Cohere Model Release (#29622)

* Cohere Model Release (#1)

Cohere Model Release

* Remove unnecessary files and code (#2)

Some cleanup

* Delete cohere-model directory (#3)

* Make Fix (#5)

* Pr fixes (#6)

* fixes for pr

* pr fixes for the format

* pr fixes for the format

* src/transformers/models/auto/tokenization_auto.py

* Tokenizer test (#8)

* tokenizer test

* format fix

* Adding Docs and other minor changes (#7)

* Add modeling tests (#9)

* Smol Fix (#11)

* tokenization tests are fixed

* format fixes

* fix pr doc tests

* fix pr doc tests

* fix pr doc tests

* fix pr style check

* small changes in cohere.md

* FIX: Address final comments for transformers integration (#13)

* fix modeling final nits and add proper test file

* for now leave empty tests

* add integration test

* push new test

* fix modeling cohere (#14)

* Update chat templates to use the new API (#15)

---------
Co-authored-by: ahmetustun <ahmetustun89@gmail.com>
Co-authored-by: Younes Belkada <49240599+younesbelkada@users.noreply.github.com>
Co-authored-by: Matt <Rocketknight1@users.noreply.github.com>

src/transformers/models/auto/configuration_auto.py

@@ -62,6 +62,7 @@ CONFIG_MAPPING_NAMES = OrderedDict(
         ("clvp", "ClvpConfig"),
         ("code_llama", "LlamaConfig"),
         ("codegen", "CodeGenConfig"),
+        ("cohere", "CohereConfig"),
         ("conditional_detr", "ConditionalDetrConfig"),
         ("convbert", "ConvBertConfig"),
         ("convnext", "ConvNextConfig"),
@@ -303,6 +304,7 @@ CONFIG_ARCHIVE_MAP_MAPPING_NAMES = OrderedDict(
         ("clipseg", "CLIPSEG_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("clvp", "CLVP_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("codegen", "CODEGEN_PRETRAINED_CONFIG_ARCHIVE_MAP"),
+        ("cohere", "COHERE_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("conditional_detr", "CONDITIONAL_DETR_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("convbert", "CONVBERT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("convnext", "CONVNEXT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
@@ -533,6 +535,7 @@ MODEL_NAMES_MAPPING = OrderedDict(
         ("clvp", "CLVP"),
         ("code_llama", "CodeLlama"),
         ("codegen", "CodeGen"),
+        ("cohere", "Cohere"),
         ("conditional_detr", "Conditional DETR"),
         ("convbert", "ConvBERT"),
         ("convnext", "ConvNeXT"),

src/transformers/models/auto/modeling_auto.py

@@ -64,6 +64,7 @@ MODEL_MAPPING_NAMES = OrderedDict(
         ("clvp", "ClvpModelForConditionalGeneration"),
         ("code_llama", "LlamaModel"),
         ("codegen", "CodeGenModel"),
+        ("cohere", "CohereModel"),
         ("conditional_detr", "ConditionalDetrModel"),
         ("convbert", "ConvBertModel"),
         ("convnext", "ConvNextModel"),
@@ -426,6 +427,7 @@ MODEL_FOR_CAUSAL_LM_MAPPING_NAMES = OrderedDict(
         ("camembert", "CamembertForCausalLM"),
         ("code_llama", "LlamaForCausalLM"),
         ("codegen", "CodeGenForCausalLM"),
+        ("cohere", "CohereForCausalLM"),
         ("cpmant", "CpmAntForCausalLM"),
         ("ctrl", "CTRLLMHeadModel"),
         ("data2vec-text", "Data2VecTextForCausalLM"),

src/transformers/models/auto/tokenization_auto.py

@@ -137,6 +137,7 @@ else:
                 ),
             ),
             ("codegen", ("CodeGenTokenizer", "CodeGenTokenizerFast" if is_tokenizers_available() else None)),
+            ("cohere", (None, "CohereTokenizerFast" if is_tokenizers_available() else None)),
             ("convbert", ("ConvBertTokenizer", "ConvBertTokenizerFast" if is_tokenizers_available() else None)),
             (
                 "cpm",

src/transformers/models/cohere/__init__.py

+# Copyright 2024 Cohere 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_cohere": ["COHERE_PRETRAINED_CONFIG_ARCHIVE_MAP", "CohereConfig"],
+}
+
+
+try:
+    if not is_tokenizers_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["tokenization_cohere_fast"] = ["CohereTokenizerFast"]
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_cohere"] = [
+        "CohereForCausalLM",
+        "CohereModel",
+        "CoherePreTrainedModel",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_cohere import COHERE_PRETRAINED_CONFIG_ARCHIVE_MAP, CohereConfig
+
+    try:
+        if not is_tokenizers_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .tokenization_cohere_fast import CohereTokenizerFast
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_cohere import (
+            CohereForCausalLM,
+            CohereModel,
+            CoherePreTrainedModel,
+        )
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)

src/transformers/models/cohere/configuration_cohere.py

+# coding=utf-8
+# Copyright 2024 Cohere team. All rights reserved.
+#
+# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
+# and OPT implementations in this library. It has been modified from its
+# original forms to accommodate minor architectural differences compared
+# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
+#
+# 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.
+""" Cohere model configuration"""
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+COHERE_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
+
+
+class CohereConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`CohereModel`]. It is used to instantiate an Cohere
+    model according to the specified arguments, defining the model architecture.
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information. Instantiating a configuration
+    with the defaults will yield a similar configuration to that of the [CohereForAI/c4ai-command-r-v01](https://huggingface.co/CohereForAI/c4ai-command-r-v01) model.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 256000):
+            Vocabulary size of the Cohere model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`CohereModel`]
+        hidden_size (`int`, *optional*, defaults to 8192):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 22528):
+            Dimension of the MLP representations.
+        logit_scale (`float`, *optional*, defaults to 0.0625):
+            The scaling factor for the output logits.
+        num_hidden_layers (`int`, *optional*, defaults to 40):
+            Number of hidden layers in the Transformer decoder.
+        num_attention_heads (`int`, *optional*, defaults to 64):
+            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 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.
+        layer_norm_eps (`float`, *optional*, defaults to 1e-05):
+            The epsilon used by the layer normalization.
+        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.
+        bos_token_id (`int`, *optional*, defaults to 5):
+            Beginning of stream token id.
+        eos_token_id (`int`, *optional*, defaults to 255001):
+            End 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.
+
+    ```python
+    >>> from transformers import CohereModel, CohereConfig
+
+    >>> # Initializing a Cohere model configuration
+    >>> configuration = CohereConfig()
+
+    >>> # Initializing a model from the Cohere configuration
+    >>> model = CohereModel(configuration) # doctest: +SKIP
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config # doctest: +SKIP
+    ```"""
+
+    model_type = "cohere"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=256000,
+        hidden_size=8192,
+        intermediate_size=22528,
+        logit_scale=0.0625,
+        num_hidden_layers=40,
+        num_attention_heads=64,
+        num_key_value_heads=None,
+        hidden_act="silu",
+        max_position_embeddings=8192,
+        initializer_range=0.02,
+        layer_norm_eps=1e-5,
+        use_cache=True,
+        pad_token_id=0,
+        bos_token_id=5,
+        eos_token_id=255001,
+        tie_word_embeddings=True,
+        rope_theta=10000.0,
+        attention_bias=False,
+        attention_dropout=0.0,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.logit_scale = logit_scale
+        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.layer_norm_eps = layer_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,
+        )

src/transformers/utils/dummy_pt_objects.py

@@ -2056,6 +2056,27 @@ class CodeGenPreTrainedModel(metaclass=DummyObject):
         requires_backends(self, ["torch"])
 
 
+class CohereForCausalLM(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class CohereModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class CoherePreTrainedModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
 CONDITIONAL_DETR_PRETRAINED_MODEL_ARCHIVE_LIST = None
 
 

src/transformers/utils/dummy_tokenizers_objects.py

@@ -86,6 +86,13 @@ class CodeGenTokenizerFast(metaclass=DummyObject):
         requires_backends(self, ["tokenizers"])
 
 
+class CohereTokenizerFast(metaclass=DummyObject):
+    _backends = ["tokenizers"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["tokenizers"])
+
+
 class ConvBertTokenizerFast(metaclass=DummyObject):
     _backends = ["tokenizers"]
 

src/transformers/utils/fx.py

@@ -135,6 +135,7 @@ _REGULAR_SUPPORTED_MODEL_NAMES_AND_TASKS = [
     "hubert",
     "layoutlm",
     "llama",
+    "cohere",
     "lxmert",
     "m2m_100",
     "marian",

tests/models/cohere/test_modeling_cohere.py

+# coding=utf-8
+# Copyright 2022 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 Cohere model. """
+
+import unittest
+
+from parameterized import parameterized
+
+from transformers import CohereConfig, is_torch_available
+from transformers.testing_utils import (
+    require_bitsandbytes,
+    require_torch,
+    require_torch_multi_gpu,
+    require_torch_sdpa,
+    slow,
+    torch_device,
+)
+
+from ...generation.test_utils import GenerationTesterMixin
+from ...test_configuration_common import ConfigTester
+from ...test_modeling_common import ModelTesterMixin, ids_tensor
+from ...test_pipeline_mixin import PipelineTesterMixin
+
+
+if is_torch_available():
+    import torch
+
+    from transformers import AutoTokenizer, CohereForCausalLM, CohereModel
+
+
+# Copied from transformers.tests.models.llama.LlamaModelTester with Llama->Cohere
+class CohereModelTester:
+    def __init__(
+        self,
+        parent,
+        batch_size=13,
+        seq_length=7,
+        is_training=True,
+        use_input_mask=True,
+        use_token_type_ids=False,
+        use_labels=True,
+        vocab_size=99,
+        hidden_size=32,
+        num_hidden_layers=2,
+        num_attention_heads=4,
+        intermediate_size=37,
+        hidden_act="gelu",
+        hidden_dropout_prob=0.1,
+        attention_probs_dropout_prob=0.1,
+        max_position_embeddings=512,
+        type_vocab_size=16,
+        type_sequence_label_size=2,
+        initializer_range=0.02,
+        num_labels=3,
+        num_choices=4,
+        pad_token_id=0,
+        scope=None,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.seq_length = seq_length
+        self.is_training = is_training
+        self.use_input_mask = use_input_mask
+        self.use_token_type_ids = use_token_type_ids
+        self.use_labels = use_labels
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.intermediate_size = intermediate_size
+        self.hidden_act = hidden_act
+        self.hidden_dropout_prob = hidden_dropout_prob
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.max_position_embeddings = max_position_embeddings
+        self.type_vocab_size = type_vocab_size
+        self.type_sequence_label_size = type_sequence_label_size
+        self.initializer_range = initializer_range
+        self.num_labels = num_labels
+        self.num_choices = num_choices
+        self.pad_token_id = pad_token_id
+        self.scope = scope
+
+    def prepare_config_and_inputs(self):
+        input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
+
+        input_mask = None
+        if self.use_input_mask:
+            input_mask = torch.tril(torch.ones(self.batch_size, self.seq_length)).to(torch_device)
+
+        token_type_ids = None
+        if self.use_token_type_ids:
+            token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size)
+
+        sequence_labels = None
+        token_labels = None
+        choice_labels = None
+        if self.use_labels:
+            sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
+            token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
+            choice_labels = ids_tensor([self.batch_size], self.num_choices)
+
+        config = self.get_config()
+
+        return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
+
+    # Ignore copy
+    def get_config(self):
+        return CohereConfig(
+            vocab_size=self.vocab_size,
+            hidden_size=self.hidden_size,
+            num_hidden_layers=self.num_hidden_layers,
+            num_attention_heads=self.num_attention_heads,
+            intermediate_size=self.intermediate_size,
+            hidden_act=self.hidden_act,
+            hidden_dropout_prob=self.hidden_dropout_prob,
+            attention_probs_dropout_prob=self.attention_probs_dropout_prob,
+            max_position_embeddings=self.max_position_embeddings,
+            type_vocab_size=self.type_vocab_size,
+            is_decoder=False,
+            initializer_range=self.initializer_range,
+            pad_token_id=self.pad_token_id,
+            eos_token_id=self.pad_token_id,
+        )
+
+    def create_and_check_model(
+        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = CohereModel(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))
+
+    def create_and_check_model_as_decoder(
+        self,
+        config,
+        input_ids,
+        token_type_ids,
+        input_mask,
+        sequence_labels,
+        token_labels,
+        choice_labels,
+        encoder_hidden_states,
+        encoder_attention_mask,
+    ):
+        config.add_cross_attention = True
+        model = CohereModel(config)
+        model.to(torch_device)
+        model.eval()
+        result = model(
+            input_ids,
+            attention_mask=input_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+        )
+        result = model(
+            input_ids,
+            attention_mask=input_mask,
+            encoder_hidden_states=encoder_hidden_states,
+        )
+        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))
+
+    def create_and_check_for_causal_lm(
+        self,
+        config,
+        input_ids,
+        token_type_ids,
+        input_mask,
+        sequence_labels,
+        token_labels,
+        choice_labels,
+        encoder_hidden_states,
+        encoder_attention_mask,
+    ):
+        model = CohereForCausalLM(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))
+
+    def create_and_check_decoder_model_past_large_inputs(
+        self,
+        config,
+        input_ids,
+        token_type_ids,
+        input_mask,
+        sequence_labels,
+        token_labels,
+        choice_labels,
+        encoder_hidden_states,
+        encoder_attention_mask,
+    ):
+        config.is_decoder = True
+        config.add_cross_attention = True
+        model = CohereForCausalLM(config=config)
+        model.to(torch_device)
+        model.eval()
+
+        # first forward pass
+        outputs = model(
+            input_ids,
+            attention_mask=input_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            use_cache=True,
+        )
+        past_key_values = outputs.past_key_values
+
+        # create hypothetical multiple next token and extent to next_input_ids
+        next_tokens = ids_tensor((self.batch_size, 3), config.vocab_size)
+        next_mask = ids_tensor((self.batch_size, 3), vocab_size=2)
+
+        # append to next input_ids and
+        next_input_ids = torch.cat([input_ids, next_tokens], dim=-1)
+        next_attention_mask = torch.cat([input_mask, next_mask], dim=-1)
+
+        output_from_no_past = model(
+            next_input_ids,
+            attention_mask=next_attention_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            output_hidden_states=True,
+        )["hidden_states"][0]
+        output_from_past = model(
+            next_tokens,
+            attention_mask=next_attention_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            past_key_values=past_key_values,
+            output_hidden_states=True,
+        )["hidden_states"][0]
+
+        # select random slice
+        random_slice_idx = ids_tensor((1,), output_from_past.shape[-1]).item()
+        output_from_no_past_slice = output_from_no_past[:, -3:, random_slice_idx].detach()
+        output_from_past_slice = output_from_past[:, :, random_slice_idx].detach()
+
+        self.parent.assertTrue(output_from_past_slice.shape[1] == next_tokens.shape[1])
+
+        # test that outputs are equal for slice
+        self.parent.assertTrue(torch.allclose(output_from_past_slice, output_from_no_past_slice, atol=1e-3))
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        (
+            config,
+            input_ids,
+            token_type_ids,
+            input_mask,
+            sequence_labels,
+            token_labels,
+            choice_labels,
+        ) = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask}
+        return config, inputs_dict
+
+
+@require_torch
+class CohereModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin, unittest.TestCase):
+    all_model_classes = (CohereModel, CohereForCausalLM) if is_torch_available() else ()
+    all_generative_model_classes = (CohereForCausalLM,) if is_torch_available() else ()
+    pipeline_model_mapping = (
+        {
+            "feature-extraction": CohereModel,
+            "text-generation": CohereForCausalLM,
+        }
+        if is_torch_available()
+        else {}
+    )
+    test_headmasking = False
+    test_pruning = False
+    fx_compatible = True
+
+    # Need to use `0.8` instead of `0.9` for `test_cpu_offload`
+    # This is because we are hitting edge cases with the causal_mask buffer
+    model_split_percents = [0.5, 0.7, 0.8]
+
+    def setUp(self):
+        self.model_tester = CohereModelTester(self)
+        self.config_tester = ConfigTester(self, config_class=CohereConfig, hidden_size=37)
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    @unittest.skip("TODO @gante fix this for Cohere")
+    @parameterized.expand([(1, False), (1, True), (4, False)])
+    def test_new_cache_format(self, num_beams, do_sample):
+        pass
+
+    def test_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_model(*config_and_inputs)
+
+    def test_model_various_embeddings(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        for type in ["absolute", "relative_key", "relative_key_query"]:
+            config_and_inputs[0].position_embedding_type = type
+            self.model_tester.create_and_check_model(*config_and_inputs)
+
+    @require_bitsandbytes
+    @require_torch_sdpa
+    @require_torch_multi_gpu
+    @slow
+    def test_eager_matches_sdpa_generate(self):
+        """
+        Overwritting the common test as the test is flaky on tiny models
+        """
+        max_new_tokens = 30
+
+        model_id = "CohereForAI/c4ai-command-r-v01-4bit"
+        tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+        model_sdpa = CohereForCausalLM.from_pretrained(
+            model_id, torch_dtype=torch.float16, low_cpu_mem_usage=True, device_map="auto"
+        )
+        self.assertTrue(model_sdpa.config._attn_implementation == "sdpa")
+
+        model_eager = CohereForCausalLM.from_pretrained(
+            model_id, torch_dtype=torch.float16, attn_implementation="eager", device_map="auto"
+        )
+
+        self.assertTrue(model_eager.config._attn_implementation == "eager")
+
+        for name, submodule in model_eager.named_modules():
+            if "SdpaAttention" in submodule.__class__.__name__:
+                raise ValueError("The eager model should not have SDPA attention layers")
+
+        has_sdpa = False
+        for name, submodule in model_sdpa.named_modules():
+            if "SdpaAttention" in submodule.__class__.__name__:
+                has_sdpa = True
+                break
+        if not has_sdpa:
+            raise ValueError("The SDPA model should have SDPA attention layers")
+
+        texts = [
+            "hi here's a longer context, getting longer and",
+            "Hello this is a very long sentence my friend, very long for real",
+            "Today I am in Paris and",
+        ]
+
+        for padding_side in ["left", "right"]:
+            tokenizer.padding_side = padding_side
+            tokenizer.pad_token = tokenizer.eos_token
+
+            inputs = tokenizer(texts, return_tensors="pt", padding=True).to(torch_device)
+
+            res_eager = model_eager.generate(**inputs, max_new_tokens=max_new_tokens, do_sample=False)
+            res_sdpa = model_sdpa.generate(**inputs, max_new_tokens=max_new_tokens, do_sample=False)
+
+            with self.subTest(f"{padding_side}"):
+                torch.testing.assert_close(
+                    res_eager,
+                    res_sdpa,
+                    msg=f"\n{tokenizer.batch_decode(res_eager)} \nvs\n{tokenizer.batch_decode(res_sdpa)}",
+                )
+
+
+@require_torch
+@slow
+class CohereIntegrationTest(unittest.TestCase):
+    @require_torch_multi_gpu
+    def test_batched_4bit(self):
+        model_id = "CohereForAI/c4ai-command-r-v01-4bit"
+
+        EXPECTED_TEXT = [
+            'Hello today I am going to show you how to make a simple and easy card using the new stamp set called "Hello" from the Occasions catalog. This set is so versatile and can be used for many occasions. I used the new In',
+            "Hi there, here we are again with another great collection of free fonts. This time we have gathered 10 free fonts that you can download and use in your designs. These fonts are free for personal and commercial use. So",
+        ]
+
+        model = CohereForCausalLM.from_pretrained(model_id)
+        tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+        tokenizer.pad_token = tokenizer.eos_token
+
+        text = ["Hello today I am going to show you how to", "Hi there, here we are"]
+        inputs = tokenizer(text, return_tensors="pt", padding=True).to(torch_device)
+
+        output = model.generate(**inputs, max_new_tokens=40, do_sample=False)
+        self.assertEqual(tokenizer.batch_decode(output, skip_special_tokens=True), EXPECTED_TEXT)
+
+    def test_batched_small_model_logits(self):
+        # Since the model is very large, we created a random cohere model so that we can do a simple
+        # logits check on it.
+        model_id = "hf-internal-testing/cohere-random"
+
+        EXPECTED_LOGITS = torch.Tensor(
+            [
+                [[0.0000, 0.1866, -0.1997], [0.0000, -0.0736, 0.1785], [0.0000, -0.1965, -0.0569]],
+                [[0.0000, -0.0302, 0.1488], [0.0000, -0.0402, 0.1351], [0.0000, -0.0341, 0.1116]],
+            ]
+        ).to(torch_device)
+
+        tokenizer = AutoTokenizer.from_pretrained(model_id)
+        model = CohereForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True, torch_dtype=torch.float16).to(
+            torch_device
+        )
+
+        tokenizer.pad_token = tokenizer.eos_token
+
+        text = ["Hello today I am going to show you how to", "Hi there, here we are"]
+        inputs = tokenizer(text, return_tensors="pt", padding=True).to(torch_device)
+
+        with torch.no_grad():
+            output = model(**inputs)
+
+        logits = output.logits
+        self.assertTrue(torch.allclose(EXPECTED_LOGITS, logits[:, :3, :3], rtol=1e-3, atol=1e-3))

tests/models/cohere/test_tokenization_cohere.py

+# coding=utf-8
+# Copyright 2022 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.
+
+import unittest
+
+from transformers import CohereTokenizerFast
+from transformers.testing_utils import require_jinja, require_tokenizers
+
+from ...test_tokenization_common import TokenizerTesterMixin
+
+
+@require_tokenizers
+class CohereTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
+    slow_tokenizer_class = None
+    rust_tokenizer_class = CohereTokenizerFast
+    tokenizer_class = CohereTokenizerFast
+    test_rust_tokenizer = True
+    test_slow_tokenizer = False
+    from_pretrained_vocab_key = "tokenizer_file"
+    from_pretrained_id = "CohereForAI/c4ai-command-r-v01"
+    special_tokens_map = {
+        "bos_token": "<BOS_TOKEN>",
+        "eos_token": "<|END_OF_TURN_TOKEN|>",
+        "unk_token": "<UNK>",
+        "pad_token": "<PAD>",
+    }
+
+    def setUp(self):
+        super().setUp()
+        tokenizer = CohereTokenizerFast.from_pretrained("CohereForAI/c4ai-command-r-v01")
+        tokenizer.save_pretrained(self.tmpdirname)
+
+    def get_rust_tokenizer(self, **kwargs):
+        kwargs.update(self.special_tokens_map)
+        return CohereTokenizerFast.from_pretrained(self.tmpdirname, **kwargs)
+
+    @unittest.skip("This needs a slow tokenizer. Cohere does not have one!")
+    def test_encode_decode_with_spaces(self):
+        return
+
+    def test_encodings_from_sample_data(self):
+        """
+        Assert that the created tokens are the same than the hard-coded ones
+        """
+        tokenizer = self.get_rust_tokenizer()
+
+        INPUT_SENTENCES = ["The quick brown fox<|END_OF_TURN_TOKEN|>", "jumps over the lazy dog<|END_OF_TURN_TOKEN|>"]
+        TARGET_TOKENS = [[5, 2162, 6629, 19883, 73388, 255001], [5, 81, 25092, 2515, 1690, 46189, 9507, 255001]]
+
+        computed_tokens = tokenizer.batch_encode_plus(INPUT_SENTENCES)["input_ids"]
+        self.assertListEqual(TARGET_TOKENS, computed_tokens)
+
+        INPUT_SENTENCES_W_BOS = [
+            "<BOS_TOKEN>The quick brown fox<|END_OF_TURN_TOKEN|>",
+            "<BOS_TOKEN>jumps over the lazy dog<|END_OF_TURN_TOKEN|>",
+        ]
+        decoded_tokens = tokenizer.batch_decode(computed_tokens)
+        self.assertListEqual(decoded_tokens, INPUT_SENTENCES_W_BOS)
+
+    def test_padding(self, max_length=10):
+        for tokenizer, pretrained_name, kwargs in self.tokenizers_list:
+            with self.subTest(f"{tokenizer.__class__.__name__} ({pretrained_name})"):
+                tokenizer_r = self.rust_tokenizer_class.from_pretrained(pretrained_name, **kwargs)
+                # tokenizer_r.pad_token = None # Hotfixing padding = None
+                # Simple input
+                s = "This is a simple input"
+                s2 = ["This is a simple input 1", "This is a simple input 2"]
+                p = ("This is a simple input", "This is a pair")
+                p2 = [
+                    ("This is a simple input 1", "This is a simple input 2"),
+                    ("This is a simple pair 1", "This is a simple pair 2"),
+                ]
+
+                # Simple input tests
+                try:
+                    tokenizer_r.encode(s, max_length=max_length)
+                    tokenizer_r.encode_plus(s, max_length=max_length)
+
+                    tokenizer_r.batch_encode_plus(s2, max_length=max_length)
+                    tokenizer_r.encode(p, max_length=max_length)
+                    tokenizer_r.batch_encode_plus(p2, max_length=max_length)
+                except ValueError:
+                    self.fail("Cohere Tokenizer should be able to deal with padding")
+
+                tokenizer_r.pad_token = None  # Hotfixing padding = None
+                self.assertRaises(ValueError, tokenizer_r.encode, s, max_length=max_length, padding="max_length")
+
+                # Simple input
+                self.assertRaises(ValueError, tokenizer_r.encode_plus, s, max_length=max_length, padding="max_length")
+
+                # Simple input
+                self.assertRaises(
+                    ValueError,
+                    tokenizer_r.batch_encode_plus,
+                    s2,
+                    max_length=max_length,
+                    padding="max_length",
+                )
+
+                # Pair input
+                self.assertRaises(ValueError, tokenizer_r.encode, p, max_length=max_length, padding="max_length")
+
+                # Pair input
+                self.assertRaises(ValueError, tokenizer_r.encode_plus, p, max_length=max_length, padding="max_length")
+
+                # Pair input
+                self.assertRaises(
+                    ValueError,
+                    tokenizer_r.batch_encode_plus,
+                    p2,
+                    max_length=max_length,
+                    padding="max_length",
+                )
+
+    def test_pretrained_model_lists(self):
+        # No `max_model_input_sizes` for Cohere model
+        self.assertGreaterEqual(len(self.tokenizer_class.pretrained_vocab_files_map), 1)
+        self.assertGreaterEqual(len(list(self.tokenizer_class.pretrained_vocab_files_map.values())[0]), 1)
+
+    @require_jinja
+    def test_tokenization_for_chat(self):
+        tokenizer = self.get_rust_tokenizer()
+        test_chats = [
+            [{"role": "system", "content": "You are a helpful chatbot."}, {"role": "user", "content": "Hello!"}],
+            [
+                {"role": "system", "content": "You are a helpful chatbot."},
+                {"role": "user", "content": "Hello!"},
+                {"role": "assistant", "content": "Nice to meet you."},
+            ],
+        ]
+        tokenized_chats = [tokenizer.apply_chat_template(test_chat) for test_chat in test_chats]
+        expected_tokens = [
+            [5, 255000, 255008, 5659, 1955, 1671, 19264, 171597, 21, 255001, 255000, 255006, 28339, 8, 255001],
+            [
+                5,
+                255000,
+                255008,
+                5659,
+                1955,
+                1671,
+                19264,
+                171597,
+                21,
+                255001,
+                255000,
+                255006,
+                28339,
+                8,
+                255001,
+                255000,
+                255007,
+                97190,
+                1726,
+                5694,
+                1933,
+                21,
+                255001,
+            ],
+        ]
+        for tokenized_chat, expected_tokens in zip(tokenized_chats, expected_tokens):
+            self.assertListEqual(tokenized_chat, expected_tokens)
+
+    @require_jinja
+    def test_tokenization_for_tool_use(self):
+        tokenizer = self.get_rust_tokenizer()
+
+        conversation = [{"role": "user", "content": "Whats the biggest penguin in the world?"}]
+
+        tools = [
+            {
+                "name": "internet_search",
+                "description": "Returns a list of relevant document snippets for a textual query retrieved from the internet",
+                "parameter_definitions": {
+                    "query": {"description": "Query to search the internet with", "type": "str", "required": True}
+                },
+            },
+            {
+                "name": "directly_answer",
+                "description": "Calls a standard (un-augmented) AI chatbot to generate a response given the conversation history",
+                "parameter_definitions": {},
+            },
+        ]
+
+        tool_use_prompt = tokenizer.apply_tool_use_template(
+            conversation,
+            tools=tools,
+            tokenize=False,
+            add_generation_prompt=True,
+        )
+
+        expected_prompt = '''<BOS_TOKEN><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|># Safety Preamble
+The instructions in this section override those in the task description and style guide sections. Don't answer questions that are harmful or immoral.
+
+# System Preamble
+## Basic Rules
+You are a powerful conversational AI trained by Cohere to help people. You are augmented by a number of tools, and your job is to use and consume the output of these tools to best help the user. You will see a conversation history between yourself and a user, ending with an utterance from the user. You will then see a specific instruction instructing you what kind of response to generate. When you answer the user's requests, you cite your sources in your answers, according to those instructions.
+
+# User Preamble
+## Task and Context
+You help people answer their questions and other requests interactively. You will be asked a very wide array of requests on all kinds of topics. You will be equipped with a wide range of search engines or similar tools to help you, which you use to research your answer. You should focus on serving the user's needs as best you can, which will be wide-ranging.
+
+## Style Guide
+Unless the user asks for a different style of answer, you should answer in full sentences, using proper grammar and spelling.
+
+## Available Tools
+Here is a list of tools that you have available to you:
+
+```python
+def internet_search(query: str) -> List[Dict]:
+    """Returns a list of relevant document snippets for a textual query retrieved from the internet
+
+    Args:
+        query (str): Query to search the internet with
+    """
+    pass
+```
+
+```python
+def directly_answer() -> List[Dict]:
+    """Calls a standard (un-augmented) AI chatbot to generate a response given the conversation history
+    """
+    pass
+```<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Whats the biggest penguin in the world?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>Write 'Action:' followed by a json-formatted list of actions that you want to perform in order to produce a good response to the user's last input. You can use any of the supplied tools any number of times, but you should aim to execute the minimum number of necessary actions for the input. You should use the `directly-answer` tool if calling the other tools is unnecessary. The list of actions you want to call should be formatted as a list of json objects, for example:
+```json
+[
+    {
+        "tool_name": title of the tool in the specification,
+        "parameters": a dict of parameters to input into the tool as they are defined in the specs, or {} if it takes no parameters
+    }
+]```<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>'''
+
+        self.assertEqual(tool_use_prompt, expected_prompt)
+
+    @require_jinja
+    def test_tokenization_for_grounded_generation(self):
+        tokenizer = self.get_rust_tokenizer()
+        conversation = [{"role": "user", "content": "Whats the biggest penguin in the world?"}]
+
+        documents = [
+            {"title": "Tall penguins", "text": "Emperor penguins are the tallest growing up to 122 cm in height."},
+            {"title": "Penguin habitats", "text": "Emperor penguins only live in Antarctica."},
+        ]
+
+        grounded_generation_prompt = tokenizer.apply_grounded_generation_template(
+            conversation,
+            documents=documents,
+            citation_mode="accurate",  # or "fast"
+            tokenize=False,
+            add_generation_prompt=True,
+        )
+
+        expected_prompt = """<BOS_TOKEN><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|># Safety Preamble
+The instructions in this section override those in the task description and style guide sections. Don't answer questions that are harmful or immoral.
+
+# System Preamble
+## Basic Rules
+You are a powerful conversational AI trained by Cohere to help people. You are augmented by a number of tools, and your job is to use and consume the output of these tools to best help the user. You will see a conversation history between yourself and a user, ending with an utterance from the user. You will then see a specific instruction instructing you what kind of response to generate. When you answer the user's requests, you cite your sources in your answers, according to those instructions.
+
+# User Preamble
+## Task and Context
+You help people answer their questions and other requests interactively. You will be asked a very wide array of requests on all kinds of topics. You will be equipped with a wide range of search engines or similar tools to help you, which you use to research your answer. You should focus on serving the user's needs as best you can, which will be wide-ranging.
+
+## Style Guide
+Unless the user asks for a different style of answer, you should answer in full sentences, using proper grammar and spelling.<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Whats the biggest penguin in the world?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|><results>
+Document: 0
+title: Tall penguins
+text: Emperor penguins are the tallest growing up to 122 cm in height.
+
+Document: 1
+title: Penguin habitats
+text: Emperor penguins only live in Antarctica.
+</results><|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|SYSTEM_TOKEN|>Carefully perform the following instructions, in order, starting each with a new line.
+Firstly, Decide which of the retrieved documents are relevant to the user's last input by writing 'Relevant Documents:' followed by comma-separated list of document numbers. If none are relevant, you should instead write 'None'.
+Secondly, Decide which of the retrieved documents contain facts that should be cited in a good answer to the user's last input by writing 'Cited Documents:' followed a comma-separated list of document numbers. If you dont want to cite any of them, you should instead write 'None'.
+Thirdly, Write 'Answer:' followed by a response to the user's last input in high quality natural english. Use the retrieved documents to help you. Do not insert any citations or grounding markup.
+Finally, Write 'Grounded answer:' followed by a response to the user's last input in high quality natural english. Use the symbols <co: doc> and </co: doc> to indicate when a fact comes from a document in the search result, e.g <co: 0>my fact</co: 0> for a fact from document 0.<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>"""
+
+        self.assertEqual(grounded_generation_prompt, expected_prompt)
+
+    def test_add_prefix_space_fast(self):
+        tokenizer_w_prefix = self.get_rust_tokenizer(add_prefix_space=True)
+        tokenizer_wo_prefix = self.get_rust_tokenizer(add_prefix_space=False)
+        tokens_w_prefix = tokenizer_w_prefix.tokenize("Hey")
+        tokens_wo_prefix = tokenizer_wo_prefix.tokenize("Hey")
+        self.assertNotEqual(tokens_w_prefix, tokens_wo_prefix)