14m400m100m/3423734.out

Model parameters: d_model 224 ffw_size 896 kv_size 32 n_heads 7 n_layers 4
Megatron-DeepSpeed/pretrain_gpt.py --tensor-model-parallel-size 1 --pipeline-model-parallel-size 1 --num-layers 4 --hidden-size 224 --num-attention-heads 7 --kv-channels 32 --ffn-hidden-size 896 --seq-length 2048 --max-position-embeddings 2048 --micro-batch-size 32 --global-batch-size 256 --train-samples 195_313 --vocab-file gpt2/vocab.json --merge-file gpt2/merges.txt --loss-scale 12 --clip-grad 1.0 --kill-switch-path kill-switch-14m400m100m --bf16 --checkpoint-activations --optimizer adam --adam-beta1 0.9 --adam-beta2 0.999 --adam-eps 1e-8 --lr 2e-4 --min-lr 2e-5 --lr-decay-style cosine --lr-decay-samples 195_313 --lr-warmup-samples 1953 --clip-grad 1.0 --weight-decay 1e-1 --log-interval 10 --save-interval 1000 --eval-interval 1000 --eval-iters 1 --tensorboard-dir tensorboard_14m400m100m --tensorboard-queue-size 5 --log-timers-to-tensorboard --log-batch-size-to-tensorboard --log-validation-ppl-to-tensorboard --save checkpoints_14m400m100m --load checkpoints_14m400m100m --train-weighted-split-paths-path train100m.txt --valid-weighted-split-paths-path val.txt --data-impl mmap --deepspeed --deepspeed_config ds_configs/3423734.json --zero-stage 0
START 3423734: Thu 27 Apr 2023 03:36:35 PM EEST
0: 
0: 
0: ======================= ROCm System Management Interface =======================
0: ================================= Concise Info =================================
0: GPU  Temp   AvgPwr  SCLK    MCLK     Fan  Perf  PwrCap  VRAM%  GPU%  
0: 0    47.0c  89.0W   800Mhz  1600Mhz  0%   auto  560.0W    0%   0%    
0: 1    48.0c  N/A     800Mhz  1600Mhz  0%   auto  0.0W      0%   0%    
0: 2    37.0c  89.0W   800Mhz  1600Mhz  0%   auto  560.0W    0%   0%    
0: 3    49.0c  N/A     800Mhz  1600Mhz  0%   auto  0.0W      0%   0%    
0: 4    47.0c  86.0W   800Mhz  1600Mhz  0%   auto  560.0W    0%   0%    
0: 5    44.0c  N/A     800Mhz  1600Mhz  0%   auto  0.0W      0%   0%    
0: 6    41.0c  83.0W   800Mhz  1600Mhz  0%   auto  560.0W    0%   0%    
0: 7    46.0c  N/A     800Mhz  1600Mhz  0%   auto  0.0W      0%   0%    
0: ================================================================================
0: ============================= End of ROCm SMI Log ==============================
0: Launching on nid005141 (0/1), master nid005141 port 9999, GPUs 8, CUDA: True
0: using world size: 8, data-parallel-size: 8, tensor-model-parallel size: 1, pipeline-model-parallel size: 1 
0: accumulate and all-reduce gradients in fp32 for bfloat16 data type.
0: using torch.bfloat16 for parameters ...
0: ------------------------ arguments ------------------------
0:   abort_on_unmet_fused_kernel_constraints ......... False
0:   accumulate_allreduce_grads_in_fp32 .............. True
0:   adam_beta1 ...................................... 0.9
0:   adam_beta2 ...................................... 0.999
0:   adam_eps ........................................ 1e-08
0:   adlr_autoresume ................................. False
0:   adlr_autoresume_interval ........................ 1000
0:   apply_query_key_layer_scaling ................... True
0:   apply_residual_connection_post_layernorm ........ False
0:   attention_dropout ............................... 0.1
0:   attention_softmax_in_fp32 ....................... False
0:   bert_binary_head ................................ True
0:   bert_load ....................................... None
0:   bf16 ............................................ True
0:   bias_dropout_fusion ............................. True
0:   bias_gelu_fusion ................................ True
0:   biencoder_projection_dim ........................ 0
0:   biencoder_shared_query_context_model ............ False
0:   block_data_path ................................. None
0:   checkpoint_activations .......................... True
0:   checkpoint_in_cpu ............................... False
0:   checkpoint_num_layers ........................... 1
0:   clip_grad ....................................... 1.0
0:   codecarbon_dir .................................. None
0:   consumed_train_samples .......................... 0
0:   consumed_train_tokens ........................... 0
0:   consumed_valid_samples .......................... 0
0:   contigious_checkpointing ........................ False
0:   cpu_optimizer ................................... False
0:   cpu_torch_adam .................................. False
0:   curriculum_learning ............................. False
0:   data_impl ....................................... mmap
0:   data_parallel_size .............................. 8
0:   data_path ....................................... None
0:   dataloader_type ................................. single
0:   DDP_impl ........................................ local
0:   decoder_seq_length .............................. None
0:   deepscale ....................................... False
0:   deepscale_config ................................ None
0:   deepspeed ....................................... True
0:   deepspeed_activation_checkpointing .............. False
0:   deepspeed_config ................................ ds_configs/3423734.json
0:   deepspeed_mpi ................................... False
0:   distribute_checkpointed_activations ............. False
0:   distributed_backend ............................. nccl
0:   embed_layernorm ................................. False
0:   embedding_path .................................. None
0:   encoder_seq_length .............................. 2048
0:   eod_mask_loss ................................... False
0:   eval_interval ................................... 1000
0:   eval_iters ...................................... 1
0:   eval_only ....................................... None
0:   evidence_data_path .............................. None
0:   exit_duration_in_mins ........................... None
0:   exit_interval ................................... None
0:   ffn_hidden_size ................................. 896
0:   finetune ........................................ False
0:   fp16 ............................................ False
0:   fp16_lm_cross_entropy ........................... False
0:   fp32_residual_connection ........................ False
0:   gigaflos_no_embeds .............................. 0
0:   global_batch_size ............................... 256
0:   glu_activation .................................. None
0:   hidden_dropout .................................. 0.1
0:   hidden_size ..................................... 224
0:   hysteresis ...................................... 2
0:   ict_head_size ................................... None
0:   ict_load ........................................ None
0:   img_dim ......................................... 224
0:   indexer_batch_size .............................. 128
0:   indexer_log_interval ............................ 1000
0:   inference ....................................... False
0:   init_method_std ................................. 0.02
0:   init_method_xavier_uniform ...................... False
0:   initial_loss_scale .............................. 4294967296
0:   kill_switch_path ................................ kill-switch-14m400m100m
0:   kv_channels ..................................... 32
0:   layer_norm_fusion ............................... True
0:   layernorm_epsilon ............................... 1e-05
0:   lazy_mpu_init ................................... None
0:   load ............................................ checkpoints_14m400m100m
0:   local_rank ...................................... None
0:   log_batch_size_to_tensorboard ................... True
0:   log_interval .................................... 10
0:   log_learning_rate_to_tensorboard ................ True
0:   log_level ....................................... None
0:   log_level_replica ............................... None
0:   log_loss_scale_to_tensorboard ................... True
0:   log_num_zeros_in_grad ........................... False
0:   log_params_norm ................................. False
0:   log_path ........................................ None
0:   log_timers_to_tensorboard ....................... True
0:   log_validation_ppl_to_tensorboard ............... True
0:   loss_on_targets_only ............................ False
0:   loss_scale ...................................... 12.0
0:   loss_scale_window ............................... 1000
0:   lr .............................................. 0.0002
0:   lr_decay_iters .................................. None
0:   lr_decay_samples ................................ 195313
0:   lr_decay_style .................................. cosine
0:   lr_decay_tokens ................................. None
0:   lr_warmup_fraction .............................. None
0:   lr_warmup_iters ................................. 0
0:   lr_warmup_samples ............................... 1953
0:   make_vocab_size_divisible_by .................... 128
0:   mask_prob ....................................... 0.15
0:   masked_softmax_fusion ........................... True
0:   max_position_embeddings ......................... 2048
0:   mean_noise_span_length .......................... None
0:   memory_centric_tiled_linear ..................... False
0:   merge_file ...................................... gpt2/merges.txt
0:   micro_batch_size ................................ 32
0:   min_loss_scale .................................. 1.0
0:   min_lr .......................................... 2e-05
0:   mmap_warmup ..................................... False
0:   no_load_optim ................................... None
0:   no_load_rng ..................................... None
0:   no_save_optim ................................... None
0:   no_save_rng ..................................... None
0:   noise_density ................................... None
0:   num_attention_heads ............................. 7
0:   num_channels .................................... 3
0:   num_classes ..................................... 1000
0:   num_layers ...................................... 4
0:   num_layers_per_virtual_pipeline_stage ........... None
0:   num_workers ..................................... 2
0:   onnx_safe ....................................... None
0:   openai_gelu ..................................... False
0:   optimizer ....................................... adam
0:   optimizer_fusion ................................ True
0:   override_lr_scheduler ........................... False
0:   pad_vocab_size_to ............................... None
0:   params_dtype .................................... torch.bfloat16
0:   partition_activations ........................... False
0:   patch_dim ....................................... 16
0:   pipeline_model_parallel_size .................... 1
0:   position_embedding_type ......................... PositionEmbeddingType.absolute
0:   pp_partition_method ............................. None
0:   profile_backward ................................ False
0:   query_in_block_prob ............................. 0.1
0:   rampup_batch_size ............................... None
0:   rank ............................................ 0
0:   remote_device ................................... none
0:   reset_attention_mask ............................ False
0:   reset_position_ids .............................. False
0:   reset_progress .................................. None
0:   retriever_report_topk_accuracies ................ []
0:   retriever_score_scaling ......................... False
0:   retriever_seq_length ............................ 256
0:   reweight_loss_based_on_position_frequency ....... False
0:   sample_rate ..................................... 1.0
0:   save ............................................ checkpoints_14m400m100m
0:   save_interval ................................... 1000
0:   scatter_gather_tensors_in_pipeline .............. True
0:   scattered_embeddings ............................ False
0:   seed ............................................ 1234
0:   seq_length ...................................... 2048
0:   sgd_momentum .................................... 0.9
0:   short_seq_prob .................................. 0.1
0:   skip_train_iteration_range ...................... None
0:   split ........................................... None
0:   split_transformers .............................. False
0:   sync_tp_duplicated_parameters ................... False
0:   synchronize_each_layer .......................... False
0:   tensor_model_parallel_size ...................... 1
0:   tensorboard_dir ................................. tensorboard_14m400m100m
0:   tensorboard_log_interval ........................ 1
0:   tensorboard_queue_size .......................... 5
0:   test_weighted_split_paths ....................... None
0:   test_weighted_split_paths_path .................. None
0:   tile_factor ..................................... 1
0:   titles_data_path ................................ None
0:   tokenizer_name_or_path .......................... None
0:   tokenizer_type .................................. GPT2BPETokenizer
0:   train_iters ..................................... None
0:   train_samples ................................... 195313
0:   train_tokens .................................... None
0:   train_weighted_split_names ...................... ['train']
0:   train_weighted_split_paths ...................... [['/scratch/project_462000119/data/c4_subsampled/gpt2tok_c4_en_100M_text_document']]
0:   train_weighted_split_paths_path ................. None
0:   train_weighted_split_splits ..................... [['0:1']]
0:   train_weighted_split_weights .................... [['1.0']]
0:   universal_checkpoint ............................ False
0:   use_bnb_optimizer ............................... False
0:   use_checkpoint_lr_scheduler ..................... False
0:   use_contiguous_buffers_in_ddp ................... True
0:   use_cpu_initialization .......................... None
0:   use_one_sent_docs ............................... False
0:   use_pin_memory .................................. False
0:   valid_num_workers ............................... 2
0:   valid_weighted_split_names ...................... ['validation']
0:   valid_weighted_split_paths ...................... [['/scratch/project_462000119/data/c4_validation/gpt2tok_c4validation_rerun_text_document']]
0:   valid_weighted_split_paths_path ................. None
0:   valid_weighted_split_splits ..................... [['0:1']]
0:   valid_weighted_split_weights .................... [['1.0']]
0:   virtual_pipeline_model_parallel_size ............ None
0:   vocab_extra_ids ................................. 0
0:   vocab_file ...................................... gpt2/vocab.json
0:   weight_decay .................................... 0.1
0:   world_size ...................................... 8
0:   zero_allgather_bucket_size ...................... 0.0
0:   zero_contigious_gradients ....................... False
0:   zero_reduce_bucket_size ......................... 0.0
0:   zero_reduce_scatter ............................. False
0:   zero_stage ...................................... 0
0: -------------------- end of arguments ---------------------
0: setting number of micro-batches to constant 1
0: > building GPT2BPETokenizer tokenizer ...
0:  > padded vocab (size: 50257) with 47 dummy tokens (new size: 50304)
0: DeepSpeed general environment info:
0: torch install path ............... ['/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/venv/lib/python3.9/site-packages/torch']
0: torch version .................... 1.13.0+rocm5.2
0: torch cuda version ............... None
0: torch hip version ................ 5.2.21151-afdc89f8
0: nvcc version ..................... None
0: deepspeed install path ........... ['/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/venv/lib/python3.9/site-packages/deepspeed']
0: deepspeed info ................... 0.7.5, unknown, unknown
0: deepspeed wheel compiled w. ...... torch 1.13, hip 5.1
0: **** Git info for Megatron: git_hash=unknown git_branch=unknown ****
0: > initializing torch distributed ...
0: [2023-04-27 15:38:51,769] [INFO] [comm.py:633:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl
0: > setting tensorboard ...
0: > initializing tensor model parallel with size 1
0: > initializing pipeline model parallel with size 1
0: > setting random seeds to 1234 ...
0: > initializing model parallel cuda seeds on global rank 0, model parallel rank 0, and data parallel rank 0 with model parallel seed: 3952 and data parallel seed: 1234
0: > compiling dataset index builder ...
0: make: Entering directory '/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/data'
0: make: Nothing to be done for 'default'.
0: make: Leaving directory '/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/data'
0: >>> done with dataset index builder. Compilation time: 0.112 seconds
0: > compiling and loading fused kernels ...
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax.cpp -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax_hip.cpp [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax_hip.h [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/compat.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/compat.h [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/type_shim.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/type_shim.h [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax_cuda.cu -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax_hip.hip [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/type_shim.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/type_shim.h [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/compat.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/compat.h [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax_hip.h [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax_hip.h [skipped, already hipified]
0: Total number of unsupported CUDA function calls: 0
0: 
0: 
0: Total number of replaced kernel launches: 87
0: ninja: no work to do.
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax.cpp -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax_hip.cpp [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax_cuda.cu -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax_hip.hip [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/type_shim.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/type_shim.h [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/compat.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/compat.h [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax_hip.h [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax_hip.h [skipped, already hipified]
0: Total number of unsupported CUDA function calls: 0
0: 
0: 
0: Total number of replaced kernel launches: 63
0: ninja: no work to do.
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/layer_norm_cuda.cpp -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/layer_norm_cuda.cpp [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/layer_norm_cuda_kernel.cu -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/layer_norm_hip_kernel.hip [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/type_shim.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/type_shim.h [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/compat.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/compat.h [skipped, no changes]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_upper_triang_masked_softmax_hip.h [skipped, already hipified]
0: /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax.h -> /pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/Megatron-DeepSpeed/megatron/fused_kernels/scaled_masked_softmax_hip.h [skipped, already hipified]
0: Total number of unsupported CUDA function calls: 0
0: 
0: 
0: Total number of replaced kernel launches: 67
0: [1/1] c++ layer_norm_hip_kernel.cuda.o layer_norm_cuda.o -shared -L/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/venv/lib/python3.9/site-packages/torch/lib -lc10 -lc10_hip -ltorch_cpu -ltorch_hip -ltorch -ltorch_python -L/opt/rocm/lib -lamdhip64 -o fused_mix_prec_layer_norm_cuda.so
0: >>> done with compiling and loading fused kernels. Compilation time: 10.743 seconds
0: time to initialize megatron (seconds): -24.979
0: [after megatron is initialized] datetime: 2023-04-27 15:39:03 
0: building GPT model ...
0: [2023-04-27 15:39:03,148] [INFO] [utils.py:827:see_memory_usage] Before Building Model
0: [2023-04-27 15:39:03,149] [INFO] [utils.py:828:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:03,149] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 37.57 GB, percent = 7.5%
0: SEED_LAYERS=False BASE_SEED=1234 SEED_FN=None
0: Using topology: {ProcessCoord(pipe=0, data=0, model=0): 0, ProcessCoord(pipe=0, data=1, model=0): 1, ProcessCoord(pipe=0, data=2, model=0): 2, ProcessCoord(pipe=0, data=3, model=0): 3, ProcessCoord(pipe=0, data=4, model=0): 4, ProcessCoord(pipe=0, data=5, model=0): 5, ProcessCoord(pipe=0, data=6, model=0): 6, ProcessCoord(pipe=0, data=7, model=0): 7}
0: [2023-04-27 15:39:03,378] [INFO] [module.py:366:_partition_layers] Partitioning pipeline stages with method type:transformer
0: stage=0 layers=11
0:      0: _to_float16
0:      1: EmbeddingPipe
0:      2: <lambda>
0:      3: ParallelTransformerLayerPipe
0:      4: ParallelTransformerLayerPipe
0:      5: ParallelTransformerLayerPipe
0:      6: ParallelTransformerLayerPipe
0:      7: undo
0:      8: MixedFusedLayerNorm
0:      9: EmbeddingPipe
0:     10: float16_to_fp32
0:   loss: CrossEntropy
0: [2023-04-27 15:39:03,577] [INFO] [utils.py:827:see_memory_usage] After Building Model
0: [2023-04-27 15:39:03,577] [INFO] [utils.py:828:see_memory_usage] MA 0.03 GB         Max_MA 0.03 GB         CA 0.05 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:03,577] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 37.58 GB, percent = 7.5%
0: setting training iterations to 762
0: > learning rate decay style: cosine
0: DeepSpeed is enabled.
0: [2023-04-27 15:39:03,578] [INFO] [logging.py:68:log_dist] [Rank 0] DeepSpeed info: version=0.7.5, git-hash=unknown, git-branch=unknown
0: [2023-04-27 15:39:08,175] [INFO] [logging.py:68:log_dist] [Rank 0] DeepSpeed Flops Profiler Enabled: False
0: [2023-04-27 15:39:08,175] [INFO] [logging.py:68:log_dist] [Rank 0] Removing param_group that has no 'params' in the client Optimizer
0: [2023-04-27 15:39:08,175] [INFO] [logging.py:68:log_dist] [Rank 0] Using client Optimizer as basic optimizer
0: [2023-04-27 15:39:08,176] [INFO] [logging.py:68:log_dist] [Rank 0] DeepSpeed Basic Optimizer = FusedAdam
0: [2023-04-27 15:39:08,176] [INFO] [logging.py:68:log_dist] [Rank 0] Creating BF16 optimizer
0: [2023-04-27 15:39:08,295] [INFO] [utils.py:827:see_memory_usage] begin bf16_optimizer
0: [2023-04-27 15:39:08,296] [INFO] [utils.py:828:see_memory_usage] MA 0.03 GB         Max_MA 0.03 GB         CA 0.05 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:08,296] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 41.07 GB, percent = 8.2%
0: ninja: no work to do.
0: Time to load utils op: 0.5155799388885498 seconds
0: Time to load utils op: 0.6356937885284424 seconds
0: Time to load utils op: 0.6354517936706543 seconds
0: Time to load utils op: 0.6361770629882812 secondsTime to load utils op: 0.6364157199859619 seconds
0: 
0: Time to load utils op: 0.6359162330627441 seconds
0: Time to load utils op: 0.6366944313049316 seconds
0: Time to load utils op: 0.6369485855102539 seconds
0: [2023-04-27 15:39:08,926] [INFO] [utils.py:827:see_memory_usage] before initializing group 0
0: [2023-04-27 15:39:08,926] [INFO] [utils.py:828:see_memory_usage] MA 0.03 GB         Max_MA 0.03 GB         CA 0.05 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:08,926] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 40.44 GB, percent = 8.0%
0: [2023-04-27 15:39:09,804] [INFO] [utils.py:827:see_memory_usage] after initializing group 0
0: [2023-04-27 15:39:09,805] [INFO] [utils.py:828:see_memory_usage] MA 0.08 GB         Max_MA 0.08 GB         CA 0.12 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:09,805] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 39.22 GB, percent = 7.8%
0: Time to load utils op: 0.005361318588256836 secondsTime to load utils op: 0.005542755126953125 seconds
0: 
0: Time to load utils op: 0.00555419921875 seconds
0: Time to load utils op: 0.005349397659301758 seconds
0: Time to load utils op: 0.005378007888793945 seconds
0: Time to load utils op: 0.005393028259277344 seconds
0: Time to load utils op: 0.005184650421142578 seconds
0: [2023-04-27 15:39:09,926] [INFO] [utils.py:827:see_memory_usage] before initializing group 1
0: [2023-04-27 15:39:09,926] [INFO] [utils.py:828:see_memory_usage] MA 0.08 GB         Max_MA 0.08 GB         CA 0.12 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:09,927] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 39.14 GB, percent = 7.8%
0: [2023-04-27 15:39:10,035] [INFO] [utils.py:827:see_memory_usage] after initializing group 1
0: [2023-04-27 15:39:10,036] [INFO] [utils.py:828:see_memory_usage] MA 0.09 GB         Max_MA 0.09 GB         CA 0.12 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:10,036] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 38.91 GB, percent = 7.7%
0: [2023-04-27 15:39:10,140] [INFO] [utils.py:827:see_memory_usage] before initializing group 2
0: [2023-04-27 15:39:10,140] [INFO] [utils.py:828:see_memory_usage] MA 0.09 GB         Max_MA 0.09 GB         CA 0.12 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:10,141] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 38.66 GB, percent = 7.7%
0: [2023-04-27 15:39:10,245] [INFO] [utils.py:827:see_memory_usage] after initializing group 2
0: [2023-04-27 15:39:10,245] [INFO] [utils.py:828:see_memory_usage] MA 0.09 GB         Max_MA 0.09 GB         CA 0.12 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:10,245] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 38.52 GB, percent = 7.7%
0: [2023-04-27 15:39:10,348] [INFO] [utils.py:827:see_memory_usage] before initialize_optimizer
0: [2023-04-27 15:39:10,348] [INFO] [utils.py:828:see_memory_usage] MA 0.09 GB         Max_MA 0.09 GB         CA 0.12 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:10,348] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 38.31 GB, percent = 7.6%
0: [2023-04-27 15:39:10,463] [INFO] [utils.py:827:see_memory_usage] end initialize_optimizer
0: [2023-04-27 15:39:10,463] [INFO] [utils.py:828:see_memory_usage] MA 0.1 GB         Max_MA 0.1 GB         CA 0.12 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:10,464] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 38.16 GB, percent = 7.6%
0: [2023-04-27 15:39:10,566] [INFO] [utils.py:827:see_memory_usage] end bf16_optimizer
0: [2023-04-27 15:39:10,566] [INFO] [utils.py:828:see_memory_usage] MA 0.1 GB         Max_MA 0.1 GB         CA 0.12 GB         Max_CA 0 GB 
0: [2023-04-27 15:39:10,566] [INFO] [utils.py:836:see_memory_usage] CPU Virtual Memory:  used = 38.0 GB, percent = 7.6%
0: [2023-04-27 15:39:10,566] [INFO] [logging.py:68:log_dist] [Rank 0] DeepSpeed Final Optimizer = FusedAdam
0: [2023-04-27 15:39:10,567] [INFO] [logging.py:68:log_dist] [Rank 0] DeepSpeed using client LR scheduler
0: [2023-04-27 15:39:10,567] [INFO] [logging.py:68:log_dist] [Rank 0] DeepSpeed LR Scheduler = <megatron.learning_rates.AnnealingLR object at 0x14d3e1c91fd0>
0: [2023-04-27 15:39:10,567] [INFO] [logging.py:68:log_dist] [Rank 0] step=0, skipped=0, lr=[0.0, 0.0, 0.0], mom=[(0.9, 0.999), (0.9, 0.999), (0.9, 0.999)]
0: [2023-04-27 15:39:10,567] [INFO] [config.py:1007:print] DeepSpeedEngine configuration:
0: [2023-04-27 15:39:10,567] [INFO] [config.py:1011:print]   activation_checkpointing_config  {
0:     "partition_activations": false, 
0:     "contiguous_memory_optimization": false, 
0:     "cpu_checkpointing": false, 
0:     "number_checkpoints": null, 
0:     "synchronize_checkpoint_boundary": false, 
0:     "profile": false
0: }
0: [2023-04-27 15:39:10,567] [INFO] [config.py:1011:print]   aio_config ................... {'block_size': 1048576, 'queue_depth': 8, 'thread_count': 1, 'single_submit': False, 'overlap_events': True}
0: [2023-04-27 15:39:10,567] [INFO] [config.py:1011:print]   amp_enabled .................. False
0: [2023-04-27 15:39:10,567] [INFO] [config.py:1011:print]   amp_params ................... False
0: [2023-04-27 15:39:10,567] [INFO] [config.py:1011:print]   autotuning_config ............ {
0:     "enabled": false, 
0:     "start_step": null, 
0:     "end_step": null, 
0:     "metric_path": null, 
0:     "arg_mappings": null, 
0:     "metric": "throughput", 
0:     "model_info": null, 
0:     "results_dir": "/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/autotuning_results", 
0:     "exps_dir": "/pfs/lustrep4/scratch/project_462000119/muennighoff/nov-2022-bettercom/autotuning_exps", 
0:     "overwrite": true, 
0:     "fast": true, 
0:     "start_profile_step": 3, 
0:     "end_profile_step": 5, 
0:     "tuner_type": "gridsearch", 
0:     "tuner_early_stopping": 5, 
0:     "tuner_num_trials": 50, 
0:     "model_info_path": null, 
0:     "mp_size": 1, 
0:     "max_train_batch_size": null, 
0:     "min_train_batch_size": 1, 
0:     "max_train_micro_batch_size_per_gpu": 1.024000e+03, 
0:     "min_train_micro_batch_size_per_gpu": 1, 
0:     "num_tuning_micro_batch_sizes": 3
0: }
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   bfloat16_enabled ............. True
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   checkpoint_parallel_write_pipeline  False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   checkpoint_tag_validation_enabled  True
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   checkpoint_tag_validation_fail  False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   comms_config ................. <deepspeed.comm.config.DeepSpeedCommsConfig object at 0x14d3e1c91d30>
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   communication_data_type ...... None
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   compression_config ........... {'weight_quantization': {'shared_parameters': {'enabled': False, 'quantizer_kernel': False, 'schedule_offset': 0, 'quantize_groups': 1, 'quantize_verbose': False, 'quantization_type': 'symmetric', 'quantize_weight_in_forward': False, 'rounding': 'nearest', 'fp16_mixed_quantize': False, 'quantize_change_ratio': 0.001}, 'different_groups': {}}, 'activation_quantization': {'shared_parameters': {'enabled': False, 'quantization_type': 'symmetric', 'range_calibration': 'dynamic', 'schedule_offset': 1000}, 'different_groups': {}}, 'sparse_pruning': {'shared_parameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'row_pruning': {'shared_parameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'head_pruning': {'shared_parameters': {'enabled': False, 'method': 'topk', 'schedule_offset': 1000}, 'different_groups': {}}, 'channel_pruning': {'shared_pa
0: rameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'layer_reduction': {'enabled': False}}
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   curriculum_enabled ........... False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   curriculum_params ............ False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   dataloader_drop_last ......... False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   disable_allgather ............ False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   dump_state ................... False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   dynamic_loss_scale_args ...... None
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   eigenvalue_enabled ........... False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   eigenvalue_gas_boundary_resolution  1
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   eigenvalue_layer_name ........ bert.encoder.layer
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   eigenvalue_layer_num ......... 0
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   eigenvalue_max_iter .......... 100
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   eigenvalue_stability ......... 1e-06
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   eigenvalue_tol ............... 0.01
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   eigenvalue_verbose ........... False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   elasticity_enabled ........... False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   flops_profiler_config ........ {
0:     "enabled": false, 
0:     "profile_step": 1, 
0:     "module_depth": -1, 
0:     "top_modules": 1, 
0:     "detailed": true, 
0:     "output_file": null
0: }
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   fp16_auto_cast ............... None
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   fp16_enabled ................. False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   fp16_master_weights_and_gradients  False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   global_rank .................. 0
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   gradient_accumulation_steps .. 1
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   gradient_clipping ............ 1.0
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   gradient_predivide_factor .... 1.0
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   initial_dynamic_scale ........ 1
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   load_universal_checkpoint .... False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   loss_scale ................... 1.0
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   memory_breakdown ............. False
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   monitor_config ............... <deepspeed.monitor.config.DeepSpeedMonitorConfig object at 0x14d3e1c91ca0>
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   nebula_config ................ {
0:     "enabled": false, 
0:     "persistent_storage_path": null, 
0:     "persistent_time_interval": 100, 
0:     "num_of_version_in_retention": 2, 
0:     "enable_nebula_load": true, 
0:     "load_path": null
0: }
0: [2023-04-27 15:39:10,568] [INFO] [config.py:1011:print]   optimizer_legacy_fusion ...... False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   optimizer_name ............... None
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   optimizer_params ............. None
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   pipeline ..................... {'stages': 'auto', 'partition': 'best', 'seed_layers': False, 'activation_checkpoint_interval': 0}
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   pld_enabled .................. False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   pld_params ................... False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   prescale_gradients ........... False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   scheduler_name ............... None
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   scheduler_params ............. None
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   sparse_attention ............. None
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   sparse_gradients_enabled ..... False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   steps_per_print .............. 2000
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   train_batch_size ............. 256
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   train_micro_batch_size_per_gpu  32
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   use_node_local_storage ....... False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   wall_clock_breakdown ......... False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   world_size ................... 8
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   zero_allow_untested_optimizer  False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   zero_config .................. stage=0 contiguous_gradients=True reduce_scatter=True reduce_bucket_size=500000000 allgather_partitions=True allgather_bucket_size=500000000 overlap_comm=False load_from_fp32_weights=True elastic_checkpoint=False offload_param=None offload_optimizer=None sub_group_size=1000000000 cpu_offload_param=None cpu_offload_use_pin_memory=None cpu_offload=None prefetch_bucket_size=50000000 param_persistence_threshold=100000 model_persistence_threshold=9223372036854775807 max_live_parameters=1000000000 max_reuse_distance=1000000000 gather_16bit_weights_on_model_save=False stage3_gather_fp16_weights_on_model_save=False ignore_unused_parameters=True legacy_stage1=False round_robin_gradients=False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   zero_enabled ................. False
0: [2023-04-27 15:39:10,569] [INFO] [config.py:1011:print]   zero_optimization_stage ...... 0
0: [2023-04-27 15:39:10,569] [INFO] [config.py:996:print_user_config]   json = {
0:     "train_micro_batch_size_per_gpu": 32, 
0:     "train_batch_size": 256, 
0:     "gradient_clipping": 1.0, 
0:     "zero_optimization": {
0:         "stage": 0
0:     }, 
0:     "bf16": {
0:         "enabled": true
0:     }, 
0:     "steps_per_print": 2.000000e+03, 
0:     "wall_clock_breakdown": false
0: }
0: Time to load utils op: 0.0004215240478515625 seconds
0: [2023-04-27 15:39:10,570] [INFO] [engine.py:87:__init__] CONFIG: micro_batches=1 micro_batch_size=32
0: [2023-04-27 15:39:10,711] [INFO] [engine.py:145:__init__] RANK=0 STAGE=0 LAYERS=11 [0, 11) STAGE_PARAMS=14147392 (14.147M) TOTAL_PARAMS=14147392 (14.147M) UNIQUE_PARAMS=14147392 (14.147M)
0: [2023-04-27 15:39:10,713] [WARNING] [engine.py:2581:load_checkpoint] Unable to find latest file at checkpoints_14m400m100m/latest, if trying to load latest checkpoint please ensure this file exists or pass an explicit checkpoint tag when loading a checkpoint.
0: WARNING: could not find the metadata file checkpoints_14m400m100m 
0:     will not load any checkpoints and will start from random
0: [2023-04-27 15:39:10,713] [WARNING] [engine.py:2581:load_checkpoint] Unable to find latest file at checkpoints_14m400m100m/latest, if trying to load latest checkpoint please ensure this file exists or pass an explicit checkpoint tag when loading a checkpoint.
0: [2023-04-27 15:39:10,713] [WARNING] [engine.py:2581:load_checkpoint] Unable to find latest file at checkpoints_14m400m100m/latest, if trying to load latest checkpoint please ensure this file exists or pass an explicit checkpoint tag when loading a checkpoint.
0: [2023-04-27 15:39:10,713] [WARNING] [engine.py:2581:load_checkpoint] Unable to find latest file at checkpoints_14m400m100m/latest, if trying to load latest checkpoint please ensure this file exists or pass an explicit checkpoint tag when loading a checkpoint.
0: [2023-04-27 15:39:10,713] [WARNING] [engine.py:2581:load_checkpoint] Unable to find latest file at checkpoints_14m400m100m/latest, if trying to load latest checkpoint please ensure this file exists or pass an explicit checkpoint tag when loading a checkpoint.
0: [2023-04-27 15:39:10,713] [WARNING] [engine.py:2581:load_checkpoint] Unable to find latest file at checkpoints_14m400m100m/latest, if trying to load latest checkpoint please ensure this file exists or pass an explicit checkpoint tag when loading a checkpoint.
0: [2023-04-27 15:39:10,714] [WARNING] [engine.py:2581:load_checkpoint] Unable to find latest file at checkpoints_14m400m100m/latest, if trying to load latest checkpoint please ensure this file exists or pass an explicit checkpoint tag when loading a checkpoint.
0: [2023-04-27 15:39:10,714] [WARNING] [engine.py:2581:load_checkpoint] Unable to find latest file at checkpoints_14m400m100m/latest, if trying to load latest checkpoint please ensure this file exists or pass an explicit checkpoint tag when loading a checkpoint.
0: time (ms) | load-checkpoint: 1.15
0: estimated model parameters: 0.014147392
0: estimated model parameters without embeddings: 0.002420544
0: [after model, optimizer, and learning rate scheduler are built] datetime: 2023-04-27 15:39:11 
0: > building train, validation, and test datasets ...
0:  > datasets target sizes (minimum size):
0:     train:      195313
0:     validation: 256
0:     test:       256
0: > building train, validation, and test datasets for GPT ...
0:  > building dataset index ...
0:     reading sizes...
0:     reading pointers...
0:     reading document index...
0:     creating numpy buffer of mmap...
0:     creating memory view of numpy buffer...
0:  > finished creating indexed dataset in 0.033320 seconds
0:     number of documents: 208931
0:  > dataset split:
0:     train:
0:      document indices in [0, 208931) total of 208931 documents
0:  > WARNING: could not find index map files, building the indices on rank 0 ...
0:  > last epoch number of samples (94) is smaller than 95.0% of number of samples per epoch (48804), setting separate_last_epoch to True
0:  > elasped time to build and save doc-idx mapping (seconds): 0.081743
0:     using:
0:      number of documents:       208931
0:      number of epochs:          5
0:      sequence length:           2048
0:      total number of samples:   244024
0:  > elasped time to build and save sample-idx mapping (seconds): 0.045406
0:  > building shuffle index with split [0, 195219) and [195219, 244024) ...
0:  > elasped time to build and save shuffle-idx mapping (seconds): 0.006500
0:  > loading doc-idx mapping from /scratch/project_462000119/data/c4_subsampled/gpt2tok_c4_en_100M_text_document_train_indexmap_195313ns_2048sl_1234s_doc_idx.npy
0:  > loading sample-idx mapping from /scratch/project_462000119/data/c4_subsampled/gpt2tok_c4_en_100M_text_document_train_indexmap_195313ns_2048sl_1234s_sample_idx.npy
0:  > loading shuffle-idx mapping from /scratch/project_462000119/data/c4_subsampled/gpt2tok_c4_en_100M_text_document_train_indexmap_195313ns_2048sl_1234s_shuffle_idx.npy
0:     loaded indexed file in 0.002 seconds
0:     total number of samples: 244025
0:     total number of epochs: 5
0:  > building dataset index ...
0:     reading sizes...
0:     reading pointers...
0:     reading document index...
0:     creating numpy buffer of mmap...
0:     creating memory view of numpy buffer...
0:  > finished creating indexed dataset in 0.038422 seconds
0:     number of documents: 364608
0:  > dataset split:
0:     validation:
0:      document indices in [0, 364608) total of 364608 documents
0:  > loading doc-idx mapping from /scratch/project_462000119/data/c4_validation/gpt2tok_c4validation_rerun_text_document_validation_indexmap_256ns_2048sl_1234s_doc_idx.npy
0:  > loading sample-idx mapping from /scratch/project_462000119/data/c4_validation/gpt2tok_c4validation_rerun_text_document_validation_indexmap_256ns_2048sl_1234s_sample_idx.npy
0:  > loading shuffle-idx mapping from /scratch/project_462000119/data/c4_validation/gpt2tok_c4validation_rerun_text_document_validation_indexmap_256ns_2048sl_1234s_shuffle_idx.npy
0:     loaded indexed file in 0.125 seconds
0:     total number of samples: 84978
0:     total number of epochs: 1
0: > finished creating GPT datasets ...
0: [after dataloaders are built] datetime: 2023-04-27 15:39:19 
0: done with setup ...
0: training ...
0: Number of parameters: [tensor rank - pipeline rank] w/ and w/o embeddings:
0: time (ms) | model-and-optimizer-setup: 8297.11 | train/valid/test-data-iterators-setup: 8101.39
0: [000-000] 0.0141B / 0.0024B
0: [before the start of training step] datetime: 2023-04-27 15:39:19 
0: [2023-04-27 15:39:20,213] [INFO] [checkpointing.py:553:forward] Activation Checkpointing Information
0: [2023-04-27 15:39:20,213] [INFO] [checkpointing.py:554:forward] ----Partition Activations False, CPU CHECKPOINTING False
0: [2023-04-27 15:39:20,213] [INFO] [checkpointing.py:557:forward] ----contiguous Memory Checkpointing False with None total layers
0: [2023-04-27 15:39:20,213] [INFO] [checkpointing.py:560:forward] ----Synchronization False
0: [2023-04-27 15:39:20,213] [INFO] [checkpointing.py:561:forward] ----Profiling time in checkpointing False
0: [Rank 0] (after 10 iterations) memory (MB) | allocated: 12710.28759765625 | max allocated: 31761.787109375 | reserved: 39838.0 | max reserved: 39838.0
0:  iteration       10/     762 | consumed samples:         2560 | consumed tokens:      5242880 | elapsed time per iteration (s): 1.09 | learning rate: 2.000E-04 | global batch size:   256 | lm loss: 1.061609E+01 | grad norm: 1.244 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 235.427 | TFLOPs: 7.01 |
0:  iteration       20/     762 | consumed samples:         5120 | consumed tokens:     10485760 | elapsed time per iteration (s): 0.47 | learning rate: 1.999E-04 | global batch size:   256 | lm loss: 1.004091E+01 | grad norm: 1.246 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 543.302 | TFLOPs: 16.17 |
0:  iteration       30/     762 | consumed samples:         7680 | consumed tokens:     15728640 | elapsed time per iteration (s): 0.47 | learning rate: 1.996E-04 | global batch size:   256 | lm loss: 9.494904E+00 | grad norm: 1.238 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 542.983 | TFLOPs: 16.16 |
0:  iteration       40/     762 | consumed samples:        10240 | consumed tokens:     20971520 | elapsed time per iteration (s): 0.47 | learning rate: 1.992E-04 | global batch size:   256 | lm loss: 8.996175E+00 | grad norm: 1.207 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 541.361 | TFLOPs: 16.11 |
0:  iteration       50/     762 | consumed samples:        12800 | consumed tokens:     26214400 | elapsed time per iteration (s): 0.47 | learning rate: 1.986E-04 | global batch size:   256 | lm loss: 8.595522E+00 | grad norm: 1.171 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 542.873 | TFLOPs: 16.15 |
0:  iteration       60/     762 | consumed samples:        15360 | consumed tokens:     31457280 | elapsed time per iteration (s): 0.47 | learning rate: 1.979E-04 | global batch size:   256 | lm loss: 8.256769E+00 | grad norm: 1.120 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 542.555 | TFLOPs: 16.14 |
0:  iteration       70/     762 | consumed samples:        17920 | consumed tokens:     36700160 | elapsed time per iteration (s): 0.47 | learning rate: 1.970E-04 | global batch size:   256 | lm loss: 7.974753E+00 | grad norm: 1.029 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 542.449 | TFLOPs: 16.14 |
0:  iteration       80/     762 | consumed samples:        20480 | consumed tokens:     41943040 | elapsed time per iteration (s): 0.47 | learning rate: 1.960E-04 | global batch size:   256 | lm loss: 7.762483E+00 | grad norm: 0.871 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 542.769 | TFLOPs: 16.15 |
0:  iteration       90/     762 | consumed samples:        23040 | consumed tokens:     47185920 | elapsed time per iteration (s): 0.47 | learning rate: 1.948E-04 | global batch size:   256 | lm loss: 7.601062E+00 | grad norm: 0.677 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 542.280 | TFLOPs: 16.14 |
0:  iteration      100/     762 | consumed samples:        25600 | consumed tokens:     52428800 | elapsed time per iteration (s): 0.47 | learning rate: 1.934E-04 | global batch size:   256 | lm loss: 7.462014E+00 | grad norm: 0.501 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 542.328 | TFLOPs: 16.14 |
0:  iteration      110/     762 | consumed samples:        28160 | consumed tokens:     57671680 | elapsed time per iteration (s): 0.47 | learning rate: 1.920E-04 | global batch size:   256 | lm loss: 7.354394E+00 | grad norm: 0.557 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 541.922 | TFLOPs: 16.13 |
0:  iteration      120/     762 | consumed samples:        30720 | consumed tokens:     62914560 | elapsed time per iteration (s): 0.47 | learning rate: 1.903E-04 | global batch size:   256 | lm loss: 7.264880E+00 | grad norm: 0.365 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 541.234 | TFLOPs: 16.11 |
0:  iteration      130/     762 | consumed samples:        33280 | consumed tokens:     68157440 | elapsed time per iteration (s): 0.47 | learning rate: 1.886E-04 | global batch size:   256 | lm loss: 7.188472E+00 | grad norm: 0.769 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 541.257 | TFLOPs: 16.11 |
0:  iteration      140/     762 | consumed samples:        35840 | consumed tokens:     73400320 | elapsed time per iteration (s): 0.47 | learning rate: 1.867E-04 | global batch size:   256 | lm loss: 7.122352E+00 | grad norm: 0.407 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 541.103 | TFLOPs: 16.10 |
0:  iteration      150/     762 | consumed samples:        38400 | consumed tokens:     78643200 | elapsed time per iteration (s): 0.47 | learning rate: 1.847E-04 | global batch size:   256 | lm loss: 7.034540E+00 | grad norm: 0.289 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 540.904 | TFLOPs: 16.10 |
0:  iteration      160/     762 | consumed samples:        40960 | consumed tokens:     83886080 | elapsed time per iteration (s): 0.47 | learning rate: 1.825E-04 | global batch size:   256 | lm loss: 6.989355E+00 | grad norm: 0.299 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 540.716 | TFLOPs: 16.09 |
0:  iteration      170/     762 | consumed samples:        43520 | consumed tokens:     89128960 | elapsed time per iteration (s): 0.47 | learning rate: 1.802E-04 | global batch size:   256 | lm loss: 6.930446E+00 | grad norm: 0.715 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 540.261 | TFLOPs: 16.08 |
0:  iteration      180/     762 | consumed samples:        46080 | consumed tokens:     94371840 | elapsed time per iteration (s): 0.47 | learning rate: 1.778E-04 | global batch size:   256 | lm loss: 6.885769E+00 | grad norm: 0.290 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 540.145 | TFLOPs: 16.07 |
0:  iteration      190/     762 | consumed samples:        48640 | consumed tokens:     99614720 | elapsed time per iteration (s): 0.47 | learning rate: 1.753E-04 | global batch size:   256 | lm loss: 6.842728E+00 | grad norm: 0.381 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 540.206 | TFLOPs: 16.07 |
0:  iteration      200/     762 | consumed samples:        51200 | consumed tokens:    104857600 | elapsed time per iteration (s): 0.47 | learning rate: 1.727E-04 | global batch size:   256 | lm loss: 6.807191E+00 | grad norm: 0.362 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.872 | TFLOPs: 16.06 |
0:  iteration      210/     762 | consumed samples:        53760 | consumed tokens:    110100480 | elapsed time per iteration (s): 0.47 | learning rate: 1.700E-04 | global batch size:   256 | lm loss: 6.770544E+00 | grad norm: 0.271 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.910 | TFLOPs: 16.07 |
0:  iteration      220/     762 | consumed samples:        56320 | consumed tokens:    115343360 | elapsed time per iteration (s): 0.47 | learning rate: 1.671E-04 | global batch size:   256 | lm loss: 6.739288E+00 | grad norm: 0.334 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.810 | TFLOPs: 16.06 |
0:  iteration      230/     762 | consumed samples:        58880 | consumed tokens:    120586240 | elapsed time per iteration (s): 0.47 | learning rate: 1.642E-04 | global batch size:   256 | lm loss: 6.705843E+00 | grad norm: 0.269 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.936 | TFLOPs: 16.07 |
0:  iteration      240/     762 | consumed samples:        61440 | consumed tokens:    125829120 | elapsed time per iteration (s): 0.47 | learning rate: 1.611E-04 | global batch size:   256 | lm loss: 6.686831E+00 | grad norm: 0.320 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.539 | TFLOPs: 16.05 |
0:  iteration      250/     762 | consumed samples:        64000 | consumed tokens:    131072000 | elapsed time per iteration (s): 0.47 | learning rate: 1.580E-04 | global batch size:   256 | lm loss: 6.666966E+00 | grad norm: 0.823 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.613 | TFLOPs: 16.06 |
0:  iteration      260/     762 | consumed samples:        66560 | consumed tokens:    136314880 | elapsed time per iteration (s): 0.47 | learning rate: 1.548E-04 | global batch size:   256 | lm loss: 6.624731E+00 | grad norm: 0.583 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.488 | TFLOPs: 16.05 |
0:  iteration      270/     762 | consumed samples:        69120 | consumed tokens:    141557760 | elapsed time per iteration (s): 0.47 | learning rate: 1.515E-04 | global batch size:   256 | lm loss: 6.607779E+00 | grad norm: 0.377 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.421 | TFLOPs: 16.05 |
0:  iteration      280/     762 | consumed samples:        71680 | consumed tokens:    146800640 | elapsed time per iteration (s): 0.47 | learning rate: 1.482E-04 | global batch size:   256 | lm loss: 6.584142E+00 | grad norm: 0.255 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.390 | TFLOPs: 16.05 |
0:  iteration      290/     762 | consumed samples:        74240 | consumed tokens:    152043520 | elapsed time per iteration (s): 0.47 | learning rate: 1.447E-04 | global batch size:   256 | lm loss: 6.567078E+00 | grad norm: 0.539 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.359 | TFLOPs: 16.05 |
0:  iteration      300/     762 | consumed samples:        76800 | consumed tokens:    157286400 | elapsed time per iteration (s): 0.47 | learning rate: 1.413E-04 | global batch size:   256 | lm loss: 6.562552E+00 | grad norm: 0.358 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.196 | TFLOPs: 16.04 |
0:  iteration      310/     762 | consumed samples:        79360 | consumed tokens:    162529280 | elapsed time per iteration (s): 0.48 | learning rate: 1.377E-04 | global batch size:   256 | lm loss: 6.553120E+00 | grad norm: 0.317 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.939 | TFLOPs: 16.04 |
0:  iteration      320/     762 | consumed samples:        81920 | consumed tokens:    167772160 | elapsed time per iteration (s): 0.47 | learning rate: 1.341E-04 | global batch size:   256 | lm loss: 6.535150E+00 | grad norm: 0.206 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 539.030 | TFLOPs: 16.04 |
0:  iteration      330/     762 | consumed samples:        84480 | consumed tokens:    173015040 | elapsed time per iteration (s): 0.48 | learning rate: 1.305E-04 | global batch size:   256 | lm loss: 6.514146E+00 | grad norm: 0.316 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.774 | TFLOPs: 16.03 |
0:  iteration      340/     762 | consumed samples:        87040 | consumed tokens:    178257920 | elapsed time per iteration (s): 0.48 | learning rate: 1.269E-04 | global batch size:   256 | lm loss: 6.501083E+00 | grad norm: 0.379 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.669 | TFLOPs: 16.03 |
0:  iteration      350/     762 | consumed samples:        89600 | consumed tokens:    183500800 | elapsed time per iteration (s): 0.48 | learning rate: 1.232E-04 | global batch size:   256 | lm loss: 6.489027E+00 | grad norm: 0.294 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.630 | TFLOPs: 16.03 |
0:  iteration      360/     762 | consumed samples:        92160 | consumed tokens:    188743680 | elapsed time per iteration (s): 0.48 | learning rate: 1.194E-04 | global batch size:   256 | lm loss: 6.469370E+00 | grad norm: 0.372 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.671 | TFLOPs: 16.03 |
0:  iteration      370/     762 | consumed samples:        94720 | consumed tokens:    193986560 | elapsed time per iteration (s): 0.48 | learning rate: 1.157E-04 | global batch size:   256 | lm loss: 6.475340E+00 | grad norm: 0.430 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.886 | TFLOPs: 16.04 |
0:  iteration      380/     762 | consumed samples:        97280 | consumed tokens:    199229440 | elapsed time per iteration (s): 0.48 | learning rate: 1.120E-04 | global batch size:   256 | lm loss: 6.449596E+00 | grad norm: 0.287 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.791 | TFLOPs: 16.03 |
0:  iteration      390/     762 | consumed samples:        99840 | consumed tokens:    204472320 | elapsed time per iteration (s): 0.48 | learning rate: 1.082E-04 | global batch size:   256 | lm loss: 6.435555E+00 | grad norm: 0.226 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.826 | TFLOPs: 16.03 |
0:  iteration      400/     762 | consumed samples:       102400 | consumed tokens:    209715200 | elapsed time per iteration (s): 0.48 | learning rate: 1.045E-04 | global batch size:   256 | lm loss: 6.432314E+00 | grad norm: 0.329 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.433 | TFLOPs: 16.02 |
0:  iteration      410/     762 | consumed samples:       104960 | consumed tokens:    214958080 | elapsed time per iteration (s): 0.48 | learning rate: 1.008E-04 | global batch size:   256 | lm loss: 6.421754E+00 | grad norm: 0.259 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.692 | TFLOPs: 16.03 |
0:  iteration      420/     762 | consumed samples:       107520 | consumed tokens:    220200960 | elapsed time per iteration (s): 0.48 | learning rate: 9.705E-05 | global batch size:   256 | lm loss: 6.429256E+00 | grad norm: 0.380 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.636 | TFLOPs: 16.03 |
0:  iteration      430/     762 | consumed samples:       110080 | consumed tokens:    225443840 | elapsed time per iteration (s): 0.48 | learning rate: 9.336E-05 | global batch size:   256 | lm loss: 6.417949E+00 | grad norm: 0.431 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.431 | TFLOPs: 16.02 |
0:  iteration      440/     762 | consumed samples:       112640 | consumed tokens:    230686720 | elapsed time per iteration (s): 0.48 | learning rate: 8.969E-05 | global batch size:   256 | lm loss: 6.409863E+00 | grad norm: 0.213 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.549 | TFLOPs: 16.03 |
0:  iteration      450/     762 | consumed samples:       115200 | consumed tokens:    235929600 | elapsed time per iteration (s): 0.48 | learning rate: 8.607E-05 | global batch size:   256 | lm loss: 6.397947E+00 | grad norm: 0.349 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.415 | TFLOPs: 16.02 |
0:  iteration      460/     762 | consumed samples:       117760 | consumed tokens:    241172480 | elapsed time per iteration (s): 0.48 | learning rate: 8.248E-05 | global batch size:   256 | lm loss: 6.390238E+00 | grad norm: 0.224 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.530 | TFLOPs: 16.02 |
0:  iteration      470/     762 | consumed samples:       120320 | consumed tokens:    246415360 | elapsed time per iteration (s): 0.48 | learning rate: 7.894E-05 | global batch size:   256 | lm loss: 6.382044E+00 | grad norm: 0.236 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.441 | TFLOPs: 16.02 |
0:  iteration      480/     762 | consumed samples:       122880 | consumed tokens:    251658240 | elapsed time per iteration (s): 0.48 | learning rate: 7.545E-05 | global batch size:   256 | lm loss: 6.380611E+00 | grad norm: 0.229 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.596 | TFLOPs: 16.03 |
0:  iteration      490/     762 | consumed samples:       125440 | consumed tokens:    256901120 | elapsed time per iteration (s): 0.48 | learning rate: 7.203E-05 | global batch size:   256 | lm loss: 6.374780E+00 | grad norm: 0.195 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.362 | TFLOPs: 16.02 |
0:  iteration      500/     762 | consumed samples:       128000 | consumed tokens:    262144000 | elapsed time per iteration (s): 0.48 | learning rate: 6.867E-05 | global batch size:   256 | lm loss: 6.364036E+00 | grad norm: 0.226 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.366 | TFLOPs: 16.02 |
0:  iteration      510/     762 | consumed samples:       130560 | consumed tokens:    267386880 | elapsed time per iteration (s): 0.48 | learning rate: 6.538E-05 | global batch size:   256 | lm loss: 6.350660E+00 | grad norm: 0.249 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.078 | TFLOPs: 16.01 |
0:  iteration      520/     762 | consumed samples:       133120 | consumed tokens:    272629760 | elapsed time per iteration (s): 0.48 | learning rate: 6.217E-05 | global batch size:   256 | lm loss: 6.343756E+00 | grad norm: 0.266 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.315 | TFLOPs: 16.02 |
0:  iteration      530/     762 | consumed samples:       135680 | consumed tokens:    277872640 | elapsed time per iteration (s): 0.48 | learning rate: 5.904E-05 | global batch size:   256 | lm loss: 6.350199E+00 | grad norm: 0.240 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.223 | TFLOPs: 16.02 |
0:  iteration      540/     762 | consumed samples:       138240 | consumed tokens:    283115520 | elapsed time per iteration (s): 0.48 | learning rate: 5.600E-05 | global batch size:   256 | lm loss: 6.352103E+00 | grad norm: 0.270 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.137 | TFLOPs: 16.01 |
0:  iteration      550/     762 | consumed samples:       140800 | consumed tokens:    288358400 | elapsed time per iteration (s): 0.48 | learning rate: 5.305E-05 | global batch size:   256 | lm loss: 6.339139E+00 | grad norm: 0.197 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.368 | TFLOPs: 16.02 |
0:  iteration      560/     762 | consumed samples:       143360 | consumed tokens:    293601280 | elapsed time per iteration (s): 0.48 | learning rate: 5.020E-05 | global batch size:   256 | lm loss: 6.343401E+00 | grad norm: 0.232 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.296 | TFLOPs: 16.02 |
0:  iteration      570/     762 | consumed samples:       145920 | consumed tokens:    298844160 | elapsed time per iteration (s): 0.48 | learning rate: 4.746E-05 | global batch size:   256 | lm loss: 6.333387E+00 | grad norm: 0.186 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.089 | TFLOPs: 16.01 |
0:  iteration      580/     762 | consumed samples:       148480 | consumed tokens:    304087040 | elapsed time per iteration (s): 0.48 | learning rate: 4.482E-05 | global batch size:   256 | lm loss: 6.321525E+00 | grad norm: 0.189 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.255 | TFLOPs: 16.02 |
0:  iteration      590/     762 | consumed samples:       151040 | consumed tokens:    309329920 | elapsed time per iteration (s): 0.48 | learning rate: 4.230E-05 | global batch size:   256 | lm loss: 6.331749E+00 | grad norm: 0.209 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.297 | TFLOPs: 16.02 |
0:  iteration      600/     762 | consumed samples:       153600 | consumed tokens:    314572800 | elapsed time per iteration (s): 0.48 | learning rate: 3.989E-05 | global batch size:   256 | lm loss: 6.320105E+00 | grad norm: 0.295 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.059 | TFLOPs: 16.01 |
0:  iteration      610/     762 | consumed samples:       156160 | consumed tokens:    319815680 | elapsed time per iteration (s): 0.48 | learning rate: 3.760E-05 | global batch size:   256 | lm loss: 6.336096E+00 | grad norm: 0.176 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.082 | TFLOPs: 16.01 |
0:  iteration      620/     762 | consumed samples:       158720 | consumed tokens:    325058560 | elapsed time per iteration (s): 0.48 | learning rate: 3.544E-05 | global batch size:   256 | lm loss: 6.320390E+00 | grad norm: 0.174 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.062 | TFLOPs: 16.01 |
0:  iteration      630/     762 | consumed samples:       161280 | consumed tokens:    330301440 | elapsed time per iteration (s): 0.48 | learning rate: 3.341E-05 | global batch size:   256 | lm loss: 6.331472E+00 | grad norm: 0.225 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.228 | TFLOPs: 16.02 |
0:  iteration      640/     762 | consumed samples:       163840 | consumed tokens:    335544320 | elapsed time per iteration (s): 0.48 | learning rate: 3.151E-05 | global batch size:   256 | lm loss: 6.302699E+00 | grad norm: 0.305 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.062 | TFLOPs: 16.01 |
0:  iteration      650/     762 | consumed samples:       166400 | consumed tokens:    340787200 | elapsed time per iteration (s): 0.48 | learning rate: 2.975E-05 | global batch size:   256 | lm loss: 6.316518E+00 | grad norm: 0.215 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.175 | TFLOPs: 16.01 |
0:  iteration      660/     762 | consumed samples:       168960 | consumed tokens:    346030080 | elapsed time per iteration (s): 0.48 | learning rate: 2.812E-05 | global batch size:   256 | lm loss: 6.302271E+00 | grad norm: 0.166 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 537.983 | TFLOPs: 16.01 |
0:  iteration      670/     762 | consumed samples:       171520 | consumed tokens:    351272960 | elapsed time per iteration (s): 0.48 | learning rate: 2.664E-05 | global batch size:   256 | lm loss: 6.307047E+00 | grad norm: 0.173 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 537.972 | TFLOPs: 16.01 |
0:  iteration      680/     762 | consumed samples:       174080 | consumed tokens:    356515840 | elapsed time per iteration (s): 0.48 | learning rate: 2.530E-05 | global batch size:   256 | lm loss: 6.308545E+00 | grad norm: 0.185 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.056 | TFLOPs: 16.01 |
0:  iteration      690/     762 | consumed samples:       176640 | consumed tokens:    361758720 | elapsed time per iteration (s): 0.48 | learning rate: 2.411E-05 | global batch size:   256 | lm loss: 6.287326E+00 | grad norm: 0.187 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.159 | TFLOPs: 16.01 |
0:  iteration      700/     762 | consumed samples:       179200 | consumed tokens:    367001600 | elapsed time per iteration (s): 0.48 | learning rate: 2.307E-05 | global batch size:   256 | lm loss: 6.303761E+00 | grad norm: 0.205 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.083 | TFLOPs: 16.01 |
0:  iteration      710/     762 | consumed samples:       181760 | consumed tokens:    372244480 | elapsed time per iteration (s): 0.48 | learning rate: 2.217E-05 | global batch size:   256 | lm loss: 6.306910E+00 | grad norm: 0.159 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.039 | TFLOPs: 16.01 |
0:  iteration      720/     762 | consumed samples:       184320 | consumed tokens:    377487360 | elapsed time per iteration (s): 0.48 | learning rate: 2.143E-05 | global batch size:   256 | lm loss: 6.284398E+00 | grad norm: 0.187 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 537.915 | TFLOPs: 16.01 |
0:  iteration      730/     762 | consumed samples:       186880 | consumed tokens:    382730240 | elapsed time per iteration (s): 0.48 | learning rate: 2.084E-05 | global batch size:   256 | lm loss: 6.303773E+00 | grad norm: 0.209 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 538.086 | TFLOPs: 16.01 |
0:  iteration      740/     762 | consumed samples:       189440 | consumed tokens:    387973120 | elapsed time per iteration (s): 0.48 | learning rate: 2.041E-05 | global batch size:   256 | lm loss: 6.302274E+00 | grad norm: 0.254 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 537.945 | TFLOPs: 16.01 |
0:  iteration      750/     762 | consumed samples:       192000 | consumed tokens:    393216000 | elapsed time per iteration (s): 0.48 | learning rate: 2.013E-05 | global batch size:   256 | lm loss: 6.297002E+00 | grad norm: 0.174 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 537.779 | TFLOPs: 16.00 |
0:  iteration      760/     762 | consumed samples:       194560 | consumed tokens:    398458880 | elapsed time per iteration (s): 0.48 | learning rate: 2.001E-05 | global batch size:   256 | lm loss: 6.282113E+00 | grad norm: 0.193 | num zeros: 0.0 | number of skipped iterations:   0 | number of nan iterations:   0 | samples per second: 537.976 | TFLOPs: 16.01 |
0: [after training is done] datetime: 2023-04-27 15:45:27 
0: saving checkpoint at iteration     762 to checkpoints_14m400m100m
0: -----------------------------------------------------------------------------------------------------------------
0: validation loss at the end of training for val data | lm loss value: 6.262801E+00 | lm loss PPL: 5.246866E+02 | 
0: -----------------------------------------------------------------------------------------------------------------
0: [2023-04-27 15:45:27,500] [INFO] [logging.py:68:log_dist] [Rank 0] [Torch] Checkpoint global_step762 is begin to save!
0: [2023-04-27 15:45:27,620] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/layer_01-model_00-model_states.pt...
0: [2023-04-27 15:45:27,665] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/layer_01-model_00-model_states.pt.
0: [2023-04-27 15:45:27,665] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/layer_03-model_00-model_states.pt...
0: [2023-04-27 15:45:27,668] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/layer_03-model_00-model_states.pt.
0: [2023-04-27 15:45:27,668] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/layer_04-model_00-model_states.pt...
0: [2023-04-27 15:45:27,672] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/layer_04-model_00-model_states.pt.
0: [2023-04-27 15:45:27,672] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/layer_05-model_00-model_states.pt...
0: [2023-04-27 15:45:27,675] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/layer_05-model_00-model_states.pt.
0: [2023-04-27 15:45:27,675] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/layer_06-model_00-model_states.pt...
0: [2023-04-27 15:45:27,678] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/layer_06-model_00-model_states.pt.
0: [2023-04-27 15:45:27,678] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/layer_08-model_00-model_states.pt...
0: [2023-04-27 15:45:27,679] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/layer_08-model_00-model_states.pt.
0: [2023-04-27 15:45:27,679] [INFO] [logging.py:68:log_dist] [Rank 0] Saving model checkpoint: checkpoints_14m400m100m/global_step762/mp_rank_00_model_states.pt
0: [2023-04-27 15:45:27,679] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/mp_rank_00_model_states.pt...
0: [2023-04-27 15:45:27,681] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/mp_rank_00_model_states.pt.
0: [2023-04-27 15:45:27,685] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt...
0: [2023-04-27 15:45:27,685] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt...
0: [2023-04-27 15:45:27,685] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt...
0: [2023-04-27 15:45:27,685] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt...
0: [2023-04-27 15:45:27,685] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt...
0: [2023-04-27 15:45:27,685] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt...
0: [2023-04-27 15:45:27,685] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt...
0: [2023-04-27 15:45:27,685] [INFO] [torch_checkpoint_engine.py:15:save] [Torch] Saving checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt...
0: [2023-04-27 15:45:27,711] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt.
0: [2023-04-27 15:45:27,711] [INFO] [engine.py:3213:_save_zero_checkpoint] bf16_zero checkpoint saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt
0: [2023-04-27 15:45:27,711] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step762 is ready now!
0: [2023-04-27 15:45:27,714] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt.
0: [2023-04-27 15:45:27,714] [INFO] [engine.py:3213:_save_zero_checkpoint] bf16_zero checkpoint saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_5_mp_rank_00_optim_states.pt
0: [2023-04-27 15:45:27,714] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step762 is ready now!
0: [2023-04-27 15:45:27,714] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt.
0: [2023-04-27 15:45:27,714] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt.
0: [2023-04-27 15:45:27,714] [INFO] [engine.py:3213:_save_zero_checkpoint] bf16_zero checkpoint saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_7_mp_rank_00_optim_states.pt
0: [2023-04-27 15:45:27,714] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step762 is ready now!
0: [2023-04-27 15:45:27,714] [INFO] [engine.py:3213:_save_zero_checkpoint] bf16_zero checkpoint saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_6_mp_rank_00_optim_states.pt
0: [2023-04-27 15:45:27,714] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step762 is ready now!
0: [2023-04-27 15:45:27,731] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt.
0: [2023-04-27 15:45:27,731] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt.
0: [2023-04-27 15:45:27,731] [INFO] [engine.py:3213:_save_zero_checkpoint] bf16_zero checkpoint saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_4_mp_rank_00_optim_states.pt
0: [2023-04-27 15:45:27,731] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step762 is ready now!
0: [2023-04-27 15:45:27,731] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt.
0: [2023-04-27 15:45:27,731] [INFO] [torch_checkpoint_engine.py:17:save] [Torch] Saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt.
0: [2023-04-27 15:45:27,731] [INFO] [engine.py:3213:_save_zero_checkpoint] bf16_zero checkpoint saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt
0: [2023-04-27 15:45:27,731] [INFO] [engine.py:3213:_save_zero_checkpoint] bf16_zero checkpoint saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt
0: [2023-04-27 15:45:27,731] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step762 is ready now!
0: [2023-04-27 15:45:27,732] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step762 is ready now!
0: [2023-04-27 15:45:27,760] [INFO] [engine.py:3213:_save_zero_checkpoint] bf16_zero checkpoint saved checkpoints_14m400m100m/global_step762/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
0: [2023-04-27 15:45:27,760] [INFO] [torch_checkpoint_engine.py:27:commit] [Torch] Checkpoint global_step762 is ready now!
0:   successfully saved checkpoint at iteration     762 to checkpoints_14m400m100m
END 3423734: Thu 27 Apr 2023 03:45:34 PM EEST