Cohere
Overview
The Cohere Command-R model was proposed in the blogpost Command-R: Retrieval Augmented Generation at Production Scale by the Cohere Team.
The abstract from the paper is the following:
Command-R is a scalable generative model targeting RAG and Tool Use to enable production-scale AI for enterprise. Today, we are introducing Command-R, a new LLM aimed at large-scale production workloads. Command-R targets the emerging “scalable” category of models that balance high efficiency with strong accuracy, enabling companies to move beyond proof of concept, and into production.
*Command-R is a generative model optimized for long context tasks such as retrieval augmented generation (RAG) and using external APIs and tools. It is designed to work in concert with our industry-leading Embed and Rerank models to provide best-in-class integration for RAG applications and excel at enterprise use cases. As a model built for companies to implement at scale, Command-R boasts:
- Strong accuracy on RAG and Tool Use
- Low latency, and high throughput
- Longer 128k context and lower pricing
- Strong capabilities across 10 key languages
- Model weights available on HuggingFace for research and evaluation
Checkout model checkpoints here. This model was contributed by Saurabh Dash and Ahmet Üstün. The code of the implementation in Hugging Face is based on GPT-NeoX here.
Usage tips
The checkpoints uploaded on the Hub use torch_dtype = 'float16', which will be
used by the AutoModel API to cast the checkpoints from torch.float32 to torch.float16.
The dtype of the online weights is mostly irrelevant unless you are using torch_dtype="auto" when initializing a model using model = AutoModelForCausalLM.from_pretrained("path", torch_dtype = "auto"). The reason is that the model will first be downloaded ( using the dtype of the checkpoints online), then it will be casted to the default dtype of torch (becomes torch.float32), and finally, if there is a torch_dtype provided in the config, it will be used.
Training the model in float16 is not recommended and is known to produce nan; as such, the model should be trained in bfloat16.
# pip install transformers
from transformers import AutoTokenizer, AutoModelForCausalLM
model_id = "CohereForAI/c4ai-command-r-v01"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)
# Format message with the command-r chat template
messages = [{"role": "user", "content": "Hello, how are you?"}]
input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
## <BOS_TOKEN><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
gen_tokens = model.generate(
input_ids,
max_new_tokens=100,
do_sample=True,
temperature=0.3,
)
gen_text = tokenizer.decode(gen_tokens[0])
print(gen_text)- When using Flash Attention 2 via
attn_implementation="flash_attention_2", don’t passtorch_dtypeto thefrom_pretrainedclass method and use Automatic Mixed-Precision training. When usingTrainer, it is simply specifying eitherfp16orbf16toTrue. Otherwise, make sure you are usingtorch.autocast. This is required because the Flash Attention only supportfp16andbf16data type.
Resources
A list of official Hugging Face and community (indicated by 🌎) resources to help you get started with Command-R. If you’re interested in submitting a resource to be included here, please feel free to open a Pull Request and we’ll review it! The resource should ideally demonstrate something new instead of duplicating an existing resource.
Loading FP16 model
# pip install transformers
from transformers import AutoTokenizer, AutoModelForCausalLM
model_id = "CohereForAI/c4ai-command-r-v01"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id)
# Format message with the command-r chat template
messages = [{"role": "user", "content": "Hello, how are you?"}]
input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
## <BOS_TOKEN><|START_OF_TURN_TOKEN|><|USER_TOKEN|>Hello, how are you?<|END_OF_TURN_TOKEN|><|START_OF_TURN_TOKEN|><|CHATBOT_TOKEN|>
gen_tokens = model.generate(
input_ids,
max_new_tokens=100,
do_sample=True,
temperature=0.3,
)
gen_text = tokenizer.decode(gen_tokens[0])
print(gen_text)Loading bitsnbytes 4bit quantized model
# pip install transformers bitsandbytes accelerate
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
bnb_config = BitsAndBytesConfig(load_in_4bit=True)
model_id = "CohereForAI/c4ai-command-r-v01"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, quantization_config=bnb_config)
gen_tokens = model.generate(
input_ids,
max_new_tokens=100,
do_sample=True,
temperature=0.3,
)
gen_text = tokenizer.decode(gen_tokens[0])
print(gen_text)CohereConfig
class transformers.CohereConfig
< source >( 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-05 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 use_qk_norm = False **kwargs )
Parameters
- vocab_size (
int, optional, defaults to 256000) — Vocabulary size of the Cohere model. Defines the number of different tokens that can be represented by theinputs_idspassed 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. Ifnum_key_value_heads=num_attention_heads, the model will use Multi Head Attention (MHA), ifnum_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 tonum_attention_heads`. - hidden_act (
strorfunction, 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 toTrue) — Whether or not the model should return the last key/values attentions (not used by all models). Only relevant ifconfig.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 toTrue) — Whether to tie weight embeddings - rope_theta (
float, optional, defaults to 10000.0) — The base period of the RoPE embeddings. - attention_bias (
bool, defaults toFalse, optional, defaults toFalse) — 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. - use_qk_norm (
bool, optional, defaults toFalse) — Whether to use query-key normalization in the attention
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 model.
>>> from transformers import CohereModel, CohereConfig
>>> # Initializing a Cohere model configuration
>>> configuration = CohereConfig()
>>> # Initializing a model from the Cohere configuration
>>> model = CohereModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.configCohereTokenizerFast
class transformers.CohereTokenizerFast
< source >( vocab_file = None merges_file = None tokenizer_file = None clean_up_tokenization_spaces = False unk_token = '<UNK>' bos_token = '<BOS_TOKEN>' eos_token = '<|END_OF_TURN_TOKEN|>' add_bos_token = True add_eos_token = False use_default_system_prompt = False add_prefix_space = False **kwargs )
Parameters
- vocab_file (
str, optional) — Path to the vocabulary file. - merges_file (
str, optional) — Path to the merges file. - tokenizer_file (
str, optional) — tokenizers file (generally has a .json extension) that contains everything needed to load the tokenizer. - clean_up_tokenization_spaces (
bool, optional, defaults toFalse) — Whether or not to cleanup spaces after decoding, cleanup consists in removing potential artifacts like extra spaces. - unk_token (
strortokenizers.AddedToken, optional, defaults to"<UNK>") — The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead. - bos_token (
strortokenizers.AddedToken, optional, defaults to"<BOS_TOKEN>") — The beginning of sequence token that was used during pretraining. Can be used a sequence classifier token. - eos_token (
strortokenizers.AddedToken, optional, defaults to"<|END_OF_TURN_TOKEN|>") — The end of sequence token. - add_bos_token (
bool, optional, defaults toTrue) — Whether or not to add anbos_tokenat the start of sequences. - add_eos_token (
bool, optional, defaults toFalse) — Whether or not to add aneos_tokenat the end of sequences. - use_default_system_prompt (
bool, optional, defaults toFalse) — Whether or not the default system prompt for Cohere tokenizer should be used. - add_prefix_space (
bool, optional, defaults toFalse) — Whether or not the tokenizer should automatically add a prefix space
Construct a Cohere tokenizer. Based on byte-level Byte-Pair-Encoding.
This uses notably ByteFallback and NFC normalization.
>>> from transformers import AutoTokenizer
>>> tokenizer = AutoTokenizer.from_pretrained("CohereForAI/c4ai-command-r-v01")
>>> tokenizer.encode("Hello this is a test")
[5, 28339, 2075, 1801, 1671, 3282]If you want to change the bos_token or the eos_token, make sure to specify them when initializing the model, or
call tokenizer.update_post_processor() to make sure that the post-processing is correctly done (otherwise the
values of the first token and final token of an encoded sequence will not be correct). For more details, checkout
[post-processors] (https://huggingface.co/docs/tokenizers/api/post-processors) documentation.
You can get around that behavior by passing add_prefix_space=True when instantiating this tokenizer, but since
the model was not pretrained this way, it might yield a decrease in performance.
When used with is_split_into_words=True, this tokenizer needs to be instantiated with add_prefix_space=True.
This tokenizer inherits from PreTrainedTokenizerFast which contains most of the main methods. Users should refer to this superclass for more information regarding those methods.
get_special_tokens_mask
< source >( token_ids_0: List token_ids_1: Optional = None already_has_special_tokens: bool = False ) → A list of integers in the range [0, 1]
Parameters
- token_ids_0 (
List[int]) — List of ids of the first sequence. - token_ids_1 (
List[int], optional) — List of ids of the second sequence. - already_has_special_tokens (
bool, optional, defaults toFalse) — Whether or not the token list is already formatted with special tokens for the model.
Returns
A list of integers in the range [0, 1]
1 for a special token, 0 for a sequence token.
Retrieves sequence ids from a token list that has no special tokens added. This method is called when adding
special tokens using the tokenizer prepare_for_model or encode_plus methods.
create_token_type_ids_from_sequences
< source >( token_ids_0: List token_ids_1: Optional = None ) → List[int]
Create the token type IDs corresponding to the sequences passed. What are token type IDs?
Should be overridden in a subclass if the model has a special way of building those.
Updates the underlying post processor with the current bos_token and eos_token.
save_vocabulary
< source >( save_directory: str filename_prefix: Optional = None ) → Tuple(str)
Save only the vocabulary of the tokenizer (vocabulary + added tokens).
This method won’t save the configuration and special token mappings of the tokenizer. Use
_save_pretrained() to save the whole state of the tokenizer.
CohereModel
class transformers.CohereModel
< source >( config: CohereConfig )
Parameters
- config (CohereConfig) — 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 from_pretrained() method to load the model weights. config — CohereConfig
The bare Cohere Model outputting raw hidden-states without any specific head on top. 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 subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.
Transformer decoder consisting of config.num_hidden_layers layers. Each layer is a CohereDecoderLayer
forward
< source >( input_ids: LongTensor = None attention_mask: Optional = None position_ids: Optional = None past_key_values: Optional = None inputs_embeds: Optional = None use_cache: Optional = None output_attentions: Optional = None output_hidden_states: Optional = None return_dict: Optional = None cache_position: Optional = None )
Parameters
- input_ids (
torch.LongTensorof 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.
- attention_mask (
torch.Tensorof 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.
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
If
past_key_valuesis used, optionally only the lastinput_idshave to be input (seepast_key_values).If you want to change padding behavior, you should read
modeling_opt._prepare_decoder_attention_maskand modify to your needs. See diagram 1 in the paper for more information on the default strategy.- 1 indicates the head is not masked,
- 0 indicates the head is masked.
- position_ids (
torch.LongTensorof 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]. - past_key_values (
Cacheortuple(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 thepast_key_valuesreturned by the model at a previous stage of decoding, whenuse_cache=Trueorconfig.use_cache=True.Two formats are allowed:
- a Cache instance;
- Tuple of
tuple(torch.FloatTensor)of lengthconfig.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_valuesare passed, the legacy cache format will be returned.If
past_key_valuesare used, the user can optionally input only the lastinput_ids(those that don’t have their past key value states given to this model) of shape(batch_size, 1)instead of allinput_idsof shape(batch_size, sequence_length). - inputs_embeds (
torch.FloatTensorof shape(batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passinginput_idsyou can choose to directly pass an embedded representation. This is useful if you want more control over how to convertinput_idsindices into associated vectors than the model’s internal embedding lookup matrix. - use_cache (
bool, optional) — If set toTrue,past_key_valueskey value states are returned and can be used to speed up decoding (seepast_key_values). - output_attentions (
bool, optional) — Whether or not to return the attentions tensors of all attention layers. Seeattentionsunder returned tensors for more detail. - output_hidden_states (
bool, optional) — Whether or not to return the hidden states of all layers. Seehidden_statesunder returned tensors for more detail. - return_dict (
bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
The CohereModel forward method, overrides the __call__ special method.
Although the recipe for forward pass needs to be defined within this function, one should call the Module
instance afterwards instead of this since the former takes care of running the pre and post processing steps while
the latter silently ignores them.
CohereForCausalLM
forward
< source >( input_ids: LongTensor = None attention_mask: Optional = None position_ids: Optional = None past_key_values: Optional = None inputs_embeds: Optional = None labels: Optional = None use_cache: Optional = None output_attentions: Optional = None output_hidden_states: Optional = None return_dict: Optional = None cache_position: Optional = None ) → transformers.modeling_outputs.CausalLMOutputWithPast or tuple(torch.FloatTensor)
Parameters
- input_ids (
torch.LongTensorof 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.
- attention_mask (
torch.Tensorof 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.
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
If
past_key_valuesis used, optionally only the lastinput_idshave to be input (seepast_key_values).If you want to change padding behavior, you should read
modeling_opt._prepare_decoder_attention_maskand modify to your needs. See diagram 1 in the paper for more information on the default strategy.- 1 indicates the head is not masked,
- 0 indicates the head is masked.
- position_ids (
torch.LongTensorof 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]. - past_key_values (
Cacheortuple(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 thepast_key_valuesreturned by the model at a previous stage of decoding, whenuse_cache=Trueorconfig.use_cache=True.Two formats are allowed:
- a Cache instance;
- Tuple of
tuple(torch.FloatTensor)of lengthconfig.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_valuesare passed, the legacy cache format will be returned.If
past_key_valuesare used, the user can optionally input only the lastinput_ids(those that don’t have their past key value states given to this model) of shape(batch_size, 1)instead of allinput_idsof shape(batch_size, sequence_length). - inputs_embeds (
torch.FloatTensorof shape(batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passinginput_idsyou can choose to directly pass an embedded representation. This is useful if you want more control over how to convertinput_idsindices into associated vectors than the model’s internal embedding lookup matrix. - use_cache (
bool, optional) — If set toTrue,past_key_valueskey value states are returned and can be used to speed up decoding (seepast_key_values). - output_attentions (
bool, optional) — Whether or not to return the attentions tensors of all attention layers. Seeattentionsunder returned tensors for more detail. - output_hidden_states (
bool, optional) — Whether or not to return the hidden states of all layers. Seehidden_statesunder returned tensors for more detail. - return_dict (
bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.Args — labels (
torch.LongTensorof 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 (seeinput_idsdocstring). Tokens with indices set to-100are ignored (masked), the loss is only computed for the tokens with labels in[0, ..., config.vocab_size].
Returns
transformers.modeling_outputs.CausalLMOutputWithPast or tuple(torch.FloatTensor)
A transformers.modeling_outputs.CausalLMOutputWithPast or a tuple of
torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various
elements depending on the configuration (CohereConfig) and inputs.
-
loss (
torch.FloatTensorof shape(1,), optional, returned whenlabelsis provided) — Language modeling loss (for next-token prediction). -
logits (
torch.FloatTensorof shape(batch_size, sequence_length, config.vocab_size)) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax). -
past_key_values (
tuple(tuple(torch.FloatTensor)), optional, returned whenuse_cache=Trueis passed or whenconfig.use_cache=True) — Tuple oftuple(torch.FloatTensor)of lengthconfig.n_layers, with each tuple having 2 tensors of shape(batch_size, num_heads, sequence_length, embed_size_per_head))Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see
past_key_valuesinput) to speed up sequential decoding. -
hidden_states (
tuple(torch.FloatTensor), optional, returned whenoutput_hidden_states=Trueis passed or whenconfig.output_hidden_states=True) — Tuple oftorch.FloatTensor(one for the output of the embeddings, if the model has an embedding layer, + one for the output of each layer) of shape(batch_size, sequence_length, hidden_size).Hidden-states of the model at the output of each layer plus the optional initial embedding outputs.
-
attentions (
tuple(torch.FloatTensor), optional, returned whenoutput_attentions=Trueis passed or whenconfig.output_attentions=True) — Tuple oftorch.FloatTensor(one for each layer) of shape(batch_size, num_heads, sequence_length, sequence_length).Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.
The CohereForCausalLM forward method, overrides the __call__ special method.
Although the recipe for forward pass needs to be defined within this function, one should call the Module
instance afterwards instead of this since the former takes care of running the pre and post processing steps while
the latter silently ignores them.
Example:
>> from transformers import AutoTokenizer, CohereForCausalLM
>> model = CohereForCausalLM.from_pretrained("CohereForAI/c4ai-command-r-v01")
>> tokenizer = AutoTokenizer.from_pretrained("CohereForAI/c4ai-command-r-v01")
>> prompt = "Hey, are you conscious? Can you talk to me?"
>> 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]
"Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."