-
Notifications
You must be signed in to change notification settings - Fork 140
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
add mistral-nemo-123b model train config
- Loading branch information
Showing
1 changed file
with
231 additions
and
0 deletions.
There are no files selected for viewing
231 changes: 231 additions & 0 deletions
231
launcher_scripts/conf/training/mistral/mistral_nemo_123b.yaml
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,231 @@ | ||
run: | ||
name: mistral_7b | ||
results_dir: ${base_results_dir}/${.name} | ||
time_limit: "0-04:00:00" | ||
dependency: "singleton" | ||
trainer: | ||
num_nodes: 16 | ||
devices: 8 | ||
accelerator: gpu | ||
precision: bf16 | ||
logger: False # logger provided by exp_manager | ||
enable_checkpointing: False | ||
use_distributed_sampler: False | ||
max_epochs: null | ||
max_steps: 300000 # consumed_samples = global_step * global_batch_size | ||
max_time: "05:23:30:00" # days:hours:minutes:seconds | ||
log_every_n_steps: 10 | ||
val_check_interval: 2000 | ||
limit_val_batches: 32 | ||
limit_test_batches: 50 | ||
accumulate_grad_batches: 1 | ||
gradient_clip_val: 1.0 | ||
exp_manager: | ||
explicit_log_dir: ${training.run.results_dir}/results | ||
exp_dir: null | ||
name: megatron_mistral | ||
create_wandb_logger: False | ||
wandb_logger_kwargs: | ||
project: nemo_mistral_pretrain | ||
name: ${training.run.name} | ||
resume_if_exists: false | ||
resume_ignore_no_checkpoint: true | ||
create_checkpoint_callback: True | ||
checkpoint_callback_params: | ||
monitor: val_loss | ||
save_top_k: 10 | ||
mode: min | ||
always_save_nemo: False # saves nemo file during validation, not implemented for model parallel | ||
save_nemo_on_train_end: False # not recommended when training large models on clusters with short time limits | ||
filename: 'megatron_mistral--{val_loss:.2f}-{step}-{consumed_samples}' | ||
model_parallel_size: ${multiply:${training.model.tensor_model_parallel_size}, ${training.model.pipeline_model_parallel_size}} | ||
log_step_timing: True | ||
step_timing_kwargs: | ||
sync_cuda: True | ||
buffer_size: 5 | ||
|
||
model: | ||
mcore_gpt: true | ||
micro_batch_size: 1 | ||
global_batch_size: 32 | ||
rampup_batch_size: null | ||
tensor_model_parallel_size: 8 | ||
pipeline_model_parallel_size: 1 | ||
virtual_pipeline_model_parallel_size: null | ||
encoder_seq_length: 4096 | ||
max_position_embeddings: 32768 | ||
num_layers: 88 | ||
hidden_size: 12288 | ||
ffn_hidden_size: 28672 | ||
num_attention_heads: 96 | ||
init_method_std: 0.02 | ||
use_scaled_init_method: true | ||
hidden_dropout: 0.0 | ||
attention_dropout: 0.0 | ||
ffn_dropout: 0 | ||
kv_channels: 128 | ||
apply_query_key_layer_scaling: false | ||
normalization: rmsnorm | ||
layernorm_epsilon: 1.0e-05 | ||
do_layer_norm_weight_decay: false | ||
make_vocab_size_divisible_by: 128 | ||
pre_process: true | ||
post_process: true | ||
persist_layer_norm: true | ||
bias: false | ||
activation: fast-swiglu | ||
headscale: false | ||
transformer_block_type: pre_ln | ||
openai_gelu: false | ||
normalize_attention_scores: true | ||
position_embedding_type: rope | ||
rotary_percentage: 1.0 | ||
attention_type: multihead | ||
share_embeddings_and_output_weights: false | ||
overlap_p2p_comm: false | ||
batch_p2p_comm: true | ||
seq_len_interpolation_factor: null | ||
num_query_groups: 8 | ||
tokenizer: | ||
library: huggingface | ||
type: mistralai/Mistral-Large-Instruct-2407 | ||
use_fast: true | ||
native_amp_init_scale: 4294967296 | ||
native_amp_growth_interval: 1000 | ||
hysteresis: 2 | ||
fp32_residual_connection: false | ||
fp16_lm_cross_entropy: false | ||
megatron_amp_O2: True | ||
grad_allreduce_chunk_size_mb: 125 | ||
grad_div_ar_fusion: true | ||
gradient_accumulation_fusion: false | ||
bias_activation_fusion: false | ||
bias_dropout_add_fusion: false | ||
masked_softmax_fusion: true | ||
get_attention_mask_from_fusion: true | ||
seed: 1234 | ||
resume_from_checkpoint: null | ||
use_cpu_initialization: false | ||
onnx_safe: false | ||
apex_transformer_log_level: 30 | ||
gradient_as_bucket_view: true | ||
sync_batch_comm: false | ||
activations_checkpoint_granularity: null | ||
activations_checkpoint_method: null | ||
activations_checkpoint_num_layers: null | ||
num_micro_batches_with_partial_activation_checkpoints: null | ||
activations_checkpoint_layers_per_pipeline: null | ||
sequence_parallel: true | ||
transformer_engine: true | ||
fp8: false | ||
fp8_e4m3: false | ||
fp8_hybrid: true | ||
fp8_margin: 0 | ||
fp8_interval: 1 | ||
fp8_amax_history_len: 1024 | ||
fp8_amax_compute_algo: max | ||
reduce_amax: true | ||
use_emha: false | ||
ub_tp_comm_overlap: false | ||
ub_tp_comm_overlap_cfg: null | ||
use_flash_attention: false | ||
nsys_profile: | ||
enabled: false | ||
start_step: 10 | ||
end_step: 10 | ||
ranks: | ||
- 0 | ||
gen_shape: false | ||
optim: | ||
name: distributed_fused_adam | ||
lr: 0.0001 | ||
weight_decay: 0.1 | ||
betas: | ||
- 0.9 | ||
- 0.95 | ||
sched: | ||
name: CosineAnnealing | ||
warmup_steps: 107 | ||
constant_steps: 11873 | ||
min_lr: 1.0e-05 | ||
window_size: | ||
- 4096 | ||
- 0 | ||
gc_interval: 0 | ||
precision: bf16 | ||
mcore_customization_config: | ||
new_decoder_architecture: false | ||
parallel_attention: true | ||
data: | ||
data_impl: mmap | ||
splits_string: "99990,8,2" | ||
seq_length: 2048 | ||
skip_warmup: true | ||
num_workers: 2 | ||
dataloader_type: single | ||
reset_position_ids: false | ||
reset_attention_mask: false | ||
eod_mask_loss: false | ||
index_mapping_dir: null | ||
data_prefix: | ||
- .0333 | ||
- ${data_dir}/my-mistral_00_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_01_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_02_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_03_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_04_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_05_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_06_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_07_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_08_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_09_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_10_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_11_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_12_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_13_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_14_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_15_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_16_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_17_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_18_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_19_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_20_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_21_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_22_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_23_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_24_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_25_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_26_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_27_text_document | ||
- .0333 | ||
- ${data_dir}/my-mistral_28_text_document | ||
- .0334 | ||
- ${data_dir}/my-mistral_29_text_document | ||
|