_modules/torchtitan/config/job_config
Run in Google Colab
Colab
Download Notebook
Notebook
View on GitHub
GitHub
Source code for torchtitan.config.job_config
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the BSD-style license found in the
# LICENSE file in the root directory of this source tree.
from dataclasses import asdict, dataclass, field
from typing import Any, Literal
[docs]
@dataclass
class Job:
config_file: str | None = None
"""Job config file"""
dump_folder: str = "./torchtitan/outputs"
"""Folder to dump job outputs"""
description: str = "default job"
"""Description of the job"""
print_config: bool = False
"""Print the configs to terminal"""
custom_config_module: str = ""
"""
This option allows users to extend the existing JobConfig with a customized
JobConfig dataclass. Users need to ensure that the path can be imported.
"""
@dataclass
class Profiling:
enable_profiling: bool = False
"""Whether to enable pytorch profile"""
save_traces_folder: str = "profile_traces"
"""Trace files location"""
profile_freq: int = 10
"""How often to collect profile traces, in iterations"""
profiler_active: int = 1
"""
The steps profiler is active for.
This is used to configure torch.profile.schedule.
"""
profiler_warmup: int = 3
"""
The number of warmup steps before the active step in each profiling cycle.
This is used to configure torch.profile.schedule.
"""
enable_memory_snapshot: bool = False
"""Whether to dump memory snapshot"""
save_memory_snapshot_folder: str = "memory_snapshot"
"""Memory snapshot files location"""
@dataclass
class Metrics:
log_freq: int = 10
"""How often to log metrics to TensorBoard, in iterations"""
enable_tensorboard: bool = False
"""Whether to log metrics to TensorBoard"""
disable_color_printing: bool = False
"""Whether to disable color printing in logs"""
save_tb_folder: str = "tb"
"""Folder to dump TensorBoard states"""
save_for_all_ranks: bool = False
"""
Whether to save TensorBoard/Wandb metrics only for rank 0 or for all ranks.
When this option is False and pipeline_parallel_degree is > 1, the metrics
component uses the 0th rank of the last stage pipeline group, which is the
only stage that computes loss metrics.
"""
enable_wandb: bool = False
"""Whether to log metrics to Weights & Biases"""
[docs]
@dataclass
class Model:
name: str = "llama3"
"""Which model to train"""
flavor: str = "debugmodel"
"""Which model config to train"""
hf_assets_path: str = "./tests/assets/tokenizer"
"""
Path to HF assets folder. This folder contains local copies of Hugging Face assets,
including model weights in .safetensors format, the model.safetensor.index.json file
(fqn to file mapping), the config.json file, generation_config.json, and tokenizer files.
"""
tokenizer_path: str | None = None
"""DEPRECATED: Use hf_assets_path instead."""
"""Tokenizer path"""
converters: list[str] = field(default_factory=list)
"""
Comma separated list of converters to apply to the model.
For instance, the `float8` converter swaps `torch.nn.Linear`
with `Float8Linear`. This feature requires you to install 'torchao'
which can be found here: https://github.com/pytorch/ao
"""
print_after_conversion: bool = False
"""
If true, model definition will be printed to stdout after all model
converters have been applied.
"""
[docs]
@dataclass
class Optimizer:
name: str = "AdamW"
"""Optimizer to use"""
lr: float = 8e-4
"""Learning rate to use"""
beta1: float = 0.9
beta2: float = 0.95
"""Exponential moving average hyperparameters to use"""
eps: float = 1e-8
"""Epsilon value to use"""
weight_decay: float = 0.1
"""Weight decay to use"""
implementation: Literal["for-loop", "foreach", "fused"] = "fused"
"""
Specify which optimizer implementation to use:
- 'fused': Use fused implementation (CUDA only) for best performance.
- 'foreach': Use some horizontal fusion of tensors for better performance.
- 'for-loop': Use the default implementation for the optimizer (slowest).
- more info: https://pytorch.org/docs/stable/optim.html
"""
early_step_in_backward: bool = False
"""
Whether to apply optimizer in the backward. Caution, optimizer_in_backward
is not compatible with gradients clipping, users should not call
register_post_accumulate_grad_hook after the optimizer is built.
"""
@dataclass
class LRScheduler:
warmup_steps: int = 200
"""
Steps for lr scheduler warmup, normally 1/5 of --training.steps
"""
decay_ratio: float | None = None
"""
Controls the proportion of the training steps allocated to the learning rate decay phase.
If `None`, the learning rate will begin decaying immediately after the warmup period.
Otherwise, the learning rate will remain stable after the warmup period and
only start decaying during the last `decay_ratio` portion of the total training steps.
This is known as the Warmup-Stable-Decay (WSD) schedule, as described in https://arxiv.org/abs/2404.06395.
"""
decay_type: Literal["linear", "sqrt", "cosine"] = "linear"
"""
Learning rate decay type to use during training:
- 'linear': linearly decays learning rate from initial to final value
- 'sqrt': decays learning rate following a 1 minus square root curve
- 'cosine': smoothly decays learning rate following a cosine curve
"""
min_lr_factor: float = 0.0
"""
Min lr ratio for lr scheduler.
If provided, the range of decay factor is scaled from 1 to `min_lr_factor`
to ensure the learning rate does not drop below `optimizer.lr * lr_scheduler.min_lr_factor`.
"""
[docs]
@dataclass
class Training:
dataset: str = "c4_test"
"""Dataset to use"""
dataset_path: str | None = None
"""
Path to the dataset in the file system. If provided, data will be
loaded from this path instead of downloaded.
"""
local_batch_size: int = 8
"""Local batch size (i.e., per-device batch size)"""
global_batch_size: int = -1
"""
Global batch size (defaults to `training.local_batch_size * data-parallel degree`)
"""
seq_len: int = 2048
"""Sequence length"""
max_norm: float | int = 1.0
"""Max norm for gradient clipping"""
steps: int = 10000
"""How many train steps to run"""
enable_cpu_offload: bool = False
"""
Whether to apply CPU offloading of parameters, gradients, and optimizer states in FSDP
"""
dtype: Literal["bfloat16", "float32"] = "float32"
"""
torch dtype for training. In contrast to mixed precision training, setting training_dtype=bfloat16 will
put all parameters, gradients, and optimizer states in bfloat16, without an extra copy of fp32 weights.
In the case of full bf16 training, RoPE calculations and logits will still be in fp32.
"""
mixed_precision_param: Literal["bfloat16", "float32"] = "bfloat16"
"""
torch dtype to use for parameters when applying mixed precision via fully_shard or torch.autocast.
This feature takes effect via fully_shard when data_parallel_shard_degree > 1 or
context_parallel_degree > 1; it takes effect via torch.autocast when data_replicate_degree >= 1
and no other parallelism is enabled, i.e. under DDP or single-device training.
"""
mixed_precision_reduce: Literal["float32"] = "float32"
"""
torch dtype to use for reductions when applying mixed precision via FSDP.
This feature only takes effect when data_parallel_shard_degree > 1
"""
gc_freq: int = 50
"""Python garbage control scheduling interval, in steps"""
gc_debug: bool = False
"""
Enable GC debugging mode. This will perform gc.collect() at every step to
detect if there is a reference cycle that includes a CUDA Tensor.
Note that you may want to lower the training steps to avoid generating too
many temporary files.
"""
seed: int | None = None
"""Choose the base RNG seed used for training"""
deterministic: bool = False
"""Use deterministic algorithms wherever possible, may be slower"""
debug_moe_force_load_balance: bool = False
"""If True, we force each experts to get the same amount of tokens via round-robin. This option is for debugging usage only."""
[docs]
@dataclass
class Parallelism:
data_parallel_replicate_degree: int = 1
"""
The `data_parallel_replicate_degree` argument specifies the degree of
data parallelism for weight replication. When this value is greater
than 1, weights will be replicated across `data_parallel_replicate_degree`
ranks. If `data_parallel_shard_degree` is also greater than 1, the parallelism
method used is HSDP (Hybrid Sharded Data Parallelism). Otherwise, the
parallelism method used is DDP (Distributed Data Parallelism).
1 means disabled.
"""
enable_compiled_autograd: bool = False
"""Enable CompiledAutograd to compile the backward."""
data_parallel_shard_degree: int = -1
"""
The `data_parallel_shard_degree` argument specifies the degree of data
parallelism for weight sharding. When this value is greater than 1, weights
will be sharded across `data_parallel_shard_degree` ranks. If
`data_parallel_replicate_degree` is also greater than 1, the parallelism
method used is HSDP (Hybrid Sharded Data Parallelism). Otherwise, the
parallelism method used is FSDP (Fully Sharded Data Parallelism).
-1 means leftover ranks will be used (After DP_REPLICATE/SP/PP). Note that
only `data_parallel_shard_degree` can be negative. 1 means disabled.
"""
fsdp_reshard_after_forward: Literal["default", "always", "never"] = "default"
"""
`reshard_after_forward` specifies the policy for applying `reshard_after_forward`
within an FSDP setup. `reshard_after_forward` controls parameter behavior after forward,
trading off memory and communication. See torch's `fully_shard` API for more documentation
on `reshard_after_forward`.
The supported policies include "default", "always" and "never":
- "default" applies default resharding behavior, implementing "smart defaults" for known optimal
scenarios.
- "always" will enable `reshard_after_forward` for all forward passes.
- "never" will disable `reshard_after_forward` for all forward passes.
"""
tensor_parallel_degree: int = 1
"""Tensor Parallelism degree. 1 means disabled."""
disable_loss_parallel: bool = False
"""Whether to apply loss parallel when sequence parallel is enabled"""
enable_async_tensor_parallel: bool = False
"""Whether to apply async tensor parallel (currently only effective when compile is enabled)"""
pipeline_parallel_degree: int = 1
"""
Pipeline Parallelism degree, or number of ranks. 1 means disabled.
If using looped schedules, this still specifies the number of physical ranks, not the number
of stages. Stages per rank are inferred from split points degree, and schedule.
"""
module_fqns_per_model_part: list[list[str]] | None = None
"""
Specify a list of lists containing the FQNs (Fully Qualified Names) of modules for each model chunk.
Each inner list represents one model chunk and contains the module names that belong to that chunk.
e.g. [['tok_embeddings', 'layers.0'], ['layers.1', 'layers.2'], ['layers.3', 'layers.4']]
will create 3 chunks: the first containing tok_embeddings and layers.0,
the second containing layers.1 and layers.2, and the third containing layers.3 and layers.4.
This provides more explicit control over which modules belong to each chunk compared to split points.
"""
pipeline_parallel_first_stage_less_layers: int = 1
"""
The number of layers to reduce in the first stage of pipeline parallelism. This is because
the first stage has the extra overhead of the embedding layer, which is not present in the other stages.
"""
pipeline_parallel_last_stage_less_layers: int = 1
"""
The number of layers to reduce in the last stage of pipeline parallelism. This is because
the last stage has the extra overhead of the output layer, which is not present in the other stages.
"""
pipeline_parallel_layers_per_stage: int | None = None
"""
The number of layers per (virtual) pipeline stage. If specified, the module_fqns_per_model_part will be
calculated from the number of layers and pipeline_parallel_degree. If not specified, the
layers per stage will be inferred from the model, schedule, and pipeline_parallel_degree.
"""
pipeline_parallel_schedule: str = "1F1B"
"""
Specify the Pipeline Parallel schedule to use. The supported schedules are:
https://github.com/pytorch/pytorch/blob/de4c2a3b4e89d96334dc678d1c3f2ae51a6630a0/torch/distributed/pipelining/schedules.py#L2161.
The schedule must be compatible with the split points and stages_per_rank.
Looped schedules (e.g. Interleaved1F1B) require specifying pipeline_parallel_degree = number of ranks,
and split_points = number of stages - 1
"""
pipeline_parallel_schedule_csv: str | None = ""
"""
Specify the path to the pipeline parallel schedule csv file to use.
The pipeline_parallel_schedule argument must be either
PipelineScheduleSingle, PipelineScheduleMulti, or _PipelineScheduleRuntime.
"""
pipeline_parallel_microbatch_size: int = 1
"""
The size of each pipeline parallel microbatch (default 1).
This value is used to compute the total number of microbatches by dividing local_batch_size with
pipeline_parallel_microbatch_size.
The global training batch size must be evenly divisible by pipeline_parallel_microbatch_size.
"""
context_parallel_degree: int = 1
"""Context parallelism degree. 1 means disabled."""
context_parallel_rotate_method: Literal["allgather", "alltoall"] = "allgather"
"""
The collective to use in context parallel SDPA for kv shards exchange.
- 'allgather' means to all-gather all kv shards on ranks after the first sub-SDPA computation,
- 'alltoall' means to all-to-all shuffle the kv shards.
The default value is 'allgather'.
"""
expert_parallel_degree: int = 1
"""
Expert parallelism degree. 1 means disabled. No effect for non-MoE models.
Currently, it is supported with the following constraints:
- when etp = tp:
- cp <= ep <= dp_shard * cp
- ep % cp == 0
- dp_shard * cp % ep == 0
- when etp = 1:
- cp * tp <= ep <= dp_shard * cp * tp
- ep % (cp * tp) == 0
- dp_shard * cp * tp % ep == 0
Note that this is still an experimental feature. Some constraints will be
relaxed soon when we have more flexible DeviceMesh support.
"""
expert_tensor_parallel_degree: int = 1
"""
Expert tensor parallelism degree. 1 means disabled. No effect for non-MoE models, or when ep = 1.
With this option, the tensor parallel degree on routed experts can be different from that on other params.
Currently, we only support either
- [partial dp -> ep] etp = tp
- [partial dp + all tp -> ep] etp = 1
Note that this is still an experimental feature.
"""
[docs]
@dataclass
class Checkpoint:
enable: bool = False
"""Whether to enable checkpoint"""
enable_ft_dataloader_checkpoints: bool = True
"""
Warning: Disabling this can have fault tolerant replicas training
over the same data multiple times. Use it with caution if training
over the same data is acceptable.
Used to enable checkpointing the dataloader index for fault tolerant training with torchft.
Fault tolerant training stores data loader index in the checkpoints, so that training can resume
without going over the same batch twice.
If enabled, data loader state is checkpointed. Otherwise, replicas
will train over the same data multiple times, which can result in
overfitting.
The failed replcia will still recover other state e.g. model
parameters from other replcias.
Note, if regular checkpointing is enabled, we also checkpoint the
data loader state. But when not using fault tolerance, the entire training starts from scratch.
"""
folder: str = "checkpoint"
"""
The folder to store the checkpoints.
When enable is set to true, checkpoints will be in {--job.dump_folder}/{--checkpoint.folder}.
"""
interval: int = 500
"""Checkpointing interval in steps."""
initial_load_path: str | None = None
"""
This option specifies the path to the initial checkpoint to load, which is
particularly useful for resuming training from a previous run with a
different output path or when loading a checkpoint from a pre-trained model.
If the checkpoint folder for the current run is not empty,
located at {--job.dump_folder}/{--checkpoint.folder}, this option will be ignored.
This feature allows users to load an initial checkpoint from a different folder and
continue training, saving new checkpoints to the specified folder without affecting
the existing ones.
Note that the path should contain the full path to the checkpoint folder,
including the step number, if any; for example,
"//pre_train/checkpoints/llama3/llama3_8b/step_10000".
"""
initial_load_model_only: bool = True
"""
This option specifies if only the model should be loaded during the initial
checkpoint load. The option is only used when `initial_load_path` is specified.
If False, the checkpoint at `initial_load_path` is treated as a standard training
checkpoint, including optimizer, lr scheduler, training states, etc.
The default setting for this option is True. Note that you will have to use
`--checkpoint.no_initial_load_model_only` to override the default setting.
"""
initial_load_in_hf: bool = False
"""
Enable the use of HuggingFace's safetensors format for checkpointing. The option
is only used when `initial_load_path` is specified. This will load checkpoints
in HF's model definition and safetensors format instead of the default torchtitan
model definition and DCP format, after necessary model state dict transformation.
`initial_load_model_only` must be true because safetensors doesn't support saving
non-tensors. The default value is False.
"""
initial_load_in_hf_quantized: bool = False
"""
Enable loading of HuggingFace's safetensors format with quantized state dict keys. The option
is only used when `initial_load_path` and `initial_load_path_in_hf` is specified. This will load
checkpoints in HF's model definition and dequantize on model weights if necessary. To support
this parameter, the model need to define proper HuggingFaceStorageReader to perform dequantize.
"""
last_save_model_only: bool = True
"""
When last_save_model_only=True, only the model will be saved at the end of training,
the last save. With this, checkpoints can be loaded using `torch.load(..., weights_only=True)`
after conversion. When last_save_model_only=False, the full checkpoint will be saved.
A full checkpoint includes model, optimizer and train_state, which can be used to resume training.
The default value is True.
"""
last_save_in_hf: bool = False
"""
Enable the use of Hugging Face's safetensors format for checkpointing. This will save the
final checkpoints in safetensors format instead of the default DCP format, after necessary
model state dict transformation. There will be a performance cost in using this as we need
to consolidate the sharded tensors to full tensors as a separate step.
last_save_model_only must be true because safetensors doesn't support saving
non-tensors. On load, this argument isn't needed as we will detect whether the loaded
checkpoint is in safetensors format or not. The default value is False.
"""
export_dtype: Literal["float16", "bfloat16", "float32"] = "float32"
"""
Converts to the specified precision when training completes and last_save_model_only=true.
"""
async_mode: Literal["disabled", "async", "async_with_pinned_mem"] = "disabled"
"""
Which async checkpoint mode to use. Currently there are 3 different modes.
- "disabled": synchronized checkpointing will be used.
- "async": torch.distributed.checkpoint.async_save will be used.
- "async_with_pinned_mem": this option utilizes a dedicated pinned memory space and creates a
separate process for faster GPU->CPU transfer performance and eliminating GIL contention.
The cost is increased CPU memory usage. If insufficient CPU memory is available, performance
may degrade due to memory paging. For most users, "async" should suffice as the performance
overhead is typically small (on the order of tens of seconds) compared to checkpointing
frequency. This mode can be employed to pursue near-zero checkpointing times
(e.g., < 1 second) given appropriate hardware support such as ample CPU memory and fast PCIe.
"disabled" is the default mode.
"""
keep_latest_k: int = 10
"""
Keeps only the latest k checkpoints, and purging older ones. If 0, keep all checkpoints.
K cannot be 1 as the last one may be in the process of being saved. As a result,
the metadata of the last one may not be ready yet. The default value is 10 to avoid
filling up the disk.
"""
load_step: int = -1
"""Load the checkpoint at the specified step. If -1, load the latest checkpoint."""
exclude_from_loading: list[str] = field(default_factory=list)
"""
Exclude specific keys from being loaded from the checkpoint.
Provide a comma-separated list of keys to exclude, e.g. 'optimizer,lr_scheduler,dataloader'.
This will load the model only, excluding the specified keys.
"""
enable_first_step_checkpoint: bool = False
"""
Enable the checkpoint save at first step. This will save a checkpoint immediately
after the first step to ensure checkpointing functions correctly. This is useful
when running on a new cluster or storage to verify checkpointing without waiting
for many steps or checkpointing too frequently. The default value is False.
"""
create_seed_checkpoint: bool = False
"""
Initializes the full model without applying parallelisms, and then saves it as a seed checkpoint.
Note: requires user to call train.py without specifying any parallelisms, e.g. NGPU=1.
Could be implemented as a separate script, but this way shares more code.
"""
load_only: bool = False
"""
In certain scenarios, you may only need to load checkpoints for verification or debugging
purposes, without saving any new checkpoints. For example, you might use seed checkpoints
to validate model correctness. Enabling this option allows checkpoints to be loaded
without saving any during the training.
"""
@dataclass
class ActivationCheckpoint:
mode: Literal["selective", "full", "memory_budget", "none"] = "selective"
"""Type of activation checkpointing to use"""
selective_ac_option: str = "2"
"""
Selective activation checkpointing options ['int', 'op'].
'int' (e.g., 2) for every nth layer, or 'op' for op level ac.
"""
per_op_sac_force_recompute_mm_shapes_by_fqns: list[str] = field(
default_factory=lambda: ["moe.router.gate"]
)
"""
When per-op selective ac is used, this list of fully qualified names is used
to determine which mm shapes to force recompute, rather than being considered
by rest of the sac policy, e.g save every other mm. Only nn.Linear modules are
supported today.
Note: this config applies to mms not limited to those matching the specified
fqns, e.g. if "moe.router.gate", corresponding to Linear(in, out), is specified,
ANY mm with shape matching (*, in) x (in, out) will be force recomputed.
"""
early_stop: bool = False
"""
Whether to stop recomputing early when all activations have already been
rematerialized.
"""
memory_budget: float = 0.5
"""
When mode is set to "memory_budget", this value determines how much
partitioner in the compiler should trade off compute for memory.
0.0 corresponds to the activation memory from applying
activation checkpointing to the full compiled region, and 1.0 corresponds to
the activation memory from the default runtime-optimized strategy. Read here:
https://pytorch.org/blog/activation-checkpointing-techniques/
"""
visualize_memory_budget_pareto: bool = False
"""
This dumps out a SVG visualization of the expected runtime vs. activation
memory tradeoffs for all memory budget values from 0 to 1 in increments of
0.05 in {--job.dump_folder}/memory_budget_pareto folder. See an example here:
https://github.com/pytorch/pytorch/pull/126320#discussion_r1625104015
"""
@dataclass
class Compile:
enable: bool = False
"""Whether to apply torch.compile"""
components: list[Literal["model", "loss"]] = field(
default_factory=lambda: ["model", "loss"]
)
"""Which components to compile"""
backend: str = "inductor"
@dataclass
class Float8Linear:
enable_fsdp_float8_all_gather: bool = False
"""Whether enable float8 all-gather in FSDP, recommended for tensorwise scaling"""
precompute_float8_dynamic_scale_for_fsdp: bool = False
"""Whether precompute float8 scales dynamically for FSDP, recommended for tensorwise scaling"""
recipe_name: Literal["tensorwise", "rowwise", "rowwise_with_gw_hp"] | None = None
"""If specified, creates float8 config from recipe name"""
filter_fqns: list[str] = field(default_factory=list)
"""
Comma-separated list of fully qualified names of modules to skip applying float8 training to.
nn.Linear modules with any dim size not divisible by 16 are always skipped due to hardware requirements.
Example: --quantize.linear.float8.filter_fqns "attention.wq,attention.wk,attention.wv,output"
"""
emulate: bool = False
"""
If True, emulation is used instead of hardware accelerated gemm. This is for test purpose only,
as the current CI does not have sm_89 capability, required by Float8.
Not compatible with torch.compile.
"""
@dataclass
class Float8GroupedMM:
fqns: list[str] | str = field(default_factory=list)
"""
*Prototype feature, performance optimization still in progress*
Comma-separated list of fully qualified names of MoE Layers to apply FP8 dynamic quantization on grouped GEMM operations.
This is a prototype feature that requires the torchao nightly build.
Example: --quantize.grouped_mm.float8.fqns="experts"
"""
@dataclass
class MXLinear:
mxfp8_dim1_cast_kernel_choice: Literal["triton", "cuda", "torch"] = "triton"
"""
Temp work around for inductor performance gap.
CUDA is recommended for best performance.
Example: --quantize.linear.mx.mxfp8_dim1_cast_kernel_choice="cuda"
"""
recipe_name: str = "mxfp8_cublas"
"""
If specified, creates MX config from recipe name. See
https://github.com/pytorch/ao/tree/main/torchao/prototype/mx_formats for more information.
Example: --quantize.linear.mx.recipe_name="mxfp8_cublas"
"""
filter_fqns: list[str] = field(default_factory=lambda: ["output"])
"""
Comma-separated list of fully qualified names of modules to skip applying mxfp8 training to.
nn.Linear modules with any dim size not divisible by 16 are also always skipped due to hardware requirements.
By default we always skip the output layer.
Example: --quantize.linear.mx.filter_fqns="attention.wq,attention.wk,attention.wv,output"
"""
@dataclass
class MXGroupedMM:
recipe_name: Literal["mxfp8"] = "mxfp8"
"""
Quantization recipe name for grouped GEMMs. Options: ["mxfp8"]
Example: --quantize.grouped_mm.mx.recipe_name="mxfp8"
"""
fqns: list[str] | str = field(default_factory=list)
"""
*Prototype feature, performance optimization still in progress*
Comma-separated list of fully qualified names of MoE modules to apply MXFP8 dynamic quantization on grouped GEMM operations.
This is a prototype feature that requires the torchao nightly build.
Example: --quantize.grouped_mm.mx.fqns="experts"
"""
@dataclass
class QuantizedLinear:
float8: Float8Linear = field(default_factory=Float8Linear)
"""FP8 training config for nn.Linear layers"""
mx: MXLinear = field(default_factory=MXLinear)
"""MX training config for nn.Linear layers"""
@dataclass
class QuantizedGroupedMM:
float8: Float8GroupedMM = field(default_factory=Float8GroupedMM)
"""FP8 training config for grouped GEMMs"""
mx: MXGroupedMM = field(default_factory=MXGroupedMM)
"""MX training config for grouped GEMMs"""
@dataclass
class Quantize:
linear: QuantizedLinear = field(default_factory=QuantizedLinear)
"""Quantized training config for nn.Linear layers"""
grouped_mm: QuantizedGroupedMM = field(default_factory=QuantizedGroupedMM)
"""Quantized training config for grouped GEMMs"""
@dataclass
class Comm:
init_timeout_seconds: int = 300
"""Timeout for communication operations, during initialization and first train step."""
train_timeout_seconds: int = 100
"""
Timeout for communication operations after the first train step --
usually a tighter bound than during initialization.
"""
trace_buf_size: int = 20000
"""Flight recorder ring buffer size, >0 means recording by default, 0 means disabled"""
save_traces_folder: str = "comm_traces"
"""Flight recorder trace files location"""
save_traces_file_prefix: str = "rank_"
"""Flight recorder trace files prefix"""
@dataclass
class MemoryEstimation:
enable: bool = False
"""Whether to estimate memory usage for FSDP"""
disable_fake_mode: bool = False
"""Whether to estimate memory under FakeTensorMode"""
@dataclass
class FaultTolerance:
enable: bool = False
"""
Enable TorchFT integration. When TorchFT is enabled, HSDP will be used.
And --fault_tolerance.data_parallel_replicate_degree should be 1 and
--fault_tolerance.group_size will be used to control the maximum
replicate group size as the replicate group size is dynamic.
Note that this is still an experimental feature.
"""
process_group: str = "gloo"
"""
The process group to use for fault tolerance. Currently, only "gloo" and "nccl" are supported.
"""
process_group_timeout_ms: int = 10000
"""
The process group will abort if operations don't succeed within this duration.
Note: This currently only works with gloo process group.
"""
replica_id: int = 0
"""The TorchFT replica ID of this run."""
group_size: int = 0
"""
The number of TorchFT replicate groups. This number will be used for
dataloader to split the dataset across the replicate groups and FSDP
dimension
"""
min_replica_size: int = 1
"""The minimum number of FT replica for each step."""
semi_sync_method: str | None = None
"""
The algorithm to use for semi-sync training. Currently, only "local_sgd" and "diloco" from
torchft are supported
(https://github.com/pytorch/torchft/blob/360c5c534bdeac959507e9d238ba9f3902d3fda9/torchft/local_sgd.py#L41)
"""
@dataclass
class Experimental:
custom_import: str = ""
"""
This option enables the importation of external modules.
Currently, it only supports dotted import modules (e.g., some_package.model_x).
It is the user's responsibility to ensure that the specified path can be
successfully imported. One method to achieve this, you can place your module
inside the ``torchtitan/torchtitan`` folder and execute ``pip install -e .`` to
make it available for import.
"""
custom_args_module: str = ""
"""
DEPRECATED (moved to Job.custom_config_module). Will be removed soon.
This option allows users to extend TorchTitan's existing JobConfig by extending
a user defined JobConfig dataclass. Similar to ``--experimental.custom_model_path``, the user
needs to ensure that the path can be imported.
"""
@dataclass
class Validation:
enable: bool = False
"""Enable validation to default run validation after each training loop"""
dataset: str = "c4_validation"
"""Dataset to use for validation"""
dataset_path: str | None = None
"""Path to dataset to use for validation"""
local_batch_size: int = 8
"""Batch size for validation"""
seq_len: int = 2048
"""Sequence length for validation"""
freq: int = 10
"""Frequency of validation"""
steps: int = -1
"""
Number of steps to take in the validation set, -1 means consuming all the data in the validation dataset
WARNING: When setting to -1 there could be hangs due to mismatch among ranks
"""
def __post_init__(self):
assert (
self.steps > 0 or self.steps == -1
), "validation steps must be positive or -1"
@dataclass
class JobConfig:
"""
Default container for training configuration.
"""
job: Job = field(default_factory=Job)
profiling: Profiling = field(default_factory=Profiling)
metrics: Metrics = field(default_factory=Metrics)
model: Model = field(default_factory=Model)
optimizer: Optimizer = field(default_factory=Optimizer)
lr_scheduler: LRScheduler = field(default_factory=LRScheduler)
training: Training = field(default_factory=Training)
parallelism: Parallelism = field(default_factory=Parallelism)
checkpoint: Checkpoint = field(default_factory=Checkpoint)
activation_checkpoint: ActivationCheckpoint = field(
default_factory=ActivationCheckpoint
)
compile: Compile = field(default_factory=Compile)
quantize: Quantize = field(default_factory=Quantize)
comm: Comm = field(default_factory=Comm)
memory_estimation: MemoryEstimation = field(default_factory=MemoryEstimation)
fault_tolerance: FaultTolerance = field(default_factory=FaultTolerance)
experimental: Experimental = field(default_factory=Experimental)
validation: Validation = field(default_factory=Validation)
def to_dict(self) -> dict[str, Any]:
return asdict(self)
