π§ Large-scale transformer-based language modelling using π€ HuggingFace
This repository contains script for training large language models. They are designed to be used with the various π€ HuggingFace libraries. It is based on the official language modeling example scripts available in the Transformers repository. These have been made more configurable and also supports integration with the PEFT library, allowing parameter efficient fine tuning with simply a few cli arguments.
First, the transformers
library needs to be installed from source.
git clone https://github.com/huggingface/transformers
cd transformers
pip install .
Alternatively, you can switch your cloned π€ Transformers to a specific version (for instance with v3.5.1) with
git checkout tags/v3.5.1
Then cd to the root of the repository and install the Python dependencies defined in the requirements.txt
pip install -r requirements.txt
Depending on the system you are using, you might need to install PyTorch from source. See here for instructions.
Setup accelerate:
accelerate config
The required hardware is dependent on a lot of factors, such as the model size, the dataset size, the batch size, the context length, training time, etc. Accelerate can be used to train models on a single machine with multiple GPUs or multiple machines with multiple GPUs.
To estimate the memory requirements, the model-memory-usage π€ space can be used. For DeepSpeed, the deepspeed-model-memory-usage π€ space can be used for calculate how much memory is required for the various Zero Redundancy Optimizer (ZeRO), given a model hosted on the π€ Hugging Face Hub and a hardware setup. can be used.
usage: run_train.py [-h] [--adapter_name ADAPTER_NAME] [--use_lora [USE_LORA]] [--rank RANK]
[--target_modules TARGET_MODULES] [--lora_alpha LORA_ALPHA] [--lora_dropout LORA_DROPOUT]
[--fan_in_fan_out [FAN_IN_FAN_OUT]] [--bias BIAS]
[--modules_to_save MODULES_TO_SAVE [MODULES_TO_SAVE ...]] [--init_lora_weights [INIT_LORA_WEIGHTS]]
[--no_init_lora_weights] [--layers_to_transform LAYERS_TO_TRANSFORM [LAYERS_TO_TRANSFORM ...]]
[--layers_pattern LAYERS_PATTERN [LAYERS_PATTERN ...]] [--rank_pattern RANK_PATTERN]
[--alpha_pattern ALPHA_PATTERN] [--use_prompt_tuning [USE_PROMPT_TUNING]]
[--num_virtual_tokens NUM_VIRTUAL_TOKENS] [--virtual_tokens_init_text VIRTUAL_TOKENS_INIT_TEXT]
[--model_name_or_path MODEL_NAME_OR_PATH] [--model_type MODEL_TYPE]
[--config_overrides CONFIG_OVERRIDES] [--config_name CONFIG_NAME] [--tokenizer_name TOKENIZER_NAME]
[--cache_dir CACHE_DIR] [--use_fast_tokenizer [USE_FAST_TOKENIZER]] [--no_use_fast_tokenizer]
[--model_revision MODEL_REVISION] [--token TOKEN] [--use_auth_token [USE_AUTH_TOKEN]]
[--trust_remote_code [TRUST_REMOTE_CODE]] [--torch_dtype {auto,bfloat16,float16,float32}]
[--low_cpu_mem_usage [LOW_CPU_MEM_USAGE]] [--dataset_name DATASET_NAME]
[--dataset_config_name DATASET_CONFIG_NAME] [--text_column_names TEXT_COLUMN_NAMES] [--target_colum_name TARGET_COLUM_NAME]
[--train_file TRAIN_FILE] [--validation_file VALIDATION_FILE] [--max_train_samples MAX_TRAIN_SAMPLES]
[--max_eval_samples MAX_EVAL_SAMPLES] [--streaming [STREAMING]] [--block_size BLOCK_SIZE]
[--overwrite_cache [OVERWRITE_CACHE]] [--validation_split_percentage VALIDATION_SPLIT_PERCENTAGE]
[--preprocessing_num_workers PREPROCESSING_NUM_WORKERS] [--keep_linebreaks [KEEP_LINEBREAKS]]
[--no_keep_linebreaks] [--ignore_pad_token_for_loss [IGNORE_PAD_TOKEN_FOR_LOSS]]
[--no_ignore_pad_token_for_loss] [--target_start_key TARGET_START_KEY]
[--log_preditions [LOG_PREDITIONS]] [--log_predition_samples LOG_PREDITION_SAMPLES] --output_dir
OUTPUT_DIR [--overwrite_output_dir [OVERWRITE_OUTPUT_DIR]] [--do_train [DO_TRAIN]]
[--do_eval [DO_EVAL]] [--do_predict [DO_PREDICT]] [--evaluation_strategy {no,steps,epoch}]
[--prediction_loss_only [PREDICTION_LOSS_ONLY]]
[--per_device_train_batch_size PER_DEVICE_TRAIN_BATCH_SIZE]
[--per_device_eval_batch_size PER_DEVICE_EVAL_BATCH_SIZE]
[--per_gpu_train_batch_size PER_GPU_TRAIN_BATCH_SIZE]
[--per_gpu_eval_batch_size PER_GPU_EVAL_BATCH_SIZE]
[--gradient_accumulation_steps GRADIENT_ACCUMULATION_STEPS]
[--eval_accumulation_steps EVAL_ACCUMULATION_STEPS] [--eval_delay EVAL_DELAY]
[--learning_rate LEARNING_RATE] [--weight_decay WEIGHT_DECAY] [--adam_beta1 ADAM_BETA1]
[--adam_beta2 ADAM_BETA2] [--adam_epsilon ADAM_EPSILON] [--max_grad_norm MAX_GRAD_NORM]
[--num_train_epochs NUM_TRAIN_EPOCHS] [--max_steps MAX_STEPS]
[--lr_scheduler_type {linear,cosine,cosine_with_restarts,polynomial,constant,constant_with_warmup,inverse_sqrt,reduce_lr_on_plateau}]
[--lr_scheduler_kwargs LR_SCHEDULER_KWARGS] [--warmup_ratio WARMUP_RATIO]
[--warmup_steps WARMUP_STEPS] [--log_level {detail,debug,info,warning,error,critical,passive}]
[--log_level_replica {detail,debug,info,warning,error,critical,passive}]
[--log_on_each_node [LOG_ON_EACH_NODE]] [--no_log_on_each_node] [--logging_dir LOGGING_DIR]
[--logging_strategy {no,steps,epoch}] [--logging_first_step [LOGGING_FIRST_STEP]]
[--logging_steps LOGGING_STEPS] [--logging_nan_inf_filter [LOGGING_NAN_INF_FILTER]]
[--no_logging_nan_inf_filter] [--save_strategy {no,steps,epoch}] [--save_steps SAVE_STEPS]
[--save_total_limit SAVE_TOTAL_LIMIT] [--save_safetensors [SAVE_SAFETENSORS]] [--no_save_safetensors]
[--save_on_each_node [SAVE_ON_EACH_NODE]] [--save_only_model [SAVE_ONLY_MODEL]] [--no_cuda [NO_CUDA]]
[--use_cpu [USE_CPU]] [--use_mps_device [USE_MPS_DEVICE]] [--seed SEED] [--data_seed DATA_SEED]
[--jit_mode_eval [JIT_MODE_EVAL]] [--use_ipex [USE_IPEX]] [--bf16 [BF16]] [--fp16 [FP16]]
[--fp16_opt_level FP16_OPT_LEVEL] [--half_precision_backend {auto,apex,cpu_amp}]
[--bf16_full_eval [BF16_FULL_EVAL]] [--fp16_full_eval [FP16_FULL_EVAL]] [--tf32 TF32]
[--local_rank LOCAL_RANK] [--ddp_backend {nccl,gloo,mpi,ccl,hccl}] [--tpu_num_cores TPU_NUM_CORES]
[--tpu_metrics_debug [TPU_METRICS_DEBUG]] [--debug DEBUG [DEBUG ...]]
[--dataloader_drop_last [DATALOADER_DROP_LAST]] [--eval_steps EVAL_STEPS]
[--dataloader_num_workers DATALOADER_NUM_WORKERS] [--past_index PAST_INDEX] [--run_name RUN_NAME]
[--disable_tqdm DISABLE_TQDM] [--remove_unused_columns [REMOVE_UNUSED_COLUMNS]]
[--no_remove_unused_columns] [--label_names LABEL_NAMES [LABEL_NAMES ...]]
[--load_best_model_at_end [LOAD_BEST_MODEL_AT_END]] [--metric_for_best_model METRIC_FOR_BEST_MODEL]
[--greater_is_better GREATER_IS_BETTER] [--ignore_data_skip [IGNORE_DATA_SKIP]] [--fsdp FSDP]
[--fsdp_min_num_params FSDP_MIN_NUM_PARAMS] [--fsdp_config FSDP_CONFIG]
[--fsdp_transformer_layer_cls_to_wrap FSDP_TRANSFORMER_LAYER_CLS_TO_WRAP] [--deepspeed DEEPSPEED]
[--label_smoothing_factor LABEL_SMOOTHING_FACTOR]
[--optim {adamw_hf,adamw_torch,adamw_torch_fused,adamw_torch_xla,adamw_torch_npu_fused,adamw_apex_fused,adafactor,adamw_anyprecision,sgd,adagrad,adamw_bnb_8bit,adamw_8bit,lion_8bit,lion_32bit,paged_adamw_32bit,paged_adamw_8bit,paged_lion_32bit,paged_lion_8bit,rmsprop}]
[--optim_args OPTIM_ARGS] [--adafactor [ADAFACTOR]] [--group_by_length [GROUP_BY_LENGTH]]
[--length_column_name LENGTH_COLUMN_NAME] [--report_to REPORT_TO [REPORT_TO ...]]
[--ddp_find_unused_parameters DDP_FIND_UNUSED_PARAMETERS] [--ddp_bucket_cap_mb DDP_BUCKET_CAP_MB]
[--ddp_broadcast_buffers DDP_BROADCAST_BUFFERS] [--dataloader_pin_memory [DATALOADER_PIN_MEMORY]]
[--no_dataloader_pin_memory] [--dataloader_persistent_workers [DATALOADER_PERSISTENT_WORKERS]]
[--skip_memory_metrics [SKIP_MEMORY_METRICS]] [--no_skip_memory_metrics]
[--use_legacy_prediction_loop [USE_LEGACY_PREDICTION_LOOP]] [--push_to_hub [PUSH_TO_HUB]]
[--resume_from_checkpoint RESUME_FROM_CHECKPOINT] [--hub_model_id HUB_MODEL_ID]
[--hub_strategy {end,every_save,checkpoint,all_checkpoints}] [--hub_token HUB_TOKEN]
[--hub_private_repo [HUB_PRIVATE_REPO]] [--hub_always_push [HUB_ALWAYS_PUSH]]
[--gradient_checkpointing [GRADIENT_CHECKPOINTING]]
[--gradient_checkpointing_kwargs GRADIENT_CHECKPOINTING_KWARGS]
[--include_inputs_for_metrics [INCLUDE_INPUTS_FOR_METRICS]] [--fp16_backend {auto,apex,cpu_amp}]
[--push_to_hub_model_id PUSH_TO_HUB_MODEL_ID] [--push_to_hub_organization PUSH_TO_HUB_ORGANIZATION]
[--push_to_hub_token PUSH_TO_HUB_TOKEN] [--mp_parameters MP_PARAMETERS]
[--auto_find_batch_size [AUTO_FIND_BATCH_SIZE]] [--full_determinism [FULL_DETERMINISM]]
[--torchdynamo TORCHDYNAMO] [--ray_scope RAY_SCOPE] [--ddp_timeout DDP_TIMEOUT]
[--torch_compile [TORCH_COMPILE]] [--torch_compile_backend TORCH_COMPILE_BACKEND]
[--torch_compile_mode TORCH_COMPILE_MODE] [--dispatch_batches DISPATCH_BATCHES]
[--split_batches [SPLIT_BATCHES]] [--include_tokens_per_second [INCLUDE_TOKENS_PER_SECOND]]
[--include_num_input_tokens_seen [INCLUDE_NUM_INPUT_TOKENS_SEEN]]
[--neftune_noise_alpha NEFTUNE_NOISE_ALPHA]
optional arguments:
-h, --help show this help message and exit
--adapter_name ADAPTER_NAME
The name to use for the adapter. If not specified, the adapter will be named `default`.
(default: None)
--use_lora [USE_LORA]
Whether to use LoRa (default: False)
--rank RANK Lora attention dimension (default: 8)
--target_modules TARGET_MODULES
List of module names or regex expression of the module names to replace with Lora.For example,
['q', 'v'] or '.*decoder.*(SelfAttention|EncDecAttention).*(q|v)$' (default: None)
--lora_alpha LORA_ALPHA
Lora alpha (default: 8)
--lora_dropout LORA_DROPOUT
Lora dropout (default: 0.0)
--fan_in_fan_out [FAN_IN_FAN_OUT]
Set this to True if the layer to replace stores weight like (fan_in, fan_out) (default: False)
--bias BIAS Bias type for Lora. Can be 'none', 'all' or 'lora_only' (default: none)
--modules_to_save MODULES_TO_SAVE [MODULES_TO_SAVE ...]
List of modules apart from LoRA layers to be set as trainable and saved in the final checkpoint.
For example, in Sequence Classification or Token Classification tasks, the final layer
`classifier/score` are randomly initialized and as such need to be trainable and saved.
(default: None)
--init_lora_weights [INIT_LORA_WEIGHTS]
Whether to initialize the weights of the Lora layers with their default initialization. Don't
change this setting, except if you know exactly what you're doing. (default: True)
--no_init_lora_weights
Whether to initialize the weights of the Lora layers with their default initialization. Don't
change this setting, except if you know exactly what you're doing. (default: False)
--layers_to_transform LAYERS_TO_TRANSFORM [LAYERS_TO_TRANSFORM ...]
The layer indexes to transform, is this argument is specified, PEFT will transform only the
layers indexes that are specified inside this list. If a single integer is passed, PEFT will
transform only the layer at this index. This only works when target_modules is a list of str.
(default: None)
--layers_pattern LAYERS_PATTERN [LAYERS_PATTERN ...]
The layer pattern name, used only if `layers_to_transform` is different to None and if the layer
pattern is not in the common layers pattern.This only works when target_modules is a list of
str. (default: None)
--rank_pattern RANK_PATTERN
The mapping from layer names or regexp expression to ranks which are different from the default
rank specified by `r`. For example, `{model.decoder.layers.0.encoder_attn.k_proj: 8`} (default:
{})
--alpha_pattern ALPHA_PATTERN
The mapping from layer names or regexp expression to alphas which are different from the default
alpha specified by `lora_alpha`. For example, `{model.decoder.layers.0.encoder_attn.k_proj: 32`}
(default: {})
--use_prompt_tuning [USE_PROMPT_TUNING]
Whether to use prompt tuning (default: False)
--num_virtual_tokens NUM_VIRTUAL_TOKENS
Number of virtual tokens to use for prompt tuning. (default: None)
--virtual_tokens_init_text VIRTUAL_TOKENS_INIT_TEXT
Initialize the virtual tokens with the given text. Otherwise, the virtual tokens will be
initialized randomly. (default: None)
--model_name_or_path MODEL_NAME_OR_PATH
The model checkpoint for weights initialization. Don't set if you want to train a model from
scratch. (default: None)
--model_type MODEL_TYPE
If training from scratch, pass a model type from the list: bart, bert, bert-generation,
big_bird, bigbird_pegasus, biogpt, blenderbot, blenderbot-small, bloom, camembert, llama,
codegen, cpmant, ctrl, data2vec-text, electra, ernie, falcon, fuyu, git, gpt2, gpt2,
gpt_bigcode, gpt_neo, gpt_neox, gpt_neox_japanese, gptj, llama, marian, mbart, mega, megatron-
bert, mistral, mixtral, mpt, musicgen, mvp, open-llama, openai-gpt, opt, pegasus, persimmon,
phi, plbart, prophetnet, qdqbert, reformer, rembert, roberta, roberta-prelayernorm, roc_bert,
roformer, rwkv, speech_to_text_2, transfo-xl, trocr, whisper, xglm, xlm, xlm-prophetnet, xlm-
roberta, xlm-roberta-xl, xlnet, xmod (default: None)
--config_overrides CONFIG_OVERRIDES
Override some existing default config settings when a model is trained from scratch. Example:
n_embd=10,resid_pdrop=0.2,scale_attn_weights=false,summary_type=cls_index (default: None)
--config_name CONFIG_NAME
Pretrained config name or path if not the same as model_name (default: None)
--tokenizer_name TOKENIZER_NAME
Pretrained tokenizer name or path if not the same as model_name (default: None)
--cache_dir CACHE_DIR
Where do you want to store the pretrained models downloaded from huggingface.co (default: None)
--use_fast_tokenizer [USE_FAST_TOKENIZER]
Whether to use one of the fast tokenizer (backed by the tokenizers library) or not. (default:
True)
--no_use_fast_tokenizer
Whether to use one of the fast tokenizer (backed by the tokenizers library) or not. (default:
False)
--model_revision MODEL_REVISION
The specific model version to use (can be a branch name, tag name or commit id). (default: main)
--token TOKEN The token to use as HTTP bearer authorization for remote files. If not specified, will use the
token generated when running `huggingface-cli login` (stored in `~/.huggingface`). (default:
None)
--use_auth_token [USE_AUTH_TOKEN]
The `use_auth_token` argument is deprecated and will be removed in v4.34. Please use `token`
instead. (default: None)
--trust_remote_code [TRUST_REMOTE_CODE]
Whether or not to allow for custom models defined on the Hub in their own modeling files. This
optionshould only be set to `True` for repositories you trust and in which you have read the
code, as it will execute code present on the Hub on your local machine. (default: False)
--torch_dtype {auto,bfloat16,float16,float32}
Override the default `torch.dtype` and load the model under this dtype. If `auto` is passed, the
dtype will be automatically derived from the model's weights. (default: None)
--low_cpu_mem_usage [LOW_CPU_MEM_USAGE]
It is an option to create the model as an empty shell, then only materialize its parameters when
the pretrained weights are loaded. set True will benefit LLM loading time and RAM consumption.
(default: False)
--dataset_name DATASET_NAME
The name of the dataset to use (via the datasets library). (default: None)
--dataset_config_name DATASET_CONFIG_NAME
The configuration name of the dataset to use (via the datasets library). (default: None)
--text_column_names TEXT_COLUMN_NAMES
The dataset column(s) name to use. (default: None)
--target_colum_name TARGET_COLUM_NAME
The dataset column name to use for labeling. Will be appended to the input text. (default: None)
--train_file TRAIN_FILE
The input training data file (a text file). (default: None)
--validation_file VALIDATION_FILE
An optional input evaluation data file to evaluate the perplexity on (a text file). (default:
None)
--max_train_samples MAX_TRAIN_SAMPLES
For debugging purposes or quicker training, truncate the number of training examples to this
value if set. (default: None)
--max_eval_samples MAX_EVAL_SAMPLES
For debugging purposes or quicker training, truncate the number of evaluation examples to this
value if set. (default: None)
--streaming [STREAMING]
Enable streaming mode (default: False)
--block_size BLOCK_SIZE
Optional input sequence length after tokenization. The training dataset will be truncated in
block of this size for training. Default to the model max input length for single sentence
inputs (take into account special tokens). (default: None)
--overwrite_cache [OVERWRITE_CACHE]
Overwrite the cached training and evaluation sets (default: False)
--validation_split_percentage VALIDATION_SPLIT_PERCENTAGE
The percentage of the train set used as validation set in case there's no validation split
(default: 5)
--preprocessing_num_workers PREPROCESSING_NUM_WORKERS
The number of processes to use for the preprocessing. (default: None)
--keep_linebreaks [KEEP_LINEBREAKS]
Whether to keep line breaks when using TXT files or not. (default: True)
--no_keep_linebreaks Whether to keep line breaks when using TXT files or not. (default: False)
--ignore_pad_token_for_loss [IGNORE_PAD_TOKEN_FOR_LOSS]
Whether to ignore the tokens corresponding to padded labels in the loss computation or not.
(default: True)
--no_ignore_pad_token_for_loss
Whether to ignore the tokens corresponding to padded labels in the loss computation or not.
(default: False)
--target_start_key TARGET_START_KEY
Key for meta column containing the index of the character to start the target sequence from when
predicting with generation. (default: None)
--log_preditions [LOG_PREDITIONS]
Whether to log predictions during training. (default: False)
--log_predition_samples LOG_PREDITION_SAMPLES
Number of samples to log during training. (default: 10)
--output_dir OUTPUT_DIR
The output directory where the model predictions and checkpoints will be written. (default:
None)
--overwrite_output_dir [OVERWRITE_OUTPUT_DIR]
Overwrite the content of the output directory. Use this to continue training if output_dir
points to a checkpoint directory. (default: False)
--do_train [DO_TRAIN]
Whether to run training. (default: False)
--do_eval [DO_EVAL] Whether to run eval on the dev set. (default: False)
--do_predict [DO_PREDICT]
Whether to run predictions on the test set. (default: False)
--evaluation_strategy {no,steps,epoch}
The evaluation strategy to use. (default: no)
--prediction_loss_only [PREDICTION_LOSS_ONLY]
When performing evaluation and predictions, only returns the loss. (default: False)
--per_device_train_batch_size PER_DEVICE_TRAIN_BATCH_SIZE
Batch size per GPU/TPU/MPS/NPU core/CPU for training. (default: 8)
--per_device_eval_batch_size PER_DEVICE_EVAL_BATCH_SIZE
Batch size per GPU/TPU/MPS/NPU core/CPU for evaluation. (default: 8)
--per_gpu_train_batch_size PER_GPU_TRAIN_BATCH_SIZE
Deprecated, the use of `--per_device_train_batch_size` is preferred. Batch size per GPU/TPU
core/CPU for training. (default: None)
--per_gpu_eval_batch_size PER_GPU_EVAL_BATCH_SIZE
Deprecated, the use of `--per_device_eval_batch_size` is preferred. Batch size per GPU/TPU
core/CPU for evaluation. (default: None)
--gradient_accumulation_steps GRADIENT_ACCUMULATION_STEPS
Number of updates steps to accumulate before performing a backward/update pass. (default: 1)
--eval_accumulation_steps EVAL_ACCUMULATION_STEPS
Number of predictions steps to accumulate before moving the tensors to the CPU. (default: None)
--eval_delay EVAL_DELAY
Number of epochs or steps to wait for before the first evaluation can be performed, depending on
the evaluation_strategy. (default: 0)
--learning_rate LEARNING_RATE
The initial learning rate for AdamW. (default: 5e-05)
--weight_decay WEIGHT_DECAY
Weight decay for AdamW if we apply some. (default: 0.0)
--adam_beta1 ADAM_BETA1
Beta1 for AdamW optimizer (default: 0.9)
--adam_beta2 ADAM_BETA2
Beta2 for AdamW optimizer (default: 0.999)
--adam_epsilon ADAM_EPSILON
Epsilon for AdamW optimizer. (default: 1e-08)
--max_grad_norm MAX_GRAD_NORM
Max gradient norm. (default: 1.0)
--num_train_epochs NUM_TRAIN_EPOCHS
Total number of training epochs to perform. (default: 3.0)
--max_steps MAX_STEPS
If > 0: set total number of training steps to perform. Override num_train_epochs. (default: -1)
--lr_scheduler_type {linear,cosine,cosine_with_restarts,polynomial,constant,constant_with_warmup,inverse_sqrt,reduce_lr_on_plateau}
The scheduler type to use. (default: linear)
--lr_scheduler_kwargs LR_SCHEDULER_KWARGS
Extra parameters for the lr_scheduler such as {'num_cycles': 1} for the cosine with hard
restarts (default: {})
--warmup_ratio WARMUP_RATIO
Linear warmup over warmup_ratio fraction of total steps. (default: 0.0)
--warmup_steps WARMUP_STEPS
Linear warmup over warmup_steps. (default: 0)
--log_level {detail,debug,info,warning,error,critical,passive}
Logger log level to use on the main node. Possible choices are the log levels as strings:
'debug', 'info', 'warning', 'error' and 'critical', plus a 'passive' level which doesn't set
anything and lets the application set the level. Defaults to 'passive'. (default: passive)
--log_level_replica {detail,debug,info,warning,error,critical,passive}
Logger log level to use on replica nodes. Same choices and defaults as ``log_level`` (default:
warning)
--log_on_each_node [LOG_ON_EACH_NODE]
When doing a multinode distributed training, whether to log once per node or just once on the
main node. (default: True)
--no_log_on_each_node
When doing a multinode distributed training, whether to log once per node or just once on the
main node. (default: False)
--logging_dir LOGGING_DIR
Tensorboard log dir. (default: None)
--logging_strategy {no,steps,epoch}
The logging strategy to use. (default: steps)
--logging_first_step [LOGGING_FIRST_STEP]
Log the first global_step (default: False)
--logging_steps LOGGING_STEPS
Log every X updates steps. Should be an integer or a float in range `[0,1)`. If smaller than 1,
will be interpreted as ratio of total training steps. (default: 500)
--logging_nan_inf_filter [LOGGING_NAN_INF_FILTER]
Filter nan and inf losses for logging. (default: True)
--no_logging_nan_inf_filter
Filter nan and inf losses for logging. (default: False)
--save_strategy {no,steps,epoch}
The checkpoint save strategy to use. (default: steps)
--save_steps SAVE_STEPS
Save checkpoint every X updates steps. Should be an integer or a float in range `[0,1)`. If
smaller than 1, will be interpreted as ratio of total training steps. (default: 500)
--save_total_limit SAVE_TOTAL_LIMIT
If a value is passed, will limit the total amount of checkpoints. Deletes the older checkpoints
in `output_dir`. When `load_best_model_at_end` is enabled, the 'best' checkpoint according to
`metric_for_best_model` will always be retained in addition to the most recent ones. For
example, for `save_total_limit=5` and `load_best_model_at_end=True`, the four last checkpoints
will always be retained alongside the best model. When `save_total_limit=1` and
`load_best_model_at_end=True`, it is possible that two checkpoints are saved: the last one and
the best one (if they are different). Default is unlimited checkpoints (default: None)
--save_safetensors [SAVE_SAFETENSORS]
Use safetensors saving and loading for state dicts instead of default torch.load and torch.save.
(default: True)
--no_save_safetensors
Use safetensors saving and loading for state dicts instead of default torch.load and torch.save.
(default: False)
--save_on_each_node [SAVE_ON_EACH_NODE]
When doing multi-node distributed training, whether to save models and checkpoints on each node,
or only on the main one (default: False)
--save_only_model [SAVE_ONLY_MODEL]
When checkpointing, whether to only save the model, or also the optimizer, scheduler & rng
state.Note that when this is true, you won't be able to resume training from checkpoint.This
enables you to save storage by not storing the optimizer, scheduler & rng state.You can only
load the model using from_pretrained with this option set to True. (default: False)
--no_cuda [NO_CUDA] This argument is deprecated. It will be removed in version 5.0 of π€ Transformers. (default:
False)
--use_cpu [USE_CPU] Whether or not to use cpu. If set to False, we will use cuda/tpu/mps/npu device if available.
(default: False)
--use_mps_device [USE_MPS_DEVICE]
This argument is deprecated. `mps` device will be used if available similar to `cuda` device. It
will be removed in version 5.0 of π€ Transformers (default: False)
--seed SEED Random seed that will be set at the beginning of training. (default: 42)
--data_seed DATA_SEED
Random seed to be used with data samplers. (default: None)
--jit_mode_eval [JIT_MODE_EVAL]
Whether or not to use PyTorch jit trace for inference (default: False)
--use_ipex [USE_IPEX]
Use Intel extension for PyTorch when it is available, installation:
'https://github.com/intel/intel-extension-for-pytorch' (default: False)
--bf16 [BF16] Whether to use bf16 (mixed) precision instead of 32-bit. Requires Ampere or higher NVIDIA
architecture or using CPU (use_cpu) or Ascend NPU. This is an experimental API and it may
change. (default: False)
--fp16 [FP16] Whether to use fp16 (mixed) precision instead of 32-bit (default: False)
--fp16_opt_level FP16_OPT_LEVEL
For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']. See details at
https://nvidia.github.io/apex/amp.html (default: O1)
--half_precision_backend {auto,apex,cpu_amp}
The backend to be used for half precision. (default: auto)
--bf16_full_eval [BF16_FULL_EVAL]
Whether to use full bfloat16 evaluation instead of 32-bit. This is an experimental API and it
may change. (default: False)
--fp16_full_eval [FP16_FULL_EVAL]
Whether to use full float16 evaluation instead of 32-bit (default: False)
--tf32 TF32 Whether to enable tf32 mode, available in Ampere and newer GPU architectures. This is an
experimental API and it may change. (default: None)
--local_rank LOCAL_RANK
For distributed training: local_rank (default: -1)
--ddp_backend {nccl,gloo,mpi,ccl,hccl}
The backend to be used for distributed training (default: None)
--tpu_num_cores TPU_NUM_CORES
TPU: Number of TPU cores (automatically passed by launcher script) (default: None)
--tpu_metrics_debug [TPU_METRICS_DEBUG]
Deprecated, the use of `--debug tpu_metrics_debug` is preferred. TPU: Whether to print debug
metrics (default: False)
--debug DEBUG [DEBUG ...]
Whether or not to enable debug mode. Current options: `underflow_overflow` (Detect underflow and
overflow in activations and weights), `tpu_metrics_debug` (print debug metrics on TPU).
(default: None)
--dataloader_drop_last [DATALOADER_DROP_LAST]
Drop the last incomplete batch if it is not divisible by the batch size. (default: False)
--eval_steps EVAL_STEPS
Run an evaluation every X steps. Should be an integer or a float in range `[0,1)`. If smaller
than 1, will be interpreted as ratio of total training steps. (default: None)
--dataloader_num_workers DATALOADER_NUM_WORKERS
Number of subprocesses to use for data loading (PyTorch only). 0 means that the data will be
loaded in the main process. (default: 0)
--past_index PAST_INDEX
If >=0, uses the corresponding part of the output as the past state for next step. (default: -1)
--run_name RUN_NAME An optional descriptor for the run. Notably used for wandb logging. (default: None)
--disable_tqdm DISABLE_TQDM
Whether or not to disable the tqdm progress bars. (default: None)
--remove_unused_columns [REMOVE_UNUSED_COLUMNS]
Remove columns not required by the model when using an nlp.Dataset. (default: True)
--no_remove_unused_columns
Remove columns not required by the model when using an nlp.Dataset. (default: False)
--label_names LABEL_NAMES [LABEL_NAMES ...]
The list of keys in your dictionary of inputs that correspond to the labels. (default: None)
--load_best_model_at_end [LOAD_BEST_MODEL_AT_END]
Whether or not to load the best model found during training at the end of training. When this
option is enabled, the best checkpoint will always be saved. See `save_total_limit` for more.
(default: False)
--metric_for_best_model METRIC_FOR_BEST_MODEL
The metric to use to compare two different models. (default: None)
--greater_is_better GREATER_IS_BETTER
Whether the `metric_for_best_model` should be maximized or not. (default: None)
--ignore_data_skip [IGNORE_DATA_SKIP]
When resuming training, whether or not to skip the first epochs and batches to get to the same
training data. (default: False)
--fsdp FSDP Whether or not to use PyTorch Fully Sharded Data Parallel (FSDP) training (in distributed
training only). The base option should be `full_shard`, `shard_grad_op` or `no_shard` and you
can add CPU-offload to `full_shard` or `shard_grad_op` like this: full_shard offload` or
`shard_grad_op offload`. You can add auto-wrap to `full_shard` or `shard_grad_op` with the same
syntax: full_shard auto_wrap` or `shard_grad_op auto_wrap`. (default: )
--fsdp_min_num_params FSDP_MIN_NUM_PARAMS
This parameter is deprecated. FSDP's minimum number of parameters for Default Auto Wrapping.
(useful only when `fsdp` field is passed). (default: 0)
--fsdp_config FSDP_CONFIG
Config to be used with FSDP (Pytorch Fully Sharded Data Parallel). The value is either a fsdp
json config file (e.g., `fsdp_config.json`) or an already loaded json file as `dict`. (default:
None)
--fsdp_transformer_layer_cls_to_wrap FSDP_TRANSFORMER_LAYER_CLS_TO_WRAP
This parameter is deprecated. Transformer layer class name (case-sensitive) to wrap, e.g,
`BertLayer`, `GPTJBlock`, `T5Block` .... (useful only when `fsdp` flag is passed). (default:
None)
--deepspeed DEEPSPEED
Enable deepspeed and pass the path to deepspeed json config file (e.g. `ds_config.json`) or an
already loaded json file as a dict (default: None)
--label_smoothing_factor LABEL_SMOOTHING_FACTOR
The label smoothing epsilon to apply (zero means no label smoothing). (default: 0.0)
--optim {adamw_hf,adamw_torch,adamw_torch_fused,adamw_torch_xla,adamw_torch_npu_fused,adamw_apex_fused,adafactor,adamw_anyprecision,sgd,adagrad,adamw_bnb_8bit,adamw_8bit,lion_8bit,lion_32bit,paged_adamw_32bit,paged_adamw_8bit,paged_lion_32bit,paged_lion_8bit,rmsprop}
The optimizer to use. (default: adamw_torch)
--optim_args OPTIM_ARGS
Optional arguments to supply to optimizer. (default: None)
--adafactor [ADAFACTOR]
Whether or not to replace AdamW by Adafactor. (default: False)
--group_by_length [GROUP_BY_LENGTH]
Whether or not to group samples of roughly the same length together when batching. (default:
False)
--length_column_name LENGTH_COLUMN_NAME
Column name with precomputed lengths to use when grouping by length. (default: length)
--report_to REPORT_TO [REPORT_TO ...]
The list of integrations to report the results and logs to. (default: None)
--ddp_find_unused_parameters DDP_FIND_UNUSED_PARAMETERS
When using distributed training, the value of the flag `find_unused_parameters` passed to
`DistributedDataParallel`. (default: None)
--ddp_bucket_cap_mb DDP_BUCKET_CAP_MB
When using distributed training, the value of the flag `bucket_cap_mb` passed to
`DistributedDataParallel`. (default: None)
--ddp_broadcast_buffers DDP_BROADCAST_BUFFERS
When using distributed training, the value of the flag `broadcast_buffers` passed to
`DistributedDataParallel`. (default: None)
--dataloader_pin_memory [DATALOADER_PIN_MEMORY]
Whether or not to pin memory for DataLoader. (default: True)
--no_dataloader_pin_memory
Whether or not to pin memory for DataLoader. (default: False)
--dataloader_persistent_workers [DATALOADER_PERSISTENT_WORKERS]
If True, the data loader will not shut down the worker processes after a dataset has been
consumed once. This allows to maintain the workers Dataset instances alive. Can potentially
speed up training, but will increase RAM usage. (default: False)
--skip_memory_metrics [SKIP_MEMORY_METRICS]
Whether or not to skip adding of memory profiler reports to metrics. (default: True)
--no_skip_memory_metrics
Whether or not to skip adding of memory profiler reports to metrics. (default: False)
--use_legacy_prediction_loop [USE_LEGACY_PREDICTION_LOOP]
Whether or not to use the legacy prediction_loop in the Trainer. (default: False)
--push_to_hub [PUSH_TO_HUB]
Whether or not to upload the trained model to the model hub after training. (default: False)
--resume_from_checkpoint RESUME_FROM_CHECKPOINT
The path to a folder with a valid checkpoint for your model. (default: None)
--hub_model_id HUB_MODEL_ID
The name of the repository to keep in sync with the local `output_dir`. (default: None)
--hub_strategy {end,every_save,checkpoint,all_checkpoints}
The hub strategy to use when `--push_to_hub` is activated. (default: every_save)
--hub_token HUB_TOKEN
The token to use to push to the Model Hub. (default: None)
--hub_private_repo [HUB_PRIVATE_REPO]
Whether the model repository is private or not. (default: False)
--hub_always_push [HUB_ALWAYS_PUSH]
Unless `True`, the Trainer will skip pushes if the previous one wasn't finished yet. (default:
False)
--gradient_checkpointing [GRADIENT_CHECKPOINTING]
If True, use gradient checkpointing to save memory at the expense of slower backward pass.
(default: False)
--gradient_checkpointing_kwargs GRADIENT_CHECKPOINTING_KWARGS
Gradient checkpointing key word arguments such as `use_reentrant`. Will be passed to
`torch.utils.checkpoint.checkpoint` through `model.gradient_checkpointing_enable`. (default:
None)
--include_inputs_for_metrics [INCLUDE_INPUTS_FOR_METRICS]
Whether or not the inputs will be passed to the `compute_metrics` function. (default: False)
--fp16_backend {auto,apex,cpu_amp}
Deprecated. Use half_precision_backend instead (default: auto)
--push_to_hub_model_id PUSH_TO_HUB_MODEL_ID
The name of the repository to which push the `Trainer`. (default: None)
--push_to_hub_organization PUSH_TO_HUB_ORGANIZATION
The name of the organization in with to which push the `Trainer`. (default: None)
--push_to_hub_token PUSH_TO_HUB_TOKEN
The token to use to push to the Model Hub. (default: None)
--mp_parameters MP_PARAMETERS
Used by the SageMaker launcher to send mp-specific args. Ignored in Trainer (default: )
--auto_find_batch_size [AUTO_FIND_BATCH_SIZE]
Whether to automatically decrease the batch size in half and rerun the training loop again each
time a CUDA Out-of-Memory was reached (default: False)
--full_determinism [FULL_DETERMINISM]
Whether to call enable_full_determinism instead of set_seed for reproducibility in distributed
training. Important: this will negatively impact the performance, so only use it for debugging.
(default: False)
--torchdynamo TORCHDYNAMO
This argument is deprecated, use `--torch_compile_backend` instead. (default: None)
--ray_scope RAY_SCOPE
The scope to use when doing hyperparameter search with Ray. By default, `"last"` will be used.
Ray will then use the last checkpoint of all trials, compare those, and select the best one.
However, other options are also available. See the Ray documentation (https://docs.ray.io/en/lat
est/tune/api_docs/analysis.html#ray.tune.ExperimentAnalysis.get_best_trial) for more options.
(default: last)
--ddp_timeout DDP_TIMEOUT
Overrides the default timeout for distributed training (value should be given in seconds).
(default: 1800)
--torch_compile [TORCH_COMPILE]
If set to `True`, the model will be wrapped in `torch.compile`. (default: False)
--torch_compile_backend TORCH_COMPILE_BACKEND
Which backend to use with `torch.compile`, passing one will trigger a model compilation.
(default: None)
--torch_compile_mode TORCH_COMPILE_MODE
Which mode to use with `torch.compile`, passing one will trigger a model compilation. (default:
None)
--dispatch_batches DISPATCH_BATCHES
Whether to dispatch batches across devices in distributed training. If set to `True`, the
dataloader prepared by the Accelerator is only iterated through on the main process and then the
batches are split and broadcast to each process. Will default to `True` for `DataLoader`
whoseunderlying dataset is an `IterableDataset`, `False` otherwise. (default: None)
--split_batches [SPLIT_BATCHES]
Whether or not the accelerator should split the batches yielded by the dataloaders across the
devices during distributed training. Ifset to `True`, the actual batch size used will be the
same on any kind of distributed processes, but it must be around multiple of the number of
processes you are using (such as GPUs). (default: False)
--include_tokens_per_second [INCLUDE_TOKENS_PER_SECOND]
If set to `True`, the speed metrics will include `tgs` (tokens per second per device). (default:
False)
--include_num_input_tokens_seen [INCLUDE_NUM_INPUT_TOKENS_SEEN]
If set to `True`, will track the number of input tokens seen throughout training. (May be slower
in distributed training) (default: False)
--neftune_noise_alpha NEFTUNE_NOISE_ALPHA
Activates neftune noise embeddings into the model. NEFTune has been proven to drastically
improve model performances for instrcution fine-tuning. Check out the original paper here:
https://arxiv.org/abs/2310.05914 and the original code here:
https://github.com/neelsjain/NEFTune. Only supported for `PreTrainedModel` and `PeftModel`
classes. (default: None)
The script supports sequence to sequence objective within the causal language modeling paradigm. To use this, simply provide both a --text_column_names
and a --target_colum_name
argument. The --text_column_names
argument should be a list of the names of the columns that contain the input text. The --target_colum_name
argument should be the name of the column that contains the target text.
Currently supported PEFT methods are LoRA and Prompt Tuning. To use these methods, simply add the --use_lora
or --use_prompt_tuning
flags to your training arcuments. The --adapter_name
argument can be used to specify the name of the saved adapter. Otherwise, the adapter will be named "default".
Models are often stored in single precision (32-bit). However, some models are stored in half-precision (16-bit) to reduce memory requirements. The --torch_dtype
argument can be used to override the default torch.dtype
and load the model under this dtype. If auto
is passed, the dtype will be automatically derived from the model config.
If a GPU with mixed precision capabilities (architecture Pascal or more recent) is available, one can use mixed precision training with PyTorch 1.6.0 or later, or by installing the Apex library for previous versions. Just add the flag --fp16 to your command! If you have a NVIDIA βAmpereβ GPU architecture, you can use Brain Floting Point (BF16) by passing the flag --bf16.
Using mixed precision training usually results in 2x-speedup for training with (more or less) the same results. See Train With Mixed Precision for more details.
To avoid overflows with mixed precision training, a scaling factor is used. This scaling factor should be as high as possible so that the model can be trained with the greatest precision. Depending on the library used, the scaling factor can often be set dynamically. See Choosing A Scaling Factor for more details.
DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective. To use DeepSpeed, you need to install the deepspeed
package. You can install it with the following command:
pip install deepspeed
There are several configs in the deepspeed_configs
directory that can be used to train the model. The ds_config_zero2.json
config is used to train the model using mixed precision and zero2 optimization. The ds_config_zero3.json
config is used to train the model using mixed precision and zero3 optimization. The ds_config_zero3_bf16.json
config is used to train the model using mixed precision, zero3 optimization, and bf16 precision. The ds_config_zero3_bf16.json
config is used to train the model using mixed precision, zero3 optimization, and bf16 precision.
You can easily log and monitor your runs code.
pip install tensorboard
To use Weights & Biases, install the wandb
package with:
pip install wandb
Then log in the command line:
wandb login
To enable logging to W&B, include "wandb"
in the report_to
of your TrainingArguments
or script. Or just pass along --report_to all
if you have wandb
installed.
Advanced configuration is possible by setting environment variables:
Environment Variable | Value |
---|---|
WANDB_LOG_MODEL | Log the model as artifact (log the model as artifact at the end of training) (false by default) |
WANDB_WATCH | one of gradients (default) to log histograms of gradients, all to log histograms of both gradients and parameters, or false for no histogram logging |
WANDB_PROJECT | Organize runs by project |
Set run names with run_name
argument present in scripts or as part of TrainingArguments
.
Additional configuration options are available through generic wandb environment variables.
Refer to related documentation & examples.
To test the model, the following command can be used:
accelerate launch \
--use_deepspeed \
--deepspeed_multinode_launcher standard \
--deepspeed_config_file ./deepspeed_configs/ds_config_z3.json \
--zero3_init_flag true \
--zero3_save_16bit_model false \
\
run_train.py \
--model_name_or_path facebook/opt-125m \
--dataset_name andstor/methods2test_small \
--dataset_config_name fm+t \
--text_column_names "source" \
--target_colum_name "target" \
--preprocessing_num_workers 10 \
\
--report_to none \
--logging_first_step --logging_steps 1 \
--output_dir .data/tmp/test-clm \
--overwrite_output_dir \
--logging_dir ./logs \
\
--block_size 2048 \
\
--do_train \
--save_strategy steps \
--save_steps 10 --max_train_samples 10 \
--warmup_ratio=0.1 \
--per_device_train_batch_size 1 \
\
--use_lora \
--adapter_name default \
--rank 32 \
--lora_alpha 16 \
--target_modules "q_proj, v_proj" \
--lora_dropout 0.1 \
--bias none \
--learning_rate 5e-3