dmx.compressor.modeling.hf.DmxPreTrainedModel

class dmx.compressor.modeling.hf.DmxPreTrainedModel(config: PretrainedConfig, *inputs, **kwargs)

__init__(config: PretrainedConfig, *inputs, **kwargs)

Initialize internal Module state, shared by both nn.Module and ScriptModule.

Methods

__init__(config, *inputs, **kwargs)	Initialize internal Module state, shared by both nn.Module and ScriptModule.
active_adapter()
active_adapters()	If you are not familiar with adapters and PEFT methods, we invite you to read more about them on the PEFT official documentation: https://huggingface.co/docs/peft
add_adapter(adapter_config[, adapter_name])	If you are not familiar with adapters and PEFT methods, we invite you to read more about them on the PEFT official documentation: https://huggingface.co/docs/peft
add_memory_hooks()	Add a memory hook before and after each sub-module forward pass to record increase in memory consumption.
add_model_tags(tags)	Add custom tags into the model that gets pushed to the Hugging Face Hub.
add_module(name, module)	Add a child module to the current module.
apply(fn)	Apply fn recursively to every submodule (as returned by .children()) as well as self.
bfloat16()	Casts all floating point parameters and buffers to bfloat16 datatype.
buffers([recurse])	Return an iterator over module buffers.
can_generate()	Returns whether this model can generate sequences with .generate().
check_dim_consistency()	A function that checks format dimension consistency and sparseness dimension consistency for all applicable dmx modules in the model
children()	Return an iterator over immediate children modules.
compile(*args, **kwargs)	Compile this Module's forward using torch.compile().
compute_transition_scores(sequences, scores)	Computes the transition scores of sequences given the generation scores (and beam indices, if beam search was used).
configure(config, *rules)	Configure Dmx-specific numerics/sparsity/logics
counting_flops([zero])
cpu()	Move all model parameters and buffers to the CPU.
create_extended_attention_mask_for_decoder(...)
create_submod_transform_forward(model, ...)	Only supported for fx path, submodule forward can be directly called in export path
cuda([device])	Move all model parameters and buffers to the GPU.
deepcopy_args(args)
delete_adapter(adapter_names)	Delete an adapter's LoRA layers from the underlying model.
dequantize()	Potentially dequantize the model in case it has been quantized by a quantization method that support dequantization.
disable_adapters()	If you are not familiar with adapters and PEFT methods, we invite you to read more about them on the PEFT official documentation: https://huggingface.co/docs/peft
disable_input_require_grads()	Removes the _require_grads_hook.
double()	Casts all floating point parameters and buffers to double datatype.
enable_adapters()	If you are not familiar with adapters and PEFT methods, we invite you to read more about them on the PEFT official documentation: https://huggingface.co/docs/peft
enable_input_require_grads()	Enables the gradients for the input embeddings.
estimate_tokens(input_dict)	Helper function to estimate the total number of tokens from the model inputs.
eval()	Set the module in evaluation mode.
extra_repr()	Return the extra representation of the module.
float(*args)	Casts all floating point parameters and buffers to float datatype.
floating_point_ops(input_dict[, ...])	Get number of (optionally, non-embeddings) floating-point operations for the forward and backward passes of a batch with this transformer model.
fold_weights_and_biases()	A function that applies the ops the weights and biases using the corresponding formats.
forward(*input)	Define the computation performed at every call.
forward_weight_hypernets()
freeze([config_file])	A function that stores the state and ops format of the model to a config file
from_pretrained(*args, **kwargs)	Instantiate a pretrained pytorch model from a pre-trained model configuration.
generate([inputs, generation_config, ...])	Generates sequences of token ids for models with a language modeling head.
get_adapter_state_dict([adapter_name])	If you are not familiar with adapters and PEFT methods, we invite you to read more about them on the PEFT official documentation: https://huggingface.co/docs/peft
get_buffer(target)	Return the buffer given by target if it exists, otherwise throw an error.
get_compiled_call(compile_config)	Return a torch.compile'd version of self.__call__.
get_extended_attention_mask(attention_mask, ...)	Makes broadcastable attention and causal masks so that future and masked tokens are ignored.
get_extra_state()	Return any extra state to include in the module's state_dict.
get_head_mask(head_mask, num_hidden_layers)	Prepare the head mask if needed.
get_input_embeddings()	Returns the model's input embeddings.
get_memory_footprint([return_buffers])	Get the memory footprint of a model.
get_monitoring_records([submodules_to_monitor])
get_output_embeddings()	Returns the model's output embeddings.
get_parameter(target)	Return the parameter given by target if it exists, otherwise throw an error.
get_position_embeddings()
get_runtime_records()
get_submodule(target)	Return the submodule given by target if it exists, otherwise throw an error.
gradient_checkpointing_disable()	Deactivates gradient checkpointing for the current model.
gradient_checkpointing_enable([...])	Activates gradient checkpointing for the current model.
half(*args)	Casts all floating point parameters and buffers to half datatype.
heal_tokens(input_ids[, tokenizer])	Generates sequences of token ids for models with a language modeling head.
init_weights()	If needed prunes and maybe initializes weights.
invert_attention_mask(encoder_attention_mask)	Invert an attention mask (e.g., switches 0.
ipu([device])	Move all model parameters and buffers to the IPU.
is_backend_compatible()
is_same_signature(_model, args, kwargs)
keep_dmx_config()
load_adapter([peft_model_id, adapter_name, ...])	Load adapter weights from file or remote Hub folder.
load_state_dict(state_dict, *[, strict, assign])
measure_runtimes(device[, submodules_to_measure])
modules()	Return an iterator over all modules in the network.
monitoring([submodules_to_monitor, ...])
mtia([device])	Move all model parameters and buffers to the MTIA.
named_buffers([prefix, recurse, ...])	Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.
named_children()	Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.
named_dmx_modules()	"Returns a list of named modules that are dmx configurable
named_modules([memo, prefix, remove_duplicate])	Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.
named_parameters([prefix, recurse, ...])	Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.
num_parameters([only_trainable, ...])	Get number of (optionally, trainable or non-embeddings) parameters in the module.
parameters([recurse])	Return an iterator over module parameters.
post_init()	A method executed at the end of each Transformer model initialization, to execute code that needs the model's modules properly initialized (such as weight initialization).
post_process_gm(_model, args, kwargs)
prepare_inputs_for_generation(input_ids[, ...])	Prepare the model inputs for generation.
print_model_tree([include_type])	A function that prints out the tree structure of a model
process_inputs_for_export(model, args, kwargs)
prune_heads(heads_to_prune)	Prunes heads of the base model.
push_to_hub(repo_id[, use_temp_dir, ...])	Upload the model file to the 🤗 Model Hub.
register_backward_hook(hook)	Register a backward hook on the module.
register_buffer(name, tensor[, persistent])	Add a buffer to the module.
register_for_auto_class([auto_class])	Register this class with a given auto class.
register_forward_hook(hook, *[, prepend, ...])	Register a forward hook on the module.
register_forward_pre_hook(hook, *[, ...])	Register a forward pre-hook on the module.
register_full_backward_hook(hook[, prepend])	Register a backward hook on the module.
register_full_backward_pre_hook(hook[, prepend])	Register a backward pre-hook on the module.
register_load_state_dict_post_hook(hook)	Register a post-hook to be run after module's load_state_dict() is called.
register_load_state_dict_pre_hook(hook)	Register a pre-hook to be run before module's load_state_dict() is called.
register_module(name, module)	Alias for add_module().
register_parameter(name, param)	Add a parameter to the module.
register_state_dict_post_hook(hook)	Register a post-hook for the state_dict() method.
register_state_dict_pre_hook(hook)	Register a pre-hook for the state_dict() method.
requires_grad_([requires_grad])	Change if autograd should record operations on parameters in this module.
reset_memory_hooks_state()	Reset the mem_rss_diff attribute of each module (see [~modeling_utils.ModuleUtilsMixin.add_memory_hooks]).
resize_position_embeddings(...)
resize_token_embeddings([new_num_tokens, ...])	Resizes input token embeddings matrix of the model if new_num_tokens != config.vocab_size.
retrieve_modules_from_names(names[, ...])
reverse_bettertransformer()	Reverts the transformation from [~PreTrainedModel.to_bettertransformer] so that the original modeling is used, for example in order to save the model.
save_pretrained(save_directory[, ...])	Save a model and its configuration file to a directory, so that it can be re-loaded using the [~PreTrainedModel.from_pretrained] class method.
set_adapter(adapter_name)	If you are not familiar with adapters and PEFT methods, we invite you to read more about them on the PEFT official documentation: https://huggingface.co/docs/peft
set_extra_state(state)	Set extra state contained in the loaded state_dict.
set_input_embeddings(value)	Set model's input embeddings.
set_submodule(target, module[, strict])	Set the submodule given by target if it exists, otherwise throw an error.
share_memory()	See torch.Tensor.share_memory_().
state_dict(*args[, destination, prefix, ...])	Return a dictionary containing references to the whole state of the module.
tensor_parallel(device_mesh)	Tensor parallelize the model across the given device mesh.
thaw([config_file])	A function that transforms the model in place from a config file.
tie_weights()	Tie the weights between the input embeddings and the output embeddings.
to(*args, **kwargs)	Move and/or cast the parameters and buffers.
to_baseline_mode()
to_basic_mode([sbfp_weight_storage])	Configures a transformed DmxModel to the BASIC mode on dmx hardware.
to_bettertransformer()	Converts the model to use [PyTorch's native attention implementation](https://pytorch.org/docs/stable/generated/torch.nn.MultiheadAttention.html), integrated to Transformers through [Optimum library](https://huggingface.co/docs/optimum/bettertransformer/overview).
to_empty(*, device[, recurse])	Move the parameters and buffers to the specified device without copying storage.
to_fp8_mode()	Configures a transformed DmxModel to the FP8 mode on dmx hardware.
to_old_forward(_m)
to_signature_key(_m, _args, _kwargs)
to_transformed_forward(_m)
train([mode])	Set the module in training mode.
transform(config, *rules)	Configure Dmx-specific numerics/sparsity/logics
type(dst_type)	Casts all parameters and buffers to dst_type.
warn_if_padding_and_no_attention_mask(...)	Shows a one-time warning if the input_ids appear to contain padding and no attention mask was given.
xpu([device])	Move all model parameters and buffers to the XPU.
zero_grad([set_to_none])	Reset gradients of all model parameters.

Attributes

T_destination
base_model	The main body of the model.
base_model_prefix
call_super_init
config_class
device	The device on which the module is (assuming that all the module parameters are on the same device).
dmx_config	"Returns the DmxConfig object for the model
dmx_module_names	"Returns a list of module names listed in a dmx_config
dtype	The dtype of the module (assuming that all the module parameters have the same dtype).
dummy_inputs	Dummy inputs to do a forward pass in the network.
dump_patches
framework	Identifies that this is a PyTorch model.
is_gradient_checkpointing	Whether gradient checkpointing is activated for this model or not.
is_parallelizable
loss_function
main_input_name
model_tags
op_set	Returns a set of unique ops present in the model
supports_gradient_checkpointing
supports_pp_plan
supports_tp_plan	Returns whether the model has a tensor parallelism plan.
training
transformed
additional_dmx_aware_mappings