3outeille's picture
3outeille HF staff
Upload llama-1B/16_GPUS/dp-1_tp-8_pp-2_mbz-16
6a35671 verified
raw
history blame
2.16 kB
general:
project: bench_cluster
seed: 42
model:
ddp_bucket_cap_mb: 25
dtype: bfloat16
init_method:
std: 0.025
make_vocab_size_divisible_by: 1
model_config:
bos_token_id: 1
eos_token_id: 2
hidden_act: silu
hidden_size: 2048
initializer_range: 0.02
intermediate_size: 4096
is_llama_config: true
max_position_embeddings: 4096
num_attention_heads: 32
num_hidden_layers: 24
num_key_value_heads: 32
pad_token_id: null
pretraining_tp: 1
rms_norm_eps: 1.0e-05
rope_scaling: null
rope_theta: 10000.0
tie_word_embeddings: true
use_cache: true
vocab_size: 50257
optimizer:
accumulate_grad_in_fp32: true
clip_grad: 1.0
learning_rate_scheduler:
learning_rate: 0.0001
lr_decay_style: linear
lr_warmup_style: linear
lr_warmup_steps: 1
min_decay_lr: 1.0e-05
optimizer_factory:
adam_beta1: 0.9
adam_beta2: 0.95
adam_eps: 1.0e-08
name: adamW
torch_adam_is_fused: true
weight_decay: 0.01
zero_stage: 1
parallelism:
dp: 1
expert_parallel_size: 1
pp: 2
pp_engine: 1f1b
tp: 8
tp_linear_async_communication: false
tp_mode: REDUCE_SCATTER
profiler:
profiler_export_path: /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-8_pp-2_mbz-16
tokenizer:
tokenizer_max_length: null
tokenizer_name_or_path: openai-community/gpt2
tokenizer_revision: null
data_stages:
- name: Training Stage
start_training_step: 1
data:
dataset:
dataset_overwrite_cache: false
dataset_processing_num_proc_per_process: 64
hf_dataset_config_name: null
hf_dataset_or_datasets: roneneldan/TinyStories
hf_dataset_splits: train
text_column_name: text
num_loading_workers: 32
seed: 42
lighteval: null
tokens:
train_steps: 20
val_check_interval: -1
batch_accumulation_per_replica: 64
limit_test_batches: 0
limit_val_batches: 0
micro_batch_size: 16
sequence_length: 4096
logging:
iteration_step_info_interval: 1
log_level: info
log_level_replica: info
checkpoints:
checkpoint_interval: 100000
checkpoints_path: /dev/null
resume_checkpoint_path: null