Add Zamba2

docs/source/en/index.md

@@ -369,6 +369,7 @@ Flax), PyTorch, and/or TensorFlow.
 |                         [YOLOS](model_doc/yolos)                         |       ✅        |         ❌         |      ❌      |
 |                          [YOSO](model_doc/yoso)                          |       ✅        |         ❌         |      ❌      |
 |                         [Zamba](model_doc/zamba)                         |       ✅        |         ❌         |      ❌      |
+|                        [Zamba2](model_doc/zamba2)                        |       ✅        |         ❌         |      ❌      |
 |                      [ZoeDepth](model_doc/zoedepth)                      |       ✅        |         ❌         |      ❌      |
 
 <!-- End table-->

docs/source/en/model_doc/zamba2.md

@@ -0,0 +1,93 @@
+<!--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.
+
+-->
+# Zamba2
+
+Zamba2 is a large language model (LLM) trained by Zyphra, and made available under an Apache 2.0 license. Please see the [Zyphra Hugging Face](https://huggingface.co/collections/zyphra/) repository for model weights.
+
+This model was contributed by [pglo](https://huggingface.co/pglo).
+
+
+## Model details
+
+Zamba2-1.2B, Zamba2-2.7B and Zamba2-7B are hybrid models combining state-space models (Specifically [Mamba](https://github.com/state-spaces/mamba)) and transformer, and were trained using next-token prediction. Zamba2 uses shared transformer layers after every 6 mamba blocks. It uses the [Mistral v0.1 tokenizer](https://huggingface.co/mistralai/Mistral-7B-v0.1). We came to this architecture after a series of ablations at small scales. Zamba2-1.2B, Zamba2-2.7B and Zamba2-7B were pre-trained on 2T and 3T tokens, respectively.
+
+<img src=https://github.com/user-attachments/assets/c2cff209-b901-483c-87aa-774b82a0769f width=30% height=40% />
+
+## Quick start
+
+
+### Presequities
+
+Zamba2 requires you use `transformers` version 4.46.0 or higher:
+```bash
+pip install transformers>=4.46.0
+```
+
+## Inference
+
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+tokenizer = AutoTokenizer.from_pretrained("Zyphra/Zamba2-7B")
+model = AutoModelForCausalLM.from_pretrained("Zyphra/Zamba2-7B", device_map="cuda", torch_dtype=torch.bfloat16)
+
+input_text = "What factors contributed to the fall of the Roman Empire?"
+input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")
+
+outputs = model.generate(**input_ids, max_new_tokens=100)
+print(tokenizer.decode(outputs[0]))
+```
+
+
+## Model card
+
+The model cards can be found at:
+* [Zamba2-1.2B](https://huggingface.co/Zyphra/Zamba2-1.2B)
+* [Zamba2-2.7B](https://huggingface.co/Zyphra/Zamba2-2.7B)
+* [Zamba2-7B](https://huggingface.co/Zyphra/Zamba2-7B)
+
+
+## Issues
+For issues with model output, or community discussion, please use the Hugging Face community [forum](https://huggingface.co/Zyphra/Zamba2-7B/discussions)
+
+
+## License
+
+The model weights are open-sourced via an Apache 2.0 license.
+
+
+## Zamba2Config
+
+[[autodoc]] Zamba2Config
+
+
+## Zamba2Model
+
+[[autodoc]] Zamba2Model
+    - forward
+
+
+## Zamba2ForCausalLM
+
+[[autodoc]] Zamba2ForCausalLM
+    - forward
+
+
+## Zamba2ForSequenceClassification
+
+[[autodoc]] transformers.Zamba2ForSequenceClassification
+    - forward

docs/source/en/perf_infer_gpu_one.md

@@ -103,6 +103,7 @@ FlashAttention-2 is currently supported for the following architectures:
 * [SigLIP](https://huggingface.co/docs/transformers/model_doc/siglip)
 * [UniSpeech](https://huggingface.co/docs/transformers/v4.39.3/en/model_doc/unispeech#transformers.UniSpeechModel)
 * [unispeech_sat](https://huggingface.co/docs/transformers/v4.39.3/en/model_doc/unispeech-sat#transformers.UniSpeechSatModel)
+* [Zamba2](https://huggingface.co/docs/transformers/model_doc/zamba2)
 
 You can request to add FlashAttention-2 support for another model by opening a GitHub Issue or Pull Request.
 
@@ -309,7 +310,7 @@ For now, Transformers supports SDPA inference and training for the following arc
 * [XLM-RoBERTa](https://huggingface.co/docs/transformers/model_doc/xlm-roberta#transformers.XLMRobertaModel)
 * [XLM-RoBERTa-XL](https://huggingface.co/docs/transformers/model_doc/xlm-roberta-xl#transformers.XLMRobertaXLModel)
 * [YOLOS](https://huggingface.co/docs/transformers/model_doc/yolos#transformers.YolosModel)
-
+* [Zamba2](https://huggingface.co/docs/transformers/model_doc/zamba2)
 <Tip>
 
 FlashAttention can only be used for models with the `fp16` or `bf16` torch type, so make sure to cast your model to the appropriate type first. The memory-efficient attention backend is able to handle `fp32` models.

src/transformers/__init__.py

@@ -863,6 +863,7 @@
     "models.yolos": ["YolosConfig"],
     "models.yoso": ["YosoConfig"],
     "models.zamba": ["ZambaConfig"],
+    "models.zamba2": ["Zamba2Config"],
     "models.zoedepth": ["ZoeDepthConfig"],
     "onnx": [],
     "pipelines": [
@@ -3842,6 +3843,14 @@
             "ZambaPreTrainedModel",
         ]
     )
+    _import_structure["models.zamba2"].extend(
+        [
+            "Zamba2ForCausalLM",
+            "Zamba2ForSequenceClassification",
+            "Zamba2Model",
+            "Zamba2PreTrainedModel",
+        ]
+    )
     _import_structure["models.zoedepth"].extend(
         [
             "ZoeDepthForDepthEstimation",
@@ -5826,6 +5835,7 @@
     from .models.yolos import YolosConfig
     from .models.yoso import YosoConfig
     from .models.zamba import ZambaConfig
+    from .models.zamba2 import Zamba2Config
     from .models.zoedepth import ZoeDepthConfig
 
     # Pipelines
@@ -8275,6 +8285,12 @@
             ZambaModel,
             ZambaPreTrainedModel,
         )
+        from .models.zamba2 import (
+            Zamba2ForCausalLM,
+            Zamba2ForSequenceClassification,
+            Zamba2Model,
+            Zamba2PreTrainedModel,
+        )
         from .models.zoedepth import (
             ZoeDepthForDepthEstimation,
             ZoeDepthPreTrainedModel,

src/transformers/models/__init__.py

@@ -288,5 +288,6 @@
     yolos,
     yoso,
     zamba,
+    zamba2,
     zoedepth,
 )

src/transformers/models/auto/configuration_auto.py

@@ -319,6 +319,7 @@
         ("yolos", "YolosConfig"),
         ("yoso", "YosoConfig"),
         ("zamba", "ZambaConfig"),
+        ("zamba2", "Zamba2Config"),
         ("zoedepth", "ZoeDepthConfig"),
     ]
 )
@@ -647,6 +648,7 @@
         ("yolos", "YOLOS"),
         ("yoso", "YOSO"),
         ("zamba", "Zamba"),
+        ("zamba2", "Zamba2"),
         ("zoedepth", "ZoeDepth"),
     ]
 )

src/transformers/models/auto/modeling_auto.py

@@ -292,6 +292,7 @@
         ("yolos", "YolosModel"),
         ("yoso", "YosoModel"),
         ("zamba", "ZambaModel"),
+        ("zamba2", "Zamba2Model"),
     ]
 )
 
@@ -559,6 +560,7 @@
         ("xlnet", "XLNetLMHeadModel"),
         ("xmod", "XmodForCausalLM"),
         ("zamba", "ZambaForCausalLM"),
+        ("zamba2", "Zamba2ForCausalLM"),
     ]
 )
 
@@ -1018,6 +1020,7 @@
         ("xmod", "XmodForSequenceClassification"),
         ("yoso", "YosoForSequenceClassification"),
         ("zamba", "ZambaForSequenceClassification"),
+        ("zamba2", "Zamba2ForSequenceClassification"),
     ]
 )
 

src/transformers/models/auto/tokenization_auto.py

@@ -568,6 +568,13 @@
                     "LlamaTokenizerFast" if is_tokenizers_available() else None,
                 ),
             ),
+            (
+                "zamba2",
+                (
+                    "LlamaTokenizer" if is_sentencepiece_available() else None,
+                    "LlamaTokenizerFast" if is_tokenizers_available() else None,
+                ),
+            ),
         ]
     )