Phi-3 (#30423)

* chore(root): Initial commit of Phi-3 files.

* fix(root): Fixes Phi-3 missing on readme.

* fix(root): Ensures files are consistent.

* fix(phi3): Fixes unit tests.

* fix(tests): Fixes style of phi-3 test file.

* chore(tests): Adds integration tests for Phi-3.

* fix(phi3): Removes additional flash-attention usage, .e.g, swiglu and rmsnorm.

* fix(phi3): Fixes incorrect docstrings.

* fix(phi3): Fixes docstring typos.

* fix(phi3): Adds support for Su and Yarn embeddings.

* fix(phi3): Improves according first batch of reviews.

* fix(phi3): Uses up_states instead of y in Phi3MLP.

* fix(phi3): Uses gemma rotary embedding to support torch.compile.

* fix(phi3): Improves how rotary embedding classes are defined.

* fix(phi3): Fixes inv_freq not being re-computed for extended RoPE.

* fix(phi3): Adds last suggestions to modeling file.

* fix(phi3): Splits inv_freq calculation in two lines.

docs/source/en/_toctree.yml

@@ -460,6 +460,8 @@
         title: Persimmon
       - local: model_doc/phi
         title: Phi
+      - local: model_doc/phi3
+        title: Phi-3
       - local: model_doc/phobert
         title: PhoBERT
       - local: model_doc/plbart

docs/source/en/index.md

@@ -236,6 +236,7 @@ Flax), PyTorch, and/or TensorFlow.
 |                     [Perceiver](model_doc/perceiver)                     |       ✅        |         ❌         |      ❌      |
 |                     [Persimmon](model_doc/persimmon)                     |       ✅        |         ❌         |      ❌      |
 |                           [Phi](model_doc/phi)                           |       ✅        |         ❌         |      ❌      |
+|                          [Phi3](model_doc/phi3)                          |       ✅        |         ❌         |      ❌      |
 |                       [PhoBERT](model_doc/phobert)                       |       ✅        |         ✅         |      ✅      |
 |                    [Pix2Struct](model_doc/pix2struct)                    |       ✅        |         ❌         |      ❌      |
 |                        [PLBart](model_doc/plbart)                        |       ✅        |         ❌         |      ❌      |

docs/source/en/model_doc/phi3.md

@@ -0,0 +1,92 @@
+<!--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.
+
+-->
+
+# Phi-3
+
+## Overview
+
+The Phi-3 model was proposed in [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://arxiv.org/abs/2404.14219) by Microsoft.
+
+### Summary
+
+The abstract from the Phi-3 paper is the following:
+
+We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens, whose overall performance, as measured by both academic benchmarks and internal testing, rivals that of models such as Mixtral 8x7B and GPT-3.5 (e.g., phi-3-mini achieves 69% on MMLU and 8.38 on MT-bench), despite being small enough to be deployed on a phone. The innovation lies entirely in our dataset for training, a scaled-up version of the one used for phi-2, composed of heavily filtered web data and synthetic data. The model is also further aligned for robustness, safety, and chat format. We also provide some initial parameter-scaling results with a 7B and 14B models trained for 4.8T tokens, called phi-3-small and phi-3-medium, both significantly more capable than phi-3-mini (e.g., respectively 75% and 78% on MMLU, and 8.7 and 8.9 on MT-bench).
+
+The original code for Phi-3 can be found [here](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct).
+
+## Usage tips
+
+- This model is very similar to `Llama` with the main difference of [`Phi3SuScaledRotaryEmbedding`] and [`Phi3YarnScaledRotaryEmbedding`], where they are used to extend the context of the rotary embeddings. The query, key and values are fused, and the MLP's up and gate projection layers are also fused.
+- The tokenizer used for this model is identical to the [`LlamaTokenizer`], with the exception of additional tokens.
+
+## How to use Phi-3
+
+<Tip warning={true}>
+
+Phi-3 has been integrated in the development version (4.40.0.dev) of `transformers`. Until the official version is released through `pip`, ensure that you are doing one of the following:
+
+* When loading the model, ensure that `trust_remote_code=True` is passed as an argument of the `from_pretrained()` function.
+
+* Update your local `transformers` to the development version: `pip uninstall -y transformers && pip install git+https://github.com/huggingface/transformers`. The previous command is an alternative to cloning and installing from the source.
+
+</Tip>
+
+```python
+>>> from transformers import AutoModelForCausalLM, AutoTokenizer
+
+>>> model = AutoModelForCausalLM.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
+>>> tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
+
+>>> messages = [{"role": "system", "content": "You are a helpful digital assistant. Please provide safe, ethical and accurate information to the user."},{"role": "user", "content": "Can you provide ways to eat combinations of bananas and dragonfruits?"}]
+>>> inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")
+
+>>> outputs = model.generate(inputs, max_new_tokens=32)
+>>> text = tokenizer.batch_decode(outputs)[0]
+>>> print(text)
+<s><|system|> You are a helpful digital assistant. Please provide safe, ethical and accurate information to the user.<|end|><|user|> Can you provide ways to eat combinations of bananas and dragonfruits?<|end|><|assistant|> Absolutely! Bananas and dragonfruits are both delicious fruits that can be combined in various ways to create tasty and nutrit
+```
+
+## Phi3Config
+
+[[autodoc]] Phi3Config
+
+<frameworkcontent>
+<pt>
+
+## Phi3Model
+
+[[autodoc]] Phi3Model
+    - forward
+
+## Phi3ForCausalLM
+
+[[autodoc]] Phi3ForCausalLM
+    - forward
+    - generate
+
+## Phi3ForSequenceClassification
+
+[[autodoc]] Phi3ForSequenceClassification
+    - forward
+
+## Phi3ForTokenClassification
+
+[[autodoc]] Phi3ForTokenClassification
+    - forward
+
+</pt>
+</frameworkcontent>

docs/source/en/perf_infer_gpu_one.md

@@ -65,6 +65,7 @@ FlashAttention-2 is currently supported for the following architectures:
 * [OLMo](https://huggingface.co/docs/transformers/model_doc/olmo#transformers.OlmoModel)
 * [OPT](https://huggingface.co/docs/transformers/model_doc/opt#transformers.OPTModel)
 * [Phi](https://huggingface.co/docs/transformers/model_doc/phi#transformers.PhiModel)
+* [Phi3](https://huggingface.co/docs/transformers/model_doc/phi3#transformers.Phi3Model)
 * [StableLm](https://huggingface.co/docs/transformers/model_doc/stablelm#transformers.StableLmModel)
 * [Starcoder2](https://huggingface.co/docs/transformers/model_doc/starcoder2#transformers.Starcoder2Model)
 * [Qwen2](https://huggingface.co/docs/transformers/model_doc/qwen2#transformers.Qwen2Model)

src/transformers/__init__.py

@@ -709,6 +709,7 @@
     ],
     "models.persimmon": ["PERSIMMON_PRETRAINED_CONFIG_ARCHIVE_MAP", "PersimmonConfig"],
     "models.phi": ["PHI_PRETRAINED_CONFIG_ARCHIVE_MAP", "PhiConfig"],
+    "models.phi3": ["PHI3_PRETRAINED_CONFIG_ARCHIVE_MAP", "Phi3Config"],
     "models.phobert": ["PhobertTokenizer"],
     "models.pix2struct": [
         "PIX2STRUCT_PRETRAINED_CONFIG_ARCHIVE_MAP",
@@ -3057,6 +3058,16 @@
             "PhiPreTrainedModel",
         ]
     )
+    _import_structure["models.phi3"].extend(
+        [
+            "PHI3_PRETRAINED_MODEL_ARCHIVE_LIST",
+            "Phi3ForCausalLM",
+            "Phi3ForSequenceClassification",
+            "Phi3ForTokenClassification",
+            "Phi3Model",
+            "Phi3PreTrainedModel",
+        ]
+    )
     _import_structure["models.pix2struct"].extend(
         [
             "PIX2STRUCT_PRETRAINED_MODEL_ARCHIVE_LIST",
@@ -5669,6 +5680,7 @@
         PersimmonConfig,
     )
     from .models.phi import PHI_PRETRAINED_CONFIG_ARCHIVE_MAP, PhiConfig
+    from .models.phi3 import PHI3_PRETRAINED_CONFIG_ARCHIVE_MAP, Phi3Config
     from .models.phobert import PhobertTokenizer
     from .models.pix2struct import (
         PIX2STRUCT_PRETRAINED_CONFIG_ARCHIVE_MAP,
@@ -7715,6 +7727,14 @@
             PhiModel,
             PhiPreTrainedModel,
         )
+        from .models.phi3 import (
+            PHI3_PRETRAINED_MODEL_ARCHIVE_LIST,
+            Phi3ForCausalLM,
+            Phi3ForSequenceClassification,
+            Phi3ForTokenClassification,
+            Phi3Model,
+            Phi3PreTrainedModel,
+        )
         from .models.pix2struct import (
             PIX2STRUCT_PRETRAINED_MODEL_ARCHIVE_LIST,
             Pix2StructForConditionalGeneration,

src/transformers/models/__init__.py

@@ -179,6 +179,7 @@
     perceiver,
     persimmon,
     phi,
+    phi3,
     phobert,
     pix2struct,
     plbart,

src/transformers/models/auto/configuration_auto.py

@@ -191,6 +191,7 @@
         ("perceiver", "PerceiverConfig"),
         ("persimmon", "PersimmonConfig"),
         ("phi", "PhiConfig"),
+        ("phi3", "Phi3Config"),
         ("pix2struct", "Pix2StructConfig"),
         ("plbart", "PLBartConfig"),
         ("poolformer", "PoolFormerConfig"),
@@ -471,6 +472,7 @@
         ("perceiver", "Perceiver"),
         ("persimmon", "Persimmon"),
         ("phi", "Phi"),
+        ("phi3", "Phi3"),
         ("phobert", "PhoBERT"),
         ("pix2struct", "Pix2Struct"),
         ("plbart", "PLBart"),

src/transformers/models/auto/modeling_auto.py

@@ -180,6 +180,7 @@
         ("perceiver", "PerceiverModel"),
         ("persimmon", "PersimmonModel"),
         ("phi", "PhiModel"),
+        ("phi3", "Phi3Model"),
         ("plbart", "PLBartModel"),
         ("poolformer", "PoolFormerModel"),
         ("prophetnet", "ProphetNetModel"),
@@ -474,6 +475,7 @@
         ("pegasus", "PegasusForCausalLM"),
         ("persimmon", "PersimmonForCausalLM"),
         ("phi", "PhiForCausalLM"),
+        ("phi3", "Phi3ForCausalLM"),
         ("plbart", "PLBartForCausalLM"),
         ("prophetnet", "ProphetNetForCausalLM"),
         ("qdqbert", "QDQBertLMHeadModel"),
@@ -884,6 +886,7 @@
         ("perceiver", "PerceiverForSequenceClassification"),
         ("persimmon", "PersimmonForSequenceClassification"),
         ("phi", "PhiForSequenceClassification"),
+        ("phi3", "Phi3ForSequenceClassification"),
         ("plbart", "PLBartForSequenceClassification"),
         ("qdqbert", "QDQBertForSequenceClassification"),
         ("qwen2", "Qwen2ForSequenceClassification"),
@@ -1049,6 +1052,7 @@
         ("nezha", "NezhaForTokenClassification"),
         ("nystromformer", "NystromformerForTokenClassification"),
         ("phi", "PhiForTokenClassification"),
+        ("phi3", "Phi3ForTokenClassification"),
         ("qdqbert", "QDQBertForTokenClassification"),
         ("rembert", "RemBertForTokenClassification"),
         ("roberta", "RobertaForTokenClassification"),

src/transformers/models/auto/tokenization_auto.py

@@ -353,6 +353,7 @@
                 ),
             ),
             ("phi", ("CodeGenTokenizer", "CodeGenTokenizerFast" if is_tokenizers_available() else None)),
+            ("phi3", ("LlamaTokenizer", "LlamaTokenizerFast" if is_tokenizers_available() else None)),
             ("phobert", ("PhobertTokenizer", None)),
             ("pix2struct", ("T5Tokenizer", "T5TokenizerFast" if is_tokenizers_available() else None)),
             ("plbart", ("PLBartTokenizer" if is_sentencepiece_available() else None, None)),

src/transformers/models/phi3/__init__.py

@@ -0,0 +1,69 @@
+# Copyright 2024 Microsoft 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 (
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    is_sentencepiece_available,
+    is_tokenizers_available,
+    is_torch_available,
+)
+
+
+_import_structure = {
+    "configuration_phi3": ["PHI3_PRETRAINED_CONFIG_ARCHIVE_MAP", "Phi3Config"],
+}
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_phi3"] = [
+        "PHI3_PRETRAINED_MODEL_ARCHIVE_LIST",
+        "Phi3PreTrainedModel",
+        "Phi3Model",
+        "Phi3ForCausalLM",
+        "Phi3ForSequenceClassification",
+        "Phi3ForTokenClassification",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_phi3 import PHI3_PRETRAINED_CONFIG_ARCHIVE_MAP, Phi3Config
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_phi3 import (
+            PHI3_PRETRAINED_MODEL_ARCHIVE_LIST,
+            Phi3ForCausalLM,
+            Phi3ForSequenceClassification,
+            Phi3ForTokenClassification,
+            Phi3Model,
+            Phi3PreTrainedModel,
+        )
+
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)