Qwen2MoE is the new model series of large language models from the Qwen team. Previously, we released the Qwen series, including Qwen-72B, Qwen-1.8B, Qwen-VL, Qwen-Audio, etc.
Qwen2MoE is a language model series including decoder language models of different model sizes. For each size, we release the base language model and the aligned chat model. Qwen2MoE has the following architectural choices:
Qwen1.5-MoE-A2.7B
is upcycled from Qwen-1.8B
. It has 14.3B parameters in total and 2.7B activated parameters during runtime, while it achieves comparable performance with Qwen1.5-7B
, with only 25% of the training resources.For more details refer to the release blog post.
Qwen1.5-MoE-A2.7B
and Qwen1.5-MoE-A2.7B-Chat
can be found on the Huggingface Hub
In the following, we demonstrate how to use Qwen1.5-MoE-A2.7B-Chat
for the inference. Note that we have used the ChatML format for dialog, in this demo we show how to leverage apply_chat_template
for this purpose.
>>> from transformers import AutoModelForCausalLM, AutoTokenizer
>>> device = "cuda" # the device to load the model onto
>>> model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen1.5-MoE-A2.7B-Chat", device_map="auto")
>>> tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen1.5-MoE-A2.7B-Chat")
>>> prompt = "Give me a short introduction to large language model."
>>> messages = [{"role": "user", "content": prompt}]
>>> text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
>>> model_inputs = tokenizer([text], return_tensors="pt").to(device)
>>> generated_ids = model.generate(model_inputs.input_ids, max_new_tokens=512, do_sample=True)
>>> generated_ids = [output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)]
>>> response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
( vocab_size = 151936 hidden_size = 2048 intermediate_size = 5632 num_hidden_layers = 24 num_attention_heads = 16 num_key_value_heads = 16 hidden_act = 'silu' max_position_embeddings = 32768 initializer_range = 0.02 rms_norm_eps = 1e-06 use_cache = True tie_word_embeddings = False rope_theta = 10000.0 use_sliding_window = False sliding_window = 4096 max_window_layers = 28 attention_dropout = 0.0 decoder_sparse_step = 1 moe_intermediate_size = 1408 shared_expert_intermediate_size = 5632 num_experts_per_tok = 4 num_experts = 60 norm_topk_prob = False output_router_logits = False router_aux_loss_coef = 0.001 **kwargs )
Parameters
int
, optional, defaults to 151936) —
Vocabulary size of the Qwen2MoE model. Defines the number of different tokens that can be represented by the
inputs_ids
passed when calling Qwen2MoeModel int
, optional, defaults to 2048) —
Dimension of the hidden representations. int
, optional, defaults to 5632) —
Dimension of the MLP representations. int
, optional, defaults to 24) —
Number of hidden layers in the Transformer encoder. int
, optional, defaults to 16) —
Number of attention heads for each attention layer in the Transformer encoder. int
, optional, defaults to 16) —
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
32`. str
or function
, optional, defaults to "silu"
) —
The non-linear activation function (function or string) in the decoder. int
, optional, defaults to 32768) —
The maximum sequence length that this model might ever be used with. float
, optional, defaults to 0.02) —
The standard deviation of the truncated_normal_initializer for initializing all weight matrices. float
, optional, defaults to 1e-06) —
The epsilon used by the rms normalization layers. 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
. bool
, optional, defaults to False
) —
Whether the model’s input and output word embeddings should be tied. float
, optional, defaults to 10000.0) —
The base period of the RoPE embeddings. bool
, optional, defaults to False
) —
Whether to use sliding window attention. int
, optional, defaults to 4096) —
Sliding window attention (SWA) window size. If not specified, will default to 4096
. int
, optional, defaults to 28) —
The number of layers that use SWA (Sliding Window Attention). The bottom layers use SWA while the top use full attention. float
, optional, defaults to 0.0) —
The dropout ratio for the attention probabilities. int
, optional, defaults to 1) —
The frequency of the MoE layer. int
, optional, defaults to 1408) —
Intermediate size of the routed expert. int
, optional, defaults to 5632) —
Intermediate size of the shared expert. int
, optional, defaults to 4) —
Number of selected experts. int
, optional, defaults to 60) —
Number of routed experts. bool
, optional, defaults to False
) —
Whether to normalize the topk probabilities. bool
, optional, defaults to False
) —
Whether or not the router logits should be returned by the model. Enabeling this will also
allow the model to output the auxiliary loss, including load balancing loss and router z-loss. float
, optional, defaults to 0.001) —
The aux loss factor for the total loss. This is the configuration class to store the configuration of a Qwen2MoeModel. It is used to instantiate a Qwen2MoE model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of Qwen1.5-MoE-A2.7B” Qwen/Qwen1.5-MoE-A2.7B”.
Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.
>>> from transformers import Qwen2MoeModel, Qwen2MoeConfig
>>> # Initializing a Qwen2MoE style configuration
>>> configuration = Qwen2MoeConfig()
>>> # Initializing a model from the Qwen1.5-MoE-A2.7B" style configuration
>>> model = Qwen2MoeModel(configuration)
>>> # Accessing the model configuration
>>> configuration = model.config
( config: Qwen2MoeConfig )
Parameters
The bare Qwen2MoE 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 Qwen2MoeDecoderLayer
( 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 output_router_logits: Optional = None return_dict: Optional = None )
Parameters
torch.LongTensor
of 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.
torch.Tensor
of shape (batch_size, sequence_length)
, optional) —
Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]
:
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
If past_key_values
is used, optionally only the last decoder_input_ids
have to be input (see
past_key_values
).
If you want to change padding behavior, you should read modeling_opt._prepare_decoder_attention_mask
and modify to your needs. See diagram 1 in the paper for more
information on the default strategy.
torch.LongTensor
of 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]
.
Cache
or tuple(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 the past_key_values
returned by the model at a previous stage of decoding, when use_cache=True
or config.use_cache=True
.
Two formats are allowed:
tuple(torch.FloatTensor)
of length config.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_values
are passed, the
legacy cache format will be returned.
If past_key_values
are used, the user can optionally input only the last input_ids
(those that don’t
have their past key value states given to this model) of shape (batch_size, 1)
instead of all input_ids
of shape (batch_size, sequence_length)
.
torch.FloatTensor
of shape (batch_size, sequence_length, hidden_size)
, optional) —
Optionally, instead of passing input_ids
you can choose to directly pass an embedded representation. This
is useful if you want more control over how to convert input_ids
indices into associated vectors than the
model’s internal embedding lookup matrix. bool
, optional) —
If set to True
, past_key_values
key value states are returned and can be used to speed up decoding (see
past_key_values
). bool
, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions
under returned
tensors for more detail. bool
, optional) —
Whether or not to return the hidden states of all layers. See hidden_states
under returned tensors for
more detail. bool
, optional) —
Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
should not be returned during inference. bool
, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. The Qwen2MoeModel 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.
( 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 output_router_logits: Optional = None return_dict: Optional = None ) → transformers.modeling_outputs.MoeCausalLMOutputWithPast
or tuple(torch.FloatTensor)
Parameters
torch.LongTensor
of 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.
torch.Tensor
of shape (batch_size, sequence_length)
, optional) —
Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]
:
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
If past_key_values
is used, optionally only the last decoder_input_ids
have to be input (see
past_key_values
).
If you want to change padding behavior, you should read modeling_opt._prepare_decoder_attention_mask
and modify to your needs. See diagram 1 in the paper for more
information on the default strategy.
torch.LongTensor
of 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]
.
Cache
or tuple(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 the past_key_values
returned by the model at a previous stage of decoding, when use_cache=True
or config.use_cache=True
.
Two formats are allowed:
tuple(torch.FloatTensor)
of length config.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_values
are passed, the
legacy cache format will be returned.
If past_key_values
are used, the user can optionally input only the last input_ids
(those that don’t
have their past key value states given to this model) of shape (batch_size, 1)
instead of all input_ids
of shape (batch_size, sequence_length)
.
torch.FloatTensor
of shape (batch_size, sequence_length, hidden_size)
, optional) —
Optionally, instead of passing input_ids
you can choose to directly pass an embedded representation. This
is useful if you want more control over how to convert input_ids
indices into associated vectors than the
model’s internal embedding lookup matrix. bool
, optional) —
If set to True
, past_key_values
key value states are returned and can be used to speed up decoding (see
past_key_values
). bool
, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions
under returned
tensors for more detail. bool
, optional) —
Whether or not to return the hidden states of all layers. See hidden_states
under returned tensors for
more detail. bool
, optional) —
Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
should not be returned during inference. bool
, optional) —
Whether or not to return a ModelOutput instead of a plain tuple.
Args —
labels (torch.LongTensor
of 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 (see input_ids
docstring). Tokens with indices set to -100
are ignored
(masked), the loss is only computed for the tokens with labels in [0, ..., config.vocab_size]
.
Returns
transformers.modeling_outputs.MoeCausalLMOutputWithPast
or tuple(torch.FloatTensor)
A transformers.modeling_outputs.MoeCausalLMOutputWithPast
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 (Qwen2MoeConfig) and inputs.
loss (torch.FloatTensor
of shape (1,)
, optional, returned when labels
is provided) — Language modeling loss (for next-token prediction).
logits (torch.FloatTensor
of shape (batch_size, sequence_length, config.vocab_size)
) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).
aux_loss (torch.FloatTensor
, optional, returned when labels
is provided) — aux_loss for the sparse modules.
router_logits (tuple(torch.FloatTensor)
, optional, returned when output_router_probs=True
and config.add_router_probs=True
is passed or when config.output_router_probs=True
) — Tuple of torch.FloatTensor
(one for each layer) of shape (batch_size, sequence_length, num_experts)
.
Raw router logtis (post-softmax) that are computed by MoE routers, these terms are used to compute the auxiliary loss for Mixture of Experts models.
past_key_values (tuple(tuple(torch.FloatTensor))
, optional, returned when use_cache=True
is passed or when config.use_cache=True
) — Tuple of tuple(torch.FloatTensor)
of length config.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_values
input) to speed up sequential decoding.
hidden_states (tuple(torch.FloatTensor)
, optional, returned when output_hidden_states=True
is passed or when config.output_hidden_states=True
) — Tuple of torch.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 when output_attentions=True
is passed or when config.output_attentions=True
) — Tuple of torch.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 Qwen2MoeForCausalLM 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, Qwen2MoeForCausalLM
>>> model = Qwen2MoeForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS)
>>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER)
>>> 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."
( config )
Parameters
The Qwen2MoE Model transformer with a sequence classification head on top (linear layer).
Qwen2MoeForSequenceClassification uses the last token in order to do the classification, as other causal models (e.g. GPT-2) do.
Since it does classification on the last token, it requires to know the position of the last token. If a
pad_token_id
is defined in the configuration, it finds the last token that is not a padding token in each row. If
no pad_token_id
is defined, it simply takes the last value in each row of the batch. Since it cannot guess the
padding tokens when inputs_embeds
are passed instead of input_ids
, it does the same (take the last value in
each row of the batch).
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.
( input_ids: LongTensor = None attention_mask: Optional = None position_ids: Optional = None past_key_values: Union = 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 )
Parameters
torch.LongTensor
of 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.
torch.Tensor
of shape (batch_size, sequence_length)
, optional) —
Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]
:
Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.
If past_key_values
is used, optionally only the last decoder_input_ids
have to be input (see
past_key_values
).
If you want to change padding behavior, you should read modeling_opt._prepare_decoder_attention_mask
and modify to your needs. See diagram 1 in the paper for more
information on the default strategy.
torch.LongTensor
of 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]
.
Cache
or tuple(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 the past_key_values
returned by the model at a previous stage of decoding, when use_cache=True
or config.use_cache=True
.
Two formats are allowed:
tuple(torch.FloatTensor)
of length config.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_values
are passed, the
legacy cache format will be returned.
If past_key_values
are used, the user can optionally input only the last input_ids
(those that don’t
have their past key value states given to this model) of shape (batch_size, 1)
instead of all input_ids
of shape (batch_size, sequence_length)
.
torch.FloatTensor
of shape (batch_size, sequence_length, hidden_size)
, optional) —
Optionally, instead of passing input_ids
you can choose to directly pass an embedded representation. This
is useful if you want more control over how to convert input_ids
indices into associated vectors than the
model’s internal embedding lookup matrix. bool
, optional) —
If set to True
, past_key_values
key value states are returned and can be used to speed up decoding (see
past_key_values
). bool
, optional) —
Whether or not to return the attentions tensors of all attention layers. See attentions
under returned
tensors for more detail. bool
, optional) —
Whether or not to return the hidden states of all layers. See hidden_states
under returned tensors for
more detail. bool
, optional) —
Whether or not to return the logits of all the routers. They are useful for computing the router loss, and
should not be returned during inference. bool
, optional) —
Whether or not to return a ModelOutput instead of a plain tuple. torch.LongTensor
of shape (batch_size,)
, optional) —
Labels for computing the sequence classification/regression loss. Indices should be in [0, ..., config.num_labels - 1]
. If config.num_labels == 1
a regression loss is computed (Mean-Square loss), If
config.num_labels > 1
a classification loss is computed (Cross-Entropy). The Qwen2MoeForSequenceClassification 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.