124Model
+
+[[autodoc]] Olmo1124Model
+    - forward
+
+## Olmo1124ForCausalLM
+
+[[autodoc]] Olmo1124ForCausalLM
+    - forward

docs/source/en/perf_infer_gpu_one.md

@@ -77,6 +77,7 @@ FlashAttention-2 is currently supported for the following architectures:
 * [Nemotron](https://huggingface.co/docs/transformers/model_doc/nemotron)
 * [NLLB](https://huggingface.co/docs/transformers/model_doc/nllb)
 * [OLMo](https://huggingface.co/docs/transformers/model_doc/olmo#transformers.OlmoModel)
+* [OLMo November 2024](https://huggingface.co/docs/transformers/model_doc/olmo_1124#transformers.Olmo1124Model)
 * [OLMoE](https://huggingface.co/docs/transformers/model_doc/olmoe#transformers.OlmoeModel)
 * [OPT](https://huggingface.co/docs/transformers/model_doc/opt#transformers.OPTModel)
 * [PaliGemma](https://huggingface.co/docs/transformers/model_doc/paligemma#transformers.PaliGemmaForConditionalGeneration)
@@ -260,6 +261,7 @@ For now, Transformers supports SDPA inference and training for the following arc
 * [MusicGen Melody](https://huggingface.co/docs/transformers/model_doc/musicgen_melody#transformers.MusicgenMelodyModel)
 * [NLLB](https://huggingface.co/docs/transformers/model_doc/nllb)
 * [OLMo](https://huggingface.co/docs/transformers/model_doc/olmo#transformers.OlmoModel)
+* [OLMo November 2024](https://huggingface.co/docs/transformers/model_doc/olmo_1124#transformers.Olmo1124Model)
 * [OLMoE](https://huggingface.co/docs/transformers/model_doc/olmoe#transformers.OlmoeModel)
 * [OPT](https://huggingface.co/docs/transformers/en/model_doc/opt)
 * [PaliGemma](https://huggingface.co/docs/transformers/model_doc/paligemma#transformers.PaliGemmaForConditionalGeneration)

src/transformers/__init__.py

@@ -621,6 +621,7 @@
     "models.nougat": ["NougatProcessor"],
     "models.nystromformer": ["NystromformerConfig"],
     "models.olmo": ["OlmoConfig"],
+    "models.olmo_1124": ["Olmo1124Config"],
     "models.olmoe": ["OlmoeConfig"],
     "models.omdet_turbo": [
         "OmDetTurboConfig",
@@ -2928,6 +2929,13 @@
             "OlmoPreTrainedModel",
         ]
     )
+    _import_structure["models.olmo_1124"].extend(
+        [
+            "Olmo1124ForCausalLM",
+            "Olmo1124Model",
+            "Olmo1124PreTrainedModel",
+        ]
+    )
     _import_structure["models.olmoe"].extend(
         [
             "OlmoeForCausalLM",
@@ -5516,6 +5524,7 @@
         NystromformerConfig,
     )
     from .models.olmo import OlmoConfig
+    from .models.olmo_1124 import Olmo1124Config
     from .models.olmoe import OlmoeConfig
     from .models.omdet_turbo import (
         OmDetTurboConfig,
@@ -7539,6 +7548,11 @@
             OlmoModel,
             OlmoPreTrainedModel,
         )
+        from .models.olmo_1124 import (
+            Olmo1124ForCausalLM,
+            Olmo1124Model,
+            Olmo1124PreTrainedModel,
+        )
         from .models.olmoe import (
             OlmoeForCausalLM,
             OlmoeModel,

src/transformers/models/__init__.py

@@ -178,6 +178,7 @@
     nougat,
     nystromformer,
     olmo,
+    olmo_1124,
     olmoe,
     omdet_turbo,
     oneformer,

src/transformers/models/auto/configuration_auto.py

@@ -196,6 +196,7 @@
         ("nougat", "VisionEncoderDecoderConfig"),
         ("nystromformer", "NystromformerConfig"),
         ("olmo", "OlmoConfig"),
+        ("olmo_1124", "Olmo1124Config"),
         ("olmoe", "OlmoeConfig"),
         ("omdet-turbo", "OmDetTurboConfig"),
         ("oneformer", "OneFormerConfig"),
@@ -512,6 +513,7 @@
         ("nougat", "Nougat"),
         ("nystromformer", "Nyströmformer"),
         ("olmo", "OLMo"),
+        ("olmo_1124", "OLMo November 2024"),
         ("olmoe", "OLMoE"),
         ("omdet-turbo", "OmDet-Turbo"),
         ("oneformer", "OneFormer"),

src/transformers/models/auto/modeling_auto.py

@@ -185,6 +185,7 @@
         ("nllb-moe", "NllbMoeModel"),
         ("nystromformer", "NystromformerModel"),
         ("olmo", "OlmoModel"),
+        ("olmo_1124", "Olmo1124Model"),
         ("olmoe", "OlmoeModel"),
         ("omdet-turbo", "OmDetTurboForObjectDetection"),
         ("oneformer", "OneFormerModel"),
@@ -517,6 +518,7 @@
         ("mvp", "MvpForCausalLM"),
         ("nemotron", "NemotronForCausalLM"),
         ("olmo", "OlmoForCausalLM"),
+        ("olmo_1124", "Olmo1124ForCausalLM"),
         ("olmoe", "OlmoeForCausalLM"),
         ("open-llama", "OpenLlamaForCausalLM"),
         ("openai-gpt", "OpenAIGPTLMHeadModel"),

src/transformers/models/auto/tokenization_auto.py

@@ -348,6 +348,7 @@
                 ),
             ),
             ("olmo", (None, "GPTNeoXTokenizerFast" if is_tokenizers_available() else None)),
+            ("olmo_1124", (None, "GPTNeoXTokenizerFast" if is_tokenizers_available() else None)),
             ("olmoe", (None, "GPTNeoXTokenizerFast" if is_tokenizers_available() else None)),
             (
                 "omdet-turbo",

src/transformers/models/olmo_1124/__init__.py

@@ -0,0 +1,27 @@
+# Copyright 2024 EleutherAI and 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 _LazyModule
+from ...utils.import_utils import define_import_structure
+
+
+if TYPE_CHECKING:
+    from .configuration_olmo_1124 import *
+    from .modeling_olmo_1124 import *
+else:
+    import sys
+
+    _file = globals()["__file__"]
+    sys.modules[__name__] = _LazyModule(__name__, _file, define_import_structure(_file), module_spec=__spec__)

src/transformers/models/olmo_1124/configuration_olmo_1124.py

@@ -0,0 +1,166 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/olmo_1124/modular_olmo_1124.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_olmo_1124.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+
+from ...configuration_utils import PretrainedConfig
+
+
+class Olmo1124Config(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`Olmo1124Model`]. It is used to instantiate an OLMo November 2024
+    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 [allenai/Olmo1124-7B-hf](https://huggingface.co/allenai/Olmo1124-7B-hf).
+
+    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 50304):
+            Vocabulary size of the Olmo1124 model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`Olmo1124Model`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 11008):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer decoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        num_key_value_heads (`int`, *optional*):
+            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`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to 2048):
+            The maximum sequence length that this model might ever be used with.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        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 1):
+            Padding token id.
+        bos_token_id (`int`, *optional*):
+            Beginning of stream token id.
+        eos_token_id (`int`, *optional*, defaults to 50279):
+            End of stream token id.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        rope_scaling (`Dict`, *optional*):
+            Dictionary containing the scaling configuration for the RoPE embeddings. Currently supports two scaling
+            strategies: linear and dynamic. Their scaling factor must be a float greater than 1. The expected format is
+            `{"type": strategy name, "factor": scaling factor}`. When using this flag, don't update
+            `max_position_embeddings` to the expected new maximum. See the following thread for more information on how
+            these scaling strategies behave:
+            https://www.reddit.com/r/LocalLLaMA/comments/14mrgpr/dynamically_scaled_rope_further_increases/. This is an
+            experimental feature, subject to breaking API changes in future versions.
+        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.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-05):
+            The epsilon used by the rms normalization layers.
+
+    ```python
+    >>> from transformers import Olmo1124Model, Olmo1124Config
+
+    >>> # Initializing a Olmo November 2024 7B style configuration
+    >>> configuration = Olmo1124Config()
+
+    >>> # Initializing a model from the Olmo November 2024 7B style configuration
+    >>> model = Olmo1124Model(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```
+    """
+
+    model_type = "olmo_1124"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=50304,
+        hidden_size=4096,
+        intermediate_size=11008,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=None,
+        hidden_act="silu",
+        max_position_embeddings=2048,
+        initializer_range=0.02,
+        use_cache=True,
+        pad_token_id=1,
+        bos_token_id=None,
+        eos_token_id=50279,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        rope_scaling=None,
+        attention_bias=False,
+        attention_dropout=0.0,
+        rms_norm_eps=1e-5,
+        **kwargs,
+    ):
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **kwargs,
+        )
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self._rope_scaling_validation()
+        self.attention_bias = attention_bias
+        self.attention_dropout = attention_dropout
+
+        self.rms_norm_eps = rms_norm_eps
+
+    def _rope_scaling_validation(self):
+        """
+        Validate the `rope_scaling` configuration.
+        """
+        if self.rope_scaling is None:
+            return
+
+        if not isinstance(self.rope_scaling, dict) or len(self.rope_scaling) != 2:
+            raise ValueError(
+                "`rope_scaling` must be a dictionary with two fields, `type` and `factor`, " f"got {self.rope_scaling}"
+            )
+        rope_scaling_type = self.rope_scaling.get("type", None)
+        rope_scaling_factor = self.rope_scaling.get("factor", None)
+        if rope_scaling_type is None or rope_scaling_type not in ["linear", "dynamic"]:
+            raise ValueError(
+                f"`rope_scaling`'s type field must be one of ['linear', 'dynamic'], got {rope_scaling_type}"
+            )
+        if rope_scaling_factor is None or not isinstance(rope_scaling_factor, float) or rope_scaling_factor <= 1.0:
+            raise ValueError(f"`rope_scaling`'s factor field must be a float > 1, got {rope_scaling_factor}")
+
+
+__all__ = ["Olmo1124Config"]