Upload llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1
Browse files
llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/bench.slurm
ADDED
|
@@ -0,0 +1,111 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/bin/bash
|
| 2 |
+
|
| 3 |
+
#SBATCH --job-name=bench_cluster
|
| 4 |
+
#SBATCH --time=00:59:00
|
| 5 |
+
#SBATCH --partition=hopper-prod
|
| 6 |
+
#SBATCH --nodes=2
|
| 7 |
+
#SBATCH --gres=gpu:8
|
| 8 |
+
#SBATCH --qos=high
|
| 9 |
+
#SBATCH --ntasks-per-node=1
|
| 10 |
+
#SBATCH --cpus-per-task=96
|
| 11 |
+
#SBATCH --exclusive
|
| 12 |
+
#SBATCH --output=/fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/log.out
|
| 13 |
+
#SBATCH --error=/fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/log.out
|
| 14 |
+
|
| 15 |
+
# Function to update status based on squeue output
|
| 16 |
+
update_status() {
|
| 17 |
+
job_id=$1
|
| 18 |
+
status_file=$2
|
| 19 |
+
# For unknown reasons, it doenst update status for pending. It only works for running
|
| 20 |
+
while true; do
|
| 21 |
+
job_status=$(squeue --job $job_id --noheader --format=%T)
|
| 22 |
+
echo "Job status: $job_status"
|
| 23 |
+
if [ -z "$job_status" ]; then
|
| 24 |
+
# Job has finished or is not found
|
| 25 |
+
break
|
| 26 |
+
elif [ "$job_status" = "RUNNING" ]; then
|
| 27 |
+
printf "running" > $status_file
|
| 28 |
+
break
|
| 29 |
+
fi
|
| 30 |
+
sleep 10
|
| 31 |
+
done
|
| 32 |
+
}
|
| 33 |
+
|
| 34 |
+
# Misc initializations.
|
| 35 |
+
echo "========================"
|
| 36 |
+
echo "START TIME: $(date)"
|
| 37 |
+
source /fsx/ferdinandmom/miniforge3/etc/profile.d/conda.sh
|
| 38 |
+
conda activate /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster
|
| 39 |
+
echo python3 version = $(python3 --version)
|
| 40 |
+
echo "========================"
|
| 41 |
+
|
| 42 |
+
# Slurm stuff
|
| 43 |
+
export HOSTNAMES=$(scontrol show hostnames "$SLURM_JOB_NODELIST")
|
| 44 |
+
export MASTER_ADDR=$(scontrol show hostnames "$SLURM_JOB_NODELIST" | head -n 1)
|
| 45 |
+
export MASTER_PORT=$((1024 + RANDOM % 64511))
|
| 46 |
+
|
| 47 |
+
export TMPDIR=/scratch
|
| 48 |
+
export HF_DATASETS_CACHE="/admin/home/ferdinand_mom/.cache"
|
| 49 |
+
export CUBLAS_WORKSPACE_CONFIG=":4096:8"
|
| 50 |
+
export CUDA_DEVICE_MAX_CONNECTIONS="1"
|
| 51 |
+
|
| 52 |
+
huggingface-cli login --token $HUGGINGFACE_TOKEN
|
| 53 |
+
|
| 54 |
+
|
| 55 |
+
NANOTRON_REPO="/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron"
|
| 56 |
+
CMD="$NANOTRON_REPO/run_train.py --config-file /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/config.yaml"
|
| 57 |
+
|
| 58 |
+
LAUNCHER="torchrun \
|
| 59 |
+
--nproc_per_node 8 \
|
| 60 |
+
--nnodes 2 \
|
| 61 |
+
--rdzv_endpoint ${MASTER_ADDR}:${MASTER_PORT} \
|
| 62 |
+
--rdzv_backend c10d \
|
| 63 |
+
--max_restarts 0 \
|
| 64 |
+
--tee 3 \
|
| 65 |
+
--node_rank ${SLURM_PROCID}"
|
| 66 |
+
|
| 67 |
+
# Checkout the bench_cluster branch
|
| 68 |
+
cd $NANOTRON_REPO
|
| 69 |
+
git checkout bench_cluster
|
| 70 |
+
cd ..
|
| 71 |
+
# Get the current job ID
|
| 72 |
+
job_id=${SLURM_JOB_ID}
|
| 73 |
+
|
| 74 |
+
# Update status to "pending" or "running" in the background
|
| 75 |
+
update_status $job_id /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/status.txt &
|
| 76 |
+
|
| 77 |
+
# Run the main command
|
| 78 |
+
srun -u $LAUNCHER $CMD
|
| 79 |
+
exit_status=$?
|
| 80 |
+
|
| 81 |
+
# Update status based on the exit status of `srun`
|
| 82 |
+
if [ $exit_status -eq 0 ]; then
|
| 83 |
+
printf "completed" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/status.txt
|
| 84 |
+
else
|
| 85 |
+
if grep -q "OutOfMemoryError" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/log.out; then
|
| 86 |
+
printf "oom" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/status.txt
|
| 87 |
+
elif grep -q " CUDA error: an illegal memory access" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/log.out; then
|
| 88 |
+
printf "oom" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/status.txt
|
| 89 |
+
elif grep -q "Timeout at NCCL" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/log.out; then
|
| 90 |
+
printf "timeout" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/status.txt
|
| 91 |
+
else
|
| 92 |
+
printf "fail" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/status.txt
|
| 93 |
+
fi
|
| 94 |
+
fi
|
| 95 |
+
|
| 96 |
+
# Run the report script if the job completed successfully
|
| 97 |
+
if [ $exit_status -eq 0 ]; then
|
| 98 |
+
python /fsx/ferdinandmom/ferdinand-hf/bench_cluster/main.py report --inp_dir /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1 --is_logs
|
| 99 |
+
python /fsx/ferdinandmom/ferdinand-hf/bench_cluster/main.py report --inp_dir /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1 --is_profiler
|
| 100 |
+
fi
|
| 101 |
+
|
| 102 |
+
|
| 103 |
+
# Push to hub the folder using huggingface_cli
|
| 104 |
+
huggingface-cli upload nanotron/bench_cluster /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1 llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1 --commit-message "Upload llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1"
|
| 105 |
+
|
| 106 |
+
# Verify the upload
|
| 107 |
+
if [ $? -eq 0 ]; then
|
| 108 |
+
echo "Uploading to Huggingface Hub successful"
|
| 109 |
+
else
|
| 110 |
+
echo "Failed to upload to Huggingface Hub"
|
| 111 |
+
fi
|
llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/config.yaml
ADDED
|
@@ -0,0 +1,90 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
general:
|
| 2 |
+
project: bench_cluster
|
| 3 |
+
seed: 42
|
| 4 |
+
model:
|
| 5 |
+
ddp_bucket_cap_mb: 25
|
| 6 |
+
dtype: bfloat16
|
| 7 |
+
init_method:
|
| 8 |
+
std: 0.025
|
| 9 |
+
make_vocab_size_divisible_by: 1
|
| 10 |
+
model_config:
|
| 11 |
+
bos_token_id: 1
|
| 12 |
+
eos_token_id: 2
|
| 13 |
+
hidden_act: silu
|
| 14 |
+
hidden_size: 2048
|
| 15 |
+
initializer_range: 0.02
|
| 16 |
+
intermediate_size: 4096
|
| 17 |
+
is_llama_config: true
|
| 18 |
+
max_position_embeddings: 4096
|
| 19 |
+
num_attention_heads: 32
|
| 20 |
+
num_hidden_layers: 24
|
| 21 |
+
num_key_value_heads: 32
|
| 22 |
+
pad_token_id: null
|
| 23 |
+
pretraining_tp: 1
|
| 24 |
+
rms_norm_eps: 1.0e-05
|
| 25 |
+
rope_scaling: null
|
| 26 |
+
rope_theta: 10000.0
|
| 27 |
+
tie_word_embeddings: true
|
| 28 |
+
use_cache: true
|
| 29 |
+
vocab_size: 50257
|
| 30 |
+
optimizer:
|
| 31 |
+
accumulate_grad_in_fp32: true
|
| 32 |
+
clip_grad: 1.0
|
| 33 |
+
learning_rate_scheduler:
|
| 34 |
+
learning_rate: 0.0001
|
| 35 |
+
lr_decay_style: linear
|
| 36 |
+
lr_warmup_style: linear
|
| 37 |
+
lr_warmup_steps: 1
|
| 38 |
+
min_decay_lr: 1.0e-05
|
| 39 |
+
optimizer_factory:
|
| 40 |
+
adam_beta1: 0.9
|
| 41 |
+
adam_beta2: 0.95
|
| 42 |
+
adam_eps: 1.0e-08
|
| 43 |
+
name: adamW
|
| 44 |
+
torch_adam_is_fused: true
|
| 45 |
+
weight_decay: 0.01
|
| 46 |
+
zero_stage: 1
|
| 47 |
+
parallelism:
|
| 48 |
+
dp: 2
|
| 49 |
+
expert_parallel_size: 1
|
| 50 |
+
pp: 1
|
| 51 |
+
pp_engine: 1f1b
|
| 52 |
+
tp: 8
|
| 53 |
+
tp_linear_async_communication: false
|
| 54 |
+
tp_mode: REDUCE_SCATTER
|
| 55 |
+
profiler:
|
| 56 |
+
profiler_export_path: /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1
|
| 57 |
+
tokenizer:
|
| 58 |
+
tokenizer_max_length: null
|
| 59 |
+
tokenizer_name_or_path: openai-community/gpt2
|
| 60 |
+
tokenizer_revision: null
|
| 61 |
+
data_stages:
|
| 62 |
+
- name: Training Stage
|
| 63 |
+
start_training_step: 1
|
| 64 |
+
data:
|
| 65 |
+
dataset:
|
| 66 |
+
dataset_overwrite_cache: false
|
| 67 |
+
dataset_processing_num_proc_per_process: 64
|
| 68 |
+
hf_dataset_config_name: null
|
| 69 |
+
hf_dataset_or_datasets: roneneldan/TinyStories
|
| 70 |
+
hf_dataset_splits: train
|
| 71 |
+
text_column_name: text
|
| 72 |
+
num_loading_workers: 32
|
| 73 |
+
seed: 42
|
| 74 |
+
lighteval: null
|
| 75 |
+
tokens:
|
| 76 |
+
train_steps: 20
|
| 77 |
+
val_check_interval: -1
|
| 78 |
+
batch_accumulation_per_replica: 512
|
| 79 |
+
limit_test_batches: 0
|
| 80 |
+
limit_val_batches: 0
|
| 81 |
+
micro_batch_size: 1
|
| 82 |
+
sequence_length: 4096
|
| 83 |
+
logging:
|
| 84 |
+
iteration_step_info_interval: 1
|
| 85 |
+
log_level: info
|
| 86 |
+
log_level_replica: info
|
| 87 |
+
checkpoints:
|
| 88 |
+
checkpoint_interval: 100000
|
| 89 |
+
checkpoints_path: /dev/null
|
| 90 |
+
resume_checkpoint_path: null
|
llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/log.out
ADDED
|
@@ -0,0 +1,602 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
========================
|
| 2 |
+
START TIME: Tue Jul 2 19:18:38 UTC 2024
|
| 3 |
+
python3 version = Python 3.10.14
|
| 4 |
+
========================
|
| 5 |
+
The token has not been saved to the git credentials helper. Pass `add_to_git_credential=True` in this function directly or `--add-to-git-credential` if using via `huggingface-cli` if you want to set the git credential as well.
|
| 6 |
+
Token is valid (permission: write).
|
| 7 |
+
Your token has been saved to /admin/home/ferdinand_mom/.cache/huggingface/token
|
| 8 |
+
Login successful
|
| 9 |
+
Already on 'bench_cluster'
|
| 10 |
+
M examples/config_tiny_llama.py
|
| 11 |
+
M examples/config_tiny_llama.yaml
|
| 12 |
+
M examples/train_tiny_llama.sh
|
| 13 |
+
M src/nanotron/models/llama.py
|
| 14 |
+
M src/nanotron/trainer.py
|
| 15 |
+
Your branch is up to date with 'origin/bench_cluster'.
|
| 16 |
+
Job status: RUNNING
|
| 17 |
+
W0702 19:18:41.233000 140131373512512 torch/distributed/run.py:757]
|
| 18 |
+
W0702 19:18:41.233000 140131373512512 torch/distributed/run.py:757] *****************************************
|
| 19 |
+
W0702 19:18:41.233000 140131373512512 torch/distributed/run.py:757] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
|
| 20 |
+
W0702 19:18:41.233000 140131373512512 torch/distributed/run.py:757] *****************************************
|
| 21 |
+
W0702 19:18:41.235000 139898939537216 torch/distributed/run.py:757]
|
| 22 |
+
W0702 19:18:41.235000 139898939537216 torch/distributed/run.py:757] *****************************************
|
| 23 |
+
W0702 19:18:41.235000 139898939537216 torch/distributed/run.py:757] Setting OMP_NUM_THREADS environment variable for each process to be 1 in default, to avoid your system being overloaded, please further tune the variable for optimal performance in your application as needed.
|
| 24 |
+
W0702 19:18:41.235000 139898939537216 torch/distributed/run.py:757] *****************************************
|
| 25 |
+
[default0]:07/02/2024 19:18:58 [WARNING|DP=0|PP=0|TP=0|ip-26-0-161-178]: [Vocab Size Padding] Padded vocab (size: 50257) with 7 dummy tokens (new size: 50264)
|
| 26 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Config:
|
| 27 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Config(general=GeneralArgs(project='bench_cluster',
|
| 28 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: run='%date_%jobid',
|
| 29 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: seed=42,
|
| 30 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: step=None,
|
| 31 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: consumed_train_samples=None,
|
| 32 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: benchmark_csv_path=None,
|
| 33 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: ignore_sanity_checks=True),
|
| 34 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: parallelism=ParallelismArgs(dp=2,
|
| 35 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: pp=1,
|
| 36 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tp=8,
|
| 37 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: pp_engine=<nanotron.parallel.pipeline_parallel.engine.OneForwardOneBackwardPipelineEngine object at 0x7ff653cf8910>,
|
| 38 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tp_mode=<TensorParallelLinearMode.REDUCE_SCATTER: 2>,
|
| 39 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tp_linear_async_communication=False,
|
| 40 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: expert_parallel_size=1),
|
| 41 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: model=ModelArgs(model_config=LlamaConfig(bos_token_id=1,
|
| 42 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: eos_token_id=2,
|
| 43 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: hidden_act='silu',
|
| 44 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: hidden_size=2048,
|
| 45 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: initializer_range=0.02,
|
| 46 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: intermediate_size=4096,
|
| 47 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: is_llama_config=True,
|
| 48 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: max_position_embeddings=4096,
|
| 49 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: num_attention_heads=32,
|
| 50 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: num_hidden_layers=24,
|
| 51 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: num_key_value_heads=32,
|
| 52 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: pad_token_id=None,
|
| 53 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: pretraining_tp=1,
|
| 54 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: rms_norm_eps=1e-05,
|
| 55 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: rope_scaling=None,
|
| 56 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: rope_theta=10000.0,
|
| 57 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tie_word_embeddings=True,
|
| 58 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: use_cache=True,
|
| 59 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: vocab_size=50264),
|
| 60 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: init_method=RandomInit(std=0.025),
|
| 61 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: dtype=torch.bfloat16,
|
| 62 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: make_vocab_size_divisible_by=1,
|
| 63 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: ddp_bucket_cap_mb=25),
|
| 64 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tokenizer=TokenizerArgs(tokenizer_name_or_path='openai-community/gpt2',
|
| 65 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tokenizer_revision=None,
|
| 66 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tokenizer_max_length=None),
|
| 67 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: checkpoints=CheckpointsArgs(checkpoints_path=Path('/dev/null'),
|
| 68 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: checkpoint_interval=100000,
|
| 69 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: save_initial_state=False,
|
| 70 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: resume_checkpoint_path=None,
|
| 71 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: checkpoints_path_is_shared_file_system=False),
|
| 72 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: logging=LoggingArgs(log_level='info',
|
| 73 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: log_level_replica='info',
|
| 74 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: iteration_step_info_interval=1),
|
| 75 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tokens=TokensArgs(sequence_length=4096,
|
| 76 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: train_steps=20,
|
| 77 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: micro_batch_size=1,
|
| 78 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: batch_accumulation_per_replica=512,
|
| 79 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: val_check_interval=-1,
|
| 80 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: limit_val_batches=0,
|
| 81 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: limit_test_batches=0),
|
| 82 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: optimizer=OptimizerArgs(optimizer_factory=AdamWOptimizerArgs(adam_eps=1e-08,
|
| 83 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: adam_beta1=0.9,
|
| 84 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: adam_beta2=0.95,
|
| 85 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: torch_adam_is_fused=True,
|
| 86 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: name='adamW'),
|
| 87 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: zero_stage=1,
|
| 88 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: weight_decay=0.01,
|
| 89 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: clip_grad=1.0,
|
| 90 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: accumulate_grad_in_fp32=True,
|
| 91 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: learning_rate_scheduler=LRSchedulerArgs(learning_rate=0.0001,
|
| 92 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: lr_warmup_steps=1,
|
| 93 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: lr_warmup_style='linear',
|
| 94 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: lr_decay_style='linear',
|
| 95 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: lr_decay_steps=19,
|
| 96 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: lr_decay_starting_step=None,
|
| 97 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: min_decay_lr=1e-05)),
|
| 98 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: data_stages=[DatasetStageArgs(name='Training Stage',
|
| 99 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: start_training_step=1,
|
| 100 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: data=DataArgs(dataset=PretrainDatasetsArgs(hf_dataset_or_datasets='roneneldan/TinyStories',
|
| 101 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: hf_dataset_splits='train',
|
| 102 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: hf_dataset_config_name=None,
|
| 103 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: dataset_processing_num_proc_per_process=64,
|
| 104 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: dataset_overwrite_cache=False,
|
| 105 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: text_column_name='text'),
|
| 106 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: seed=42,
|
| 107 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: num_loading_workers=32))],
|
| 108 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: profiler=ProfilerArgs(profiler_export_path=Path('/fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1')),
|
| 109 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: lighteval=None)
|
| 110 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Model Config:
|
| 111 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: LlamaConfig(bos_token_id=1,
|
| 112 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: eos_token_id=2,
|
| 113 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: hidden_act='silu',
|
| 114 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: hidden_size=2048,
|
| 115 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: initializer_range=0.02,
|
| 116 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: intermediate_size=4096,
|
| 117 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: is_llama_config=True,
|
| 118 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: max_position_embeddings=4096,
|
| 119 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: num_attention_heads=32,
|
| 120 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: num_hidden_layers=24,
|
| 121 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: num_key_value_heads=32,
|
| 122 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: pad_token_id=None,
|
| 123 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: pretraining_tp=1,
|
| 124 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: rms_norm_eps=1e-05,
|
| 125 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: rope_scaling=None,
|
| 126 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: rope_theta=10000.0,
|
| 127 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: tie_word_embeddings=True,
|
| 128 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: use_cache=True,
|
| 129 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: vocab_size=50264)
|
| 130 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Building model..
|
| 131 |
+
[default0]:07/02/2024 19:18:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Setting PP block ranks...
|
| 132 |
+
[default4]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=4|ip-26-0-161-178]: Local number of parameters: 139M (264.73MiB)
|
| 133 |
+
[default0]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Total number of parameters: 1.11G (2117.88MiB)
|
| 134 |
+
[default0]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Local number of parameters: 139M (264.73MiB)
|
| 135 |
+
[default4]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=4|ip-26-0-161-178]: [After model building] Memory usage: 290.76MiB. Peak allocated: 317.33MiB Peak reserved: 324.00MiB
|
| 136 |
+
[default4]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=4|ip-26-0-161-178]: No checkpoint path provided.
|
| 137 |
+
[default0]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [After model building] Memory usage: 290.76MiB. Peak allocated: 317.33MiB Peak reserved: 324.00MiB
|
| 138 |
+
[default0]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: No checkpoint path provided.
|
| 139 |
+
[default0]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Parametrizing model parameters using StandardParametrizator
|
| 140 |
+
[default1]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=1|ip-26-0-161-178]: Local number of parameters: 139M (264.73MiB)
|
| 141 |
+
[default1]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=1|ip-26-0-161-178]: [After model building] Memory usage: 290.76MiB. Peak allocated: 317.33MiB Peak reserved: 324.00MiB
|
| 142 |
+
[default1]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=1|ip-26-0-161-178]: No checkpoint path provided.
|
| 143 |
+
[default3]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=3|ip-26-0-161-178]: Local number of parameters: 139M (264.73MiB)
|
| 144 |
+
[default3]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=3|ip-26-0-161-178]: [After model building] Memory usage: 290.76MiB. Peak allocated: 317.33MiB Peak reserved: 324.00MiB
|
| 145 |
+
[default3]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=3|ip-26-0-161-178]: No checkpoint path provided.
|
| 146 |
+
[default2]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=2|ip-26-0-161-178]: Local number of parameters: 139M (264.73MiB)
|
| 147 |
+
[default2]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=2|ip-26-0-161-178]: [After model building] Memory usage: 290.76MiB. Peak allocated: 317.33MiB Peak reserved: 324.00MiB
|
| 148 |
+
[default2]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=2|ip-26-0-161-178]: No checkpoint path provided.
|
| 149 |
+
[default7]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=7|ip-26-0-161-178]: Local number of parameters: 139M (264.73MiB)
|
| 150 |
+
[default7]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=7|ip-26-0-161-178]: [After model building] Memory usage: 290.76MiB. Peak allocated: 317.33MiB Peak reserved: 324.00MiB
|
| 151 |
+
[default7]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=7|ip-26-0-161-178]: No checkpoint path provided.
|
| 152 |
+
[default5]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=5|ip-26-0-161-178]: Local number of parameters: 139M (264.73MiB)
|
| 153 |
+
[default5]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=5|ip-26-0-161-178]: [After model building] Memory usage: 290.76MiB. Peak allocated: 317.33MiB Peak reserved: 324.00MiB
|
| 154 |
+
[default5]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=5|ip-26-0-161-178]: No checkpoint path provided.
|
| 155 |
+
[default6]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=6|ip-26-0-161-178]: Local number of parameters: 139M (264.73MiB)
|
| 156 |
+
[default6]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=6|ip-26-0-161-178]: [After model building] Memory usage: 290.76MiB. Peak allocated: 317.33MiB Peak reserved: 324.00MiB
|
| 157 |
+
[default6]:07/02/2024 19:19:14 [INFO|DP=0|PP=0|TP=6|ip-26-0-161-178]: No checkpoint path provided.
|
| 158 |
+
[default5]:07/02/2024 19:19:14 [INFO|DP=1|PP=0|TP=5|ip-26-0-162-233]: No checkpoint path provided.
|
| 159 |
+
[default1]:07/02/2024 19:19:14 [INFO|DP=1|PP=0|TP=1|ip-26-0-162-233]: No checkpoint path provided.
|
| 160 |
+
[default4]:07/02/2024 19:19:14 [INFO|DP=1|PP=0|TP=4|ip-26-0-162-233]: No checkpoint path provided.
|
| 161 |
+
[default3]:07/02/2024 19:19:14 [INFO|DP=1|PP=0|TP=3|ip-26-0-162-233]: No checkpoint path provided.
|
| 162 |
+
[default7]:07/02/2024 19:19:14 [INFO|DP=1|PP=0|TP=7|ip-26-0-162-233]: No checkpoint path provided.
|
| 163 |
+
[default0]:07/02/2024 19:19:14 [INFO|DP=1|PP=0|TP=0|ip-26-0-162-233]: No checkpoint path provided.
|
| 164 |
+
[default6]:07/02/2024 19:19:14 [INFO|DP=1|PP=0|TP=6|ip-26-0-162-233]: No checkpoint path provided.
|
| 165 |
+
[default2]:07/02/2024 19:19:14 [INFO|DP=1|PP=0|TP=2|ip-26-0-162-233]: No checkpoint path provided.
|
| 166 |
+
[default0]:07/02/2024 19:19:15 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [Optimizer Building] Using LearningRateForSP as learning rate
|
| 167 |
+
[default0]:07/02/2024 19:19:15 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [ZeRO sharding] Size of optimizer params per rank:
|
| 168 |
+
[default0]:07/02/2024 19:19:15 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [ZeRO sharding] DP Rank 0 has 69.4M out of 139M (50.00%) params' optimizer states
|
| 169 |
+
[default0]:07/02/2024 19:19:15 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [ZeRO sharding] DP Rank 1 has 69.4M out of 139M (50.00%) params' optimizer states
|
| 170 |
+
[default0]:07/02/2024 19:19:17 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [Training Plan] Stage Training Stage has 19 remaining training steps and has consumed 0 samples
|
| 171 |
+
[default0]:07/02/2024 19:19:17 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Using `datasets` library
|
| 172 |
+
[default0]:07/02/2024 19:19:17 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Loading tokenizer from openai-community/gpt2 and transformers/hf_hub versions ('4.41.2', '0.23.4')
|
| 173 |
+
[default0]:07/02/2024 19:19:17 [WARNING|DP=0|PP=0|TP=0|ip-26-0-161-178]: Repo card metadata block was not found. Setting CardData to empty.
|
| 174 |
+
[default0]:Repo card metadata block was not found. Setting CardData to empty.
|
| 175 |
+
[default0]:07/02/2024 19:19:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [Training Plan] There are 1 training stages
|
| 176 |
+
[default0]:07/02/2024 19:19:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [Stage Training Stage] start from step 1
|
| 177 |
+
[default0]:07/02/2024 19:19:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]:
|
| 178 |
+
[default0]:07/02/2024 19:19:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: [Start training] datetime: 2024-07-02 19:19:18.537077 | mbs: 1 | grad_accum: 512 | global_batch_size: 1024 | sequence_length: 4096 | train_steps: 20 | start_iteration_step: 0 | consumed_train_samples: 0
|
| 179 |
+
[default0]:07/02/2024 19:19:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Resuming training from stage Training Stage, it has trained for 0 samples and has 19 remaining train steps
|
| 180 |
+
[default0]:07/02/2024 19:19:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1085.49MiB. Peak allocated 1085.49MiB. Peak reserved: 1120.00MiB
|
| 181 |
+
[default2]:07/02/2024 19:19:18 [WARNING|DP=0|PP=0|TP=2|ip-26-0-161-178]: Repo card metadata block was not found. Setting CardData to empty.
|
| 182 |
+
[default4]:07/02/2024 19:19:18 [WARNING|DP=0|PP=0|TP=4|ip-26-0-161-178]: Repo card metadata block was not found. Setting CardData to empty.
|
| 183 |
+
[default1]:07/02/2024 19:19:18 [WARNING|DP=1|PP=0|TP=1|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 184 |
+
[default4]:07/02/2024 19:19:18 [WARNING|DP=1|PP=0|TP=4|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 185 |
+
[default3]:07/02/2024 19:19:18 [WARNING|DP=1|PP=0|TP=3|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 186 |
+
[default2]:Repo card metadata block was not found. Setting CardData to empty.
|
| 187 |
+
[default0]:07/02/2024 19:19:18 [WARNING|DP=1|PP=0|TP=0|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 188 |
+
[default1]:07/02/2024 19:19:18 [WARNING|DP=0|PP=0|TP=1|ip-26-0-161-178]: Repo card metadata block was not found. Setting CardData to empty.
|
| 189 |
+
[default0]:Repo card metadata block was not found. Setting CardData to empty.
|
| 190 |
+
[default5]:07/02/2024 19:19:18 [WARNING|DP=1|PP=0|TP=5|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 191 |
+
[default7]:07/02/2024 19:19:18 [WARNING|DP=0|PP=0|TP=7|ip-26-0-161-178]: Repo card metadata block was not found. Setting CardData to empty.
|
| 192 |
+
[default7]:07/02/2024 19:19:18 [WARNING|DP=1|PP=0|TP=7|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 193 |
+
[default3]:07/02/2024 19:19:18 [WARNING|DP=0|PP=0|TP=3|ip-26-0-161-178]: Repo card metadata block was not found. Setting CardData to empty.
|
| 194 |
+
[default3]:Repo card metadata block was not found. Setting CardData to empty.
|
| 195 |
+
[default6]:07/02/2024 19:19:18 [WARNING|DP=0|PP=0|TP=6|ip-26-0-161-178]: Repo card metadata block was not found. Setting CardData to empty.
|
| 196 |
+
[default1]:Repo card metadata block was not found. Setting CardData to empty.
|
| 197 |
+
[default5]:07/02/2024 19:19:18 [WARNING|DP=0|PP=0|TP=5|ip-26-0-161-178]: Repo card metadata block was not found. Setting CardData to empty.
|
| 198 |
+
[default4]:Repo card metadata block was not found. Setting CardData to empty.
|
| 199 |
+
[default5]:Repo card metadata block was not found. Setting CardData to empty.
|
| 200 |
+
[default1]:Repo card metadata block was not found. Setting CardData to empty.
|
| 201 |
+
[default2]:07/02/2024 19:19:18 [WARNING|DP=1|PP=0|TP=2|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 202 |
+
[default6]:07/02/2024 19:19:18 [WARNING|DP=1|PP=0|TP=6|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 203 |
+
[default7]:Repo card metadata block was not found. Setting CardData to empty.
|
| 204 |
+
[default6]:Repo card metadata block was not found. Setting CardData to empty.
|
| 205 |
+
[default3]:Repo card metadata block was not found. Setting CardData to empty.
|
| 206 |
+
[default6]:Repo card metadata block was not found. Setting CardData to empty.
|
| 207 |
+
[default2]:Repo card metadata block was not found. Setting CardData to empty.
|
| 208 |
+
[default7]:Repo card metadata block was not found. Setting CardData to empty.
|
| 209 |
+
[default5]:Repo card metadata block was not found. Setting CardData to empty.
|
| 210 |
+
[default4]:Repo card metadata block was not found. Setting CardData to empty.
|
| 211 |
+
[default2]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 212 |
+
[default2]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 213 |
+
[default3]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 214 |
+
[default3]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 215 |
+
[default5]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 216 |
+
[default5]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 217 |
+
[default1]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 218 |
+
[default1]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 219 |
+
[default7]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 220 |
+
[default7]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 221 |
+
[default6]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 222 |
+
[default6]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 223 |
+
[default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 224 |
+
[default0]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 225 |
+
[default4]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 226 |
+
[default4]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 227 |
+
[default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 228 |
+
[default0]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 229 |
+
[default4]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 230 |
+
[default4]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 231 |
+
[default1]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 232 |
+
[default1]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 233 |
+
[default3]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 234 |
+
[default3]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 235 |
+
[default7]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 236 |
+
[default7]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 237 |
+
[default2]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 238 |
+
[default2]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 239 |
+
[default6]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 240 |
+
[default6]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 241 |
+
[default5]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 242 |
+
[default5]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 243 |
+
[default0]:07/02/2024 19:20:34 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1161.81MiB. Peak allocated 2975.18MiB. Peak reserved: 3232.00MiB
|
| 244 |
+
[default2]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 245 |
+
[default2]: warnings.warn(
|
| 246 |
+
[default3]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 247 |
+
[default3]: warnings.warn(
|
| 248 |
+
[default6]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 249 |
+
[default6]: warnings.warn(
|
| 250 |
+
[default7]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 251 |
+
[default7]: warnings.warn(
|
| 252 |
+
[default1]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 253 |
+
[default1]: warnings.warn(
|
| 254 |
+
[default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 255 |
+
[default0]: warnings.warn(
|
| 256 |
+
[default5]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 257 |
+
[default5]: warnings.warn(
|
| 258 |
+
[default4]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 259 |
+
[default4]: warnings.warn(
|
| 260 |
+
[default5]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 261 |
+
[default5]: warnings.warn(
|
| 262 |
+
[default7]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 263 |
+
[default7]: warnings.warn(
|
| 264 |
+
[default4]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 265 |
+
[default4]: warnings.warn(
|
| 266 |
+
[default1]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 267 |
+
[default1]: warnings.warn(
|
| 268 |
+
[default3]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 269 |
+
[default3]: warnings.warn(
|
| 270 |
+
[default0]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 271 |
+
[default0]: warnings.warn(
|
| 272 |
+
[default6]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 273 |
+
[default6]: warnings.warn(
|
| 274 |
+
[default2]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/distributed_c10d.py:2261: UserWarning: torch.distributed.all_reduce_coalesced will be deprecated. If you must use it, please revisit our documentation later at https://pytorch.org/docs/master/distributed.html#collective-functions
|
| 275 |
+
[default2]: warnings.warn(
|
| 276 |
+
[default0]:07/02/2024 19:20:44 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: iteration: 1 / 20 | consumed_tokens: 4.19M | elapsed_time_per_iteration_ms: 85.7K | tokens_per_sec: 49K | tokens_per_sec_per_gpu: 3.06K | global_batch_size: 1.02K | lm_loss: 11.5 | lr: 0.0001 | model_tflops_per_gpu: 27.8 | hardware_tflops_per_gpu: 27.8 | grad_norm: 15.7 | cuda_memory_allocated: 1.78G | cuda_max_memory_reserved: 3.83G | hd_total_memory_tb: 312G | hd_used_memory_tb: 69.1G | hd_free_memory_tb: 243G
|
| 277 |
+
[default0]:07/02/2024 19:20:44 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1697.35MiB. Peak allocated 2359.19MiB. Peak reserved: 3656.00MiB
|
| 278 |
+
[default0]:07/02/2024 19:21:54 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1697.60MiB. Peak allocated 3450.67MiB. Peak reserved: 3680.00MiB
|
| 279 |
+
[default0]:07/02/2024 19:21:57 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: iteration: 2 / 20 | consumed_tokens: 8.39M | elapsed_time_per_iteration_ms: 72.8K | tokens_per_sec: 57.6K | tokens_per_sec_per_gpu: 3.6K | global_batch_size: 1.02K | lm_loss: 11.5 | lr: 9.53e-05 | model_tflops_per_gpu: 32.7 | hardware_tflops_per_gpu: 32.7 | grad_norm: 16 | cuda_memory_allocated: 1.78G | cuda_max_memory_reserved: 3.86G | hd_total_memory_tb: 312G | hd_used_memory_tb: 69.1G | hd_free_memory_tb: 243G
|
| 280 |
+
[default0]:07/02/2024 19:21:57 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1697.35MiB. Peak allocated 2359.44MiB. Peak reserved: 3680.00MiB
|
| 281 |
+
[default0]:07/02/2024 19:23:17 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1697.60MiB. Peak allocated 3450.67MiB. Peak reserved: 3680.00MiB
|
| 282 |
+
[default0]:STAGE:2024-07-02 19:23:18 396232:396232 ActivityProfilerController.cpp:314] Completed Stage: Warm Up
|
| 283 |
+
[default0]:07/02/2024 19:23:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: iteration: 3 / 20 | consumed_tokens: 12.6M | elapsed_time_per_iteration_ms: 81.4K | tokens_per_sec: 51.5K | tokens_per_sec_per_gpu: 3.22K | global_batch_size: 1.02K | lm_loss: 12.8 | lr: 9.05e-05 | model_tflops_per_gpu: 29.2 | hardware_tflops_per_gpu: 29.2 | grad_norm: 137 | cuda_memory_allocated: 1.78G | cuda_max_memory_reserved: 3.86G | hd_total_memory_tb: 312G | hd_used_memory_tb: 69.1G | hd_free_memory_tb: 243G
|
| 284 |
+
[default0]:07/02/2024 19:23:18 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1697.35MiB. Peak allocated 2359.44MiB. Peak reserved: 3680.00MiB
|
| 285 |
+
[default0]:07/02/2024 19:25:05 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1697.60MiB. Peak allocated 3450.67MiB. Peak reserved: 3680.00MiB
|
| 286 |
+
[default0]:07/02/2024 19:25:07 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: iteration: 4 / 20 | consumed_tokens: 16.8M | elapsed_time_per_iteration_ms: 109K | tokens_per_sec: 38.6K | tokens_per_sec_per_gpu: 2.42K | global_batch_size: 1.02K | lm_loss: 12.2 | lr: 8.58e-05 | model_tflops_per_gpu: 21.9 | hardware_tflops_per_gpu: 21.9 | grad_norm: 22.4 | cuda_memory_allocated: 1.78G | cuda_max_memory_reserved: 3.86G | hd_total_memory_tb: 312G | hd_used_memory_tb: 69.1G | hd_free_memory_tb: 243G
|
| 287 |
+
[default0]:07/02/2024 19:25:07 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1697.35MiB. Peak allocated 2359.44MiB. Peak reserved: 3680.00MiB
|
| 288 |
+
[default0]:07/02/2024 19:26:56 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: iteration: 5 / 20 | consumed_tokens: 21M | elapsed_time_per_iteration_ms: 109K | tokens_per_sec: 38.3K | tokens_per_sec_per_gpu: 2.4K | global_batch_size: 1.02K | lm_loss: 12.4 | lr: 8.11e-05 | model_tflops_per_gpu: 21.7 | hardware_tflops_per_gpu: 21.7 | grad_norm: 42.8
|
| 289 |
+
[default0]:07/02/2024 19:26:56 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: Memory usage: 1697.35MiB. Peak allocated 3450.67MiB. Peak reserved: 3680.00MiB
|
| 290 |
+
[default0]:07/02/2024 19:28:46 [INFO|DP=0|PP=0|TP=0|ip-26-0-161-178]: iteration: 6 / 20 | consumed_tokens: 25.2M | elapsed_time_per_iteration_ms: 110K | tokens_per_sec: 38.2K | tokens_per_sec_per_gpu: 2.39K | global_batch_size: 1.02K | lm_loss: 11.1 | lr: 7.63e-05 | model_tflops_per_gpu: 21.6 | hardware_tflops_per_gpu: 21.6 | grad_norm: 24.8
|
| 291 |
+
[default0]:STAGE:2024-07-02 19:33:58 396232:396232 ActivityProfilerController.cpp:320] Completed Stage: Collection
|
| 292 |
+
[default0]:STAGE:2024-07-02 19:34:51 396232:396232 ActivityProfilerController.cpp:324] Completed Stage: Post Processing
|
| 293 |
+
[default3]:[rank3]:[E ProcessGroupNCCL.cpp:563] [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600053 milliseconds before timing out.
|
| 294 |
+
[default2]:[rank2]:[E ProcessGroupNCCL.cpp:563] [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600050 milliseconds before timing out.
|
| 295 |
+
[default5]:[rank5]:[E ProcessGroupNCCL.cpp:563] [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600002 milliseconds before timing out.
|
| 296 |
+
[default1]:[rank1]:[E ProcessGroupNCCL.cpp:563] [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600047 milliseconds before timing out.
|
| 297 |
+
[default4]:[rank4]:[E ProcessGroupNCCL.cpp:563] [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600033 milliseconds before timing out.
|
| 298 |
+
[default6]:[rank6]:[E ProcessGroupNCCL.cpp:563] [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600061 milliseconds before timing out.
|
| 299 |
+
[default7]:[rank7]:[E ProcessGroupNCCL.cpp:563] [Rank 7] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600060 milliseconds before timing out.
|
| 300 |
+
[default4]:[rank4]:[E ProcessGroupNCCL.cpp:1537] [PG 2 Rank 4] Timeout at NCCL work: 611392, last enqueued NCCL work: 611514, last completed NCCL work: 611391.
|
| 301 |
+
[default4]:[rank4]:[E ProcessGroupNCCL.cpp:577] [Rank 4] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
|
| 302 |
+
[default4]:[rank4]:[E ProcessGroupNCCL.cpp:583] [Rank 4] To avoid data inconsistency, we are taking the entire process down.
|
| 303 |
+
[default4]:[rank4]:[E ProcessGroupNCCL.cpp:1414] [PG 2 Rank 4] Process group watchdog thread terminated with exception: [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600033 milliseconds before timing out.
|
| 304 |
+
[default4]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 305 |
+
[default4]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7fca0ab4f897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 306 |
+
[default4]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7fca0be28c62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 307 |
+
[default4]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7fca0be2da80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 308 |
+
[default4]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7fca0be2edcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 309 |
+
[default4]:frame #4: <unknown function> + 0xd3e95 (0x7fca578c7e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 310 |
+
[default4]:frame #5: <unknown function> + 0x8609 (0x7fca5c90e609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 311 |
+
[default4]:frame #6: clone + 0x43 (0x7fca5c6d9353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 312 |
+
[default4]:
|
| 313 |
+
[default4]:terminate called after throwing an instance of 'c10::DistBackendError'
|
| 314 |
+
[default4]: what(): [PG 2 Rank 4] Process group watchdog thread terminated with exception: [Rank 4] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600033 milliseconds before timing out.
|
| 315 |
+
[default4]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 316 |
+
[default4]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7fca0ab4f897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 317 |
+
[default4]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7fca0be28c62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 318 |
+
[default4]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7fca0be2da80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 319 |
+
[default4]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7fca0be2edcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 320 |
+
[default4]:frame #4: <unknown function> + 0xd3e95 (0x7fca578c7e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 321 |
+
[default4]:frame #5: <unknown function> + 0x8609 (0x7fca5c90e609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 322 |
+
[default4]:frame #6: clone + 0x43 (0x7fca5c6d9353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 323 |
+
[default4]:
|
| 324 |
+
[default4]:Exception raised from ncclCommWatchdog at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1418 (most recent call first):
|
| 325 |
+
[default4]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7fca0ab4f897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 326 |
+
[default4]:frame #1: <unknown function> + 0xe32119 (0x7fca0bab2119 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 327 |
+
[default4]:frame #2: <unknown function> + 0xd3e95 (0x7fca578c7e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 328 |
+
[default4]:frame #3: <unknown function> + 0x8609 (0x7fca5c90e609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 329 |
+
[default4]:frame #4: clone + 0x43 (0x7fca5c6d9353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 330 |
+
[default4]:
|
| 331 |
+
[default6]:[rank6]:[E ProcessGroupNCCL.cpp:1537] [PG 2 Rank 6] Timeout at NCCL work: 611392, last enqueued NCCL work: 611514, last completed NCCL work: 611391.
|
| 332 |
+
[default6]:[rank6]:[E ProcessGroupNCCL.cpp:577] [Rank 6] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
|
| 333 |
+
[default6]:[rank6]:[E ProcessGroupNCCL.cpp:583] [Rank 6] To avoid data inconsistency, we are taking the entire process down.
|
| 334 |
+
[default6]:[rank6]:[E ProcessGroupNCCL.cpp:1414] [PG 2 Rank 6] Process group watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600061 milliseconds before timing out.
|
| 335 |
+
[default6]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 336 |
+
[default6]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f88522b8897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 337 |
+
[default6]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7f8853591c62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 338 |
+
[default6]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7f8853596a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 339 |
+
[default6]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7f8853597dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 340 |
+
[default6]:frame #4: <unknown function> + 0xd3e95 (0x7f889f030e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 341 |
+
[default6]:frame #5: <unknown function> + 0x8609 (0x7f88a4077609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 342 |
+
[default6]:frame #6: clone + 0x43 (0x7f88a3e42353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 343 |
+
[default6]:
|
| 344 |
+
[default6]:terminate called after throwing an instance of 'c10::DistBackendError'
|
| 345 |
+
[default6]: what(): [PG 2 Rank 6] Process group watchdog thread terminated with exception: [Rank 6] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600061 milliseconds before timing out.
|
| 346 |
+
[default6]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 347 |
+
[default6]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f88522b8897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 348 |
+
[default6]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7f8853591c62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 349 |
+
[default6]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7f8853596a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 350 |
+
[default6]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7f8853597dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 351 |
+
[default6]:frame #4: <unknown function> + 0xd3e95 (0x7f889f030e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 352 |
+
[default6]:frame #5: <unknown function> + 0x8609 (0x7f88a4077609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 353 |
+
[default6]:frame #6: clone + 0x43 (0x7f88a3e42353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 354 |
+
[default6]:
|
| 355 |
+
[default6]:Exception raised from ncclCommWatchdog at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1418 (most recent call first):
|
| 356 |
+
[default6]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f88522b8897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 357 |
+
[default6]:frame #1: <unknown function> + 0xe32119 (0x7f885321b119 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 358 |
+
[default6]:frame #2: <unknown function> + 0xd3e95 (0x7f889f030e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 359 |
+
[default6]:frame #3: <unknown function> + 0x8609 (0x7f88a4077609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 360 |
+
[default6]:frame #4: clone + 0x43 (0x7f88a3e42353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 361 |
+
[default6]:
|
| 362 |
+
[default5]:[rank5]:[E ProcessGroupNCCL.cpp:1537] [PG 2 Rank 5] Timeout at NCCL work: 611392, last enqueued NCCL work: 611514, last completed NCCL work: 611391.
|
| 363 |
+
[default3]:[rank3]:[E ProcessGroupNCCL.cpp:1537] [PG 2 Rank 3] Timeout at NCCL work: 611392, last enqueued NCCL work: 611514, last completed NCCL work: 611391.
|
| 364 |
+
[default2]:[rank2]:[E ProcessGroupNCCL.cpp:1537] [PG 2 Rank 2] Timeout at NCCL work: 611392, last enqueued NCCL work: 611514, last completed NCCL work: 611391.
|
| 365 |
+
[default2]:[rank2]:[E ProcessGroupNCCL.cpp:577] [Rank 2] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
|
| 366 |
+
[default3]:[rank3]:[E ProcessGroupNCCL.cpp:577] [Rank 3] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
|
| 367 |
+
[default5]:[rank5]:[E ProcessGroupNCCL.cpp:577] [Rank 5] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
|
| 368 |
+
[default5]:[rank5]:[E ProcessGroupNCCL.cpp:583] [Rank 5] To avoid data inconsistency, we are taking the entire process down.
|
| 369 |
+
[default3]:[rank3]:[E ProcessGroupNCCL.cpp:583] [Rank 3] To avoid data inconsistency, we are taking the entire process down.
|
| 370 |
+
[default2]:[rank2]:[E ProcessGroupNCCL.cpp:583] [Rank 2] To avoid data inconsistency, we are taking the entire process down.
|
| 371 |
+
[default5]:[rank5]:[E ProcessGroupNCCL.cpp:1414] [PG 2 Rank 5] Process group watchdog thread terminated with exception: [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600002 milliseconds before timing out.
|
| 372 |
+
[default2]:[rank2]:[E ProcessGroupNCCL.cpp:1414] [PG 2 Rank 2] Process group watchdog thread terminated with exception: [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600050 milliseconds before timing out.
|
| 373 |
+
[default5]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 374 |
+
[default5]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f67d4555897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 375 |
+
[default2]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 376 |
+
[default5]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7f67d582ec62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 377 |
+
[default3]:[rank3]:[E ProcessGroupNCCL.cpp:1414] [PG 2 Rank 3] Process group watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600053 milliseconds before timing out.
|
| 378 |
+
[default3]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 379 |
+
[default5]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7f67d5833a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 380 |
+
[default2]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7fdeb9545897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 381 |
+
[default2]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7fdeba81ec62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 382 |
+
[default5]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7f67d5834dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 383 |
+
[default2]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7fdeba823a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 384 |
+
[default2]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7fdeba824dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 385 |
+
[default3]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f604a86a897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 386 |
+
[default2]:frame #4: <unknown function> + 0xd3e95 (0x7fdf062bde95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 387 |
+
[default5]:frame #4: <unknown function> + 0xd3e95 (0x7f68212cde95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 388 |
+
[default3]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7f604bb43c62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 389 |
+
[default2]:frame #5: <unknown function> + 0x8609 (0x7fdf0b304609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 390 |
+
[default5]:frame #5: <unknown function> + 0x8609 (0x7f6826314609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 391 |
+
[default3]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7f604bb48a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 392 |
+
[default2]:frame #6: clone + 0x43 (0x7fdf0b0cf353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 393 |
+
[default3]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7f604bb49dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 394 |
+
[default5]:frame #6: clone + 0x43 (0x7f68260df353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 395 |
+
[default3]:frame #4: <unknown function> + 0xd3e95 (0x7f60975e2e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 396 |
+
[default2]:
|
| 397 |
+
[default5]:
|
| 398 |
+
[default3]:frame #5: <unknown function> + 0x8609 (0x7f609c629609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 399 |
+
[default5]:terminate called after throwing an instance of 'c10::DistBackendError'
|
| 400 |
+
[default3]:frame #6: clone + 0x43 (0x7f609c3f4353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 401 |
+
[default2]:terminate called after throwing an instance of 'c10::DistBackendError'
|
| 402 |
+
[default3]:
|
| 403 |
+
[default5]: what(): [PG 2 Rank 5] Process group watchdog thread terminated with exception: [Rank 5] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600002 milliseconds before timing out.
|
| 404 |
+
[default3]:terminate called after throwing an instance of 'c10::DistBackendError'
|
| 405 |
+
[default2]: what(): [PG 2 Rank 2] Process group watchdog thread terminated with exception: [Rank 2] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600050 milliseconds before timing out.
|
| 406 |
+
[default5]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 407 |
+
[default3]: what(): [PG 2 Rank 3] Process group watchdog thread terminated with exception: [Rank 3] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600053 milliseconds before timing out.
|
| 408 |
+
[default2]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 409 |
+
[default5]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f67d4555897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 410 |
+
[default3]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 411 |
+
[default2]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7fdeb9545897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 412 |
+
[default5]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7f67d582ec62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 413 |
+
[default3]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f604a86a897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 414 |
+
[default2]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7fdeba81ec62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 415 |
+
[default5]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7f67d5833a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 416 |
+
[default3]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7f604bb43c62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 417 |
+
[default5]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7f67d5834dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 418 |
+
[default2]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7fdeba823a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 419 |
+
[default3]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7f604bb48a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 420 |
+
[default2]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7fdeba824dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 421 |
+
[default5]:frame #4: <unknown function> + 0xd3e95 (0x7f68212cde95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 422 |
+
[default2]:frame #4: <unknown function> + 0xd3e95 (0x7fdf062bde95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 423 |
+
[default5]:frame #5: <unknown function> + 0x8609 (0x7f6826314609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 424 |
+
[default5]:frame #6: clone + 0x43 (0x7f68260df353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 425 |
+
[default2]:frame #5: <unknown function> + 0x8609 (0x7fdf0b304609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 426 |
+
[default5]:
|
| 427 |
+
[default3]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7f604bb49dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 428 |
+
[default2]:frame #6: clone + 0x43 (0x7fdf0b0cf353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 429 |
+
[default3]:frame #4: <unknown function> + 0xd3e95 (0x7f60975e2e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 430 |
+
[default5]:Exception raised from ncclCommWatchdog at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1418 (most recent call first):
|
| 431 |
+
[default3]:frame #5: <unknown function> + 0x8609 (0x7f609c629609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 432 |
+
[default5]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f67d4555897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 433 |
+
[default3]:frame #6: clone + 0x43 (0x7f609c3f4353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 434 |
+
[default2]:
|
| 435 |
+
[default5]:frame #1: <unknown function> + 0xe32119 (0x7f67d54b8119 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 436 |
+
[default3]:
|
| 437 |
+
[default2]:Exception raised from ncclCommWatchdog at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1418 (most recent call first):
|
| 438 |
+
[default5]:frame #2: <unknown function> + 0xd3e95 (0x7f68212cde95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 439 |
+
[default5]:frame #3: <unknown function> + 0x8609 (0x7f6826314609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 440 |
+
[default2]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7fdeb9545897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 441 |
+
[default5]:frame #4: clone + 0x43 (0x7f68260df353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 442 |
+
[default3]:Exception raised from ncclCommWatchdog at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1418 (most recent call first):
|
| 443 |
+
[default3]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f604a86a897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 444 |
+
[default5]:
|
| 445 |
+
[default3]:frame #1: <unknown function> + 0xe32119 (0x7f604b7cd119 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 446 |
+
[default3]:frame #2: <unknown function> + 0xd3e95 (0x7f60975e2e95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 447 |
+
[default2]:frame #1: <unknown function> + 0xe32119 (0x7fdeba4a8119 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 448 |
+
[default2]:frame #2: <unknown function> + 0xd3e95 (0x7fdf062bde95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 449 |
+
[default2]:frame #3: <unknown function> + 0x8609 (0x7fdf0b304609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 450 |
+
[default2]:frame #4: clone + 0x43 (0x7fdf0b0cf353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 451 |
+
[default2]:
|
| 452 |
+
[default3]:frame #3: <unknown function> + 0x8609 (0x7f609c629609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 453 |
+
[default3]:frame #4: clone + 0x43 (0x7f609c3f4353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 454 |
+
[default3]:
|
| 455 |
+
[default1]:[rank1]:[E ProcessGroupNCCL.cpp:1537] [PG 2 Rank 1] Timeout at NCCL work: 611392, last enqueued NCCL work: 611514, last completed NCCL work: 611391.
|
| 456 |
+
[default1]:[rank1]:[E ProcessGroupNCCL.cpp:577] [Rank 1] Some NCCL operations have failed or timed out. Due to the asynchronous nature of CUDA kernels, subsequent GPU operations might run on corrupted/incomplete data.
|
| 457 |
+
[default1]:[rank1]:[E ProcessGroupNCCL.cpp:583] [Rank 1] To avoid data inconsistency, we are taking the entire process down.
|
| 458 |
+
[default1]:[rank1]:[E ProcessGroupNCCL.cpp:1414] [PG 2 Rank 1] Process group watchdog thread terminated with exception: [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600047 milliseconds before timing out.
|
| 459 |
+
[default1]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 460 |
+
[default1]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f42b1853897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 461 |
+
[default1]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7f42b2b2cc62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 462 |
+
[default1]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7f42b2b31a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 463 |
+
[default1]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7f42b2b32dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 464 |
+
[default1]:frame #4: <unknown function> + 0xd3e95 (0x7f42fe5cbe95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 465 |
+
[default1]:frame #5: <unknown function> + 0x8609 (0x7f4303612609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 466 |
+
[default1]:frame #6: clone + 0x43 (0x7f43033dd353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 467 |
+
[default1]:
|
| 468 |
+
[default1]:terminate called after throwing an instance of 'c10::DistBackendError'
|
| 469 |
+
[default1]: what(): [PG 2 Rank 1] Process group watchdog thread terminated with exception: [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=611392, OpType=_REDUCE_SCATTER_BASE, NumelIn=8388608, NumelOut=1048576, Timeout(ms)=600000) ran for 600047 milliseconds before timing out.
|
| 470 |
+
[default1]:Exception raised from checkTimeout at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:565 (most recent call first):
|
| 471 |
+
[default1]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f42b1853897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 472 |
+
[default1]:frame #1: c10d::ProcessGroupNCCL::WorkNCCL::checkTimeout(std::optional<std::chrono::duration<long, std::ratio<1l, 1000l> > >) + 0x1d2 (0x7f42b2b2cc62 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 473 |
+
[default1]:frame #2: c10d::ProcessGroupNCCL::watchdogHandler() + 0x1a0 (0x7f42b2b31a80 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 474 |
+
[default1]:frame #3: c10d::ProcessGroupNCCL::ncclCommWatchdog() + 0x10c (0x7f42b2b32dcc in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 475 |
+
[default1]:frame #4: <unknown function> + 0xd3e95 (0x7f42fe5cbe95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 476 |
+
[default1]:frame #5: <unknown function> + 0x8609 (0x7f4303612609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 477 |
+
[default1]:frame #6: clone + 0x43 (0x7f43033dd353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 478 |
+
[default1]:
|
| 479 |
+
[default1]:Exception raised from ncclCommWatchdog at ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1418 (most recent call first):
|
| 480 |
+
[default1]:frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x57 (0x7f42b1853897 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libc10.so)
|
| 481 |
+
[default1]:frame #1: <unknown function> + 0xe32119 (0x7f42b27b6119 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/lib/libtorch_cuda.so)
|
| 482 |
+
[default1]:frame #2: <unknown function> + 0xd3e95 (0x7f42fe5cbe95 in /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/../lib/libstdc++.so.6)
|
| 483 |
+
[default1]:frame #3: <unknown function> + 0x8609 (0x7f4303612609 in /lib/x86_64-linux-gnu/libpthread.so.0)
|
| 484 |
+
[default1]:frame #4: clone + 0x43 (0x7f43033dd353 in /lib/x86_64-linux-gnu/libc.so.6)
|
| 485 |
+
[default1]:
|
| 486 |
+
W0702 19:39:48.621000 140131373512512 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 396232 closing signal SIGTERM
|
| 487 |
+
W0702 19:39:48.624000 140131373512512 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 396239 closing signal SIGTERM
|
| 488 |
+
E0702 19:39:56.784000 140131373512512 torch/distributed/elastic/multiprocessing/api.py:826] failed (exitcode: -6) local_rank: 1 (pid: 396233) of binary: /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/python3.10
|
| 489 |
+
Traceback (most recent call last):
|
| 490 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/torchrun", line 8, in <module>
|
| 491 |
+
sys.exit(main())
|
| 492 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 347, in wrapper
|
| 493 |
+
return f(*args, **kwargs)
|
| 494 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 879, in main
|
| 495 |
+
run(args)
|
| 496 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 870, in run
|
| 497 |
+
elastic_launch(
|
| 498 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
|
| 499 |
+
return launch_agent(self._config, self._entrypoint, list(args))
|
| 500 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 263, in launch_agent
|
| 501 |
+
raise ChildFailedError(
|
| 502 |
+
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
|
| 503 |
+
============================================================
|
| 504 |
+
/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py FAILED
|
| 505 |
+
------------------------------------------------------------
|
| 506 |
+
Failures:
|
| 507 |
+
[1]:
|
| 508 |
+
time : 2024-07-02_19:39:48
|
| 509 |
+
host : ip-26-0-161-178.ec2.internal
|
| 510 |
+
rank : 2 (local_rank: 2)
|
| 511 |
+
exitcode : -6 (pid: 396234)
|
| 512 |
+
error_file: <N/A>
|
| 513 |
+
traceback : Signal 6 (SIGABRT) received by PID 396234
|
| 514 |
+
[2]:
|
| 515 |
+
time : 2024-07-02_19:39:48
|
| 516 |
+
host : ip-26-0-161-178.ec2.internal
|
| 517 |
+
rank : 3 (local_rank: 3)
|
| 518 |
+
exitcode : -6 (pid: 396235)
|
| 519 |
+
error_file: <N/A>
|
| 520 |
+
traceback : Signal 6 (SIGABRT) received by PID 396235
|
| 521 |
+
[3]:
|
| 522 |
+
time : 2024-07-02_19:39:48
|
| 523 |
+
host : ip-26-0-161-178.ec2.internal
|
| 524 |
+
rank : 4 (local_rank: 4)
|
| 525 |
+
exitcode : -6 (pid: 396236)
|
| 526 |
+
error_file: <N/A>
|
| 527 |
+
traceback : Signal 6 (SIGABRT) received by PID 396236
|
| 528 |
+
[4]:
|
| 529 |
+
time : 2024-07-02_19:39:48
|
| 530 |
+
host : ip-26-0-161-178.ec2.internal
|
| 531 |
+
rank : 5 (local_rank: 5)
|
| 532 |
+
exitcode : -6 (pid: 396237)
|
| 533 |
+
error_file: <N/A>
|
| 534 |
+
traceback : Signal 6 (SIGABRT) received by PID 396237
|
| 535 |
+
[5]:
|
| 536 |
+
time : 2024-07-02_19:39:48
|
| 537 |
+
host : ip-26-0-161-178.ec2.internal
|
| 538 |
+
rank : 6 (local_rank: 6)
|
| 539 |
+
exitcode : -6 (pid: 396238)
|
| 540 |
+
error_file: <N/A>
|
| 541 |
+
traceback : Signal 6 (SIGABRT) received by PID 396238
|
| 542 |
+
------------------------------------------------------------
|
| 543 |
+
Root Cause (first observed failure):
|
| 544 |
+
[0]:
|
| 545 |
+
time : 2024-07-02_19:39:48
|
| 546 |
+
host : ip-26-0-161-178.ec2.internal
|
| 547 |
+
rank : 1 (local_rank: 1)
|
| 548 |
+
exitcode : -6 (pid: 396233)
|
| 549 |
+
error_file: <N/A>
|
| 550 |
+
traceback : Signal 6 (SIGABRT) received by PID 396233
|
| 551 |
+
============================================================
|
| 552 |
+
W0702 19:39:57.348000 139893272717056 torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1252] The node 'ip-26-0-162-233.ec2.internal_1294232_0' has failed to send a keep-alive heartbeat to the rendezvous 'none' due to an error of type RendezvousConnectionError.
|
| 553 |
+
srun: error: ip-26-0-161-178: task 0: Exited with exit code 1
|
| 554 |
+
[default7]:[rank15]:[E ProcessGroupNCCL.cpp:563] [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=713280, OpType=COALESCED, NumelIn=2048, NumelOut=2048, Timeout(ms)=600000) ran for 600025 milliseconds before timing out.
|
| 555 |
+
[default0]:[rank8]:[E ProcessGroupNCCL.cpp:563] [Rank 1] Watchdog caught collective operation timeout: WorkNCCL(SeqNum=713280, OpType=COALESCED, NumelIn=2048, NumelOut=2048, Timeout(ms)=600000) ran for 600032 milliseconds before timing out.
|
| 556 |
+
W0702 19:39:58.717000 139898939537216 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1294301 closing signal SIGTERM
|
| 557 |
+
W0702 19:39:58.722000 139898939537216 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1294302 closing signal SIGTERM
|
| 558 |
+
W0702 19:39:58.729000 139898939537216 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1294303 closing signal SIGTERM
|
| 559 |
+
W0702 19:39:58.732000 139898939537216 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1294304 closing signal SIGTERM
|
| 560 |
+
W0702 19:39:58.777000 139898939537216 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1294305 closing signal SIGTERM
|
| 561 |
+
W0702 19:39:58.782000 139898939537216 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1294306 closing signal SIGTERM
|
| 562 |
+
W0702 19:39:58.788000 139898939537216 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1294307 closing signal SIGTERM
|
| 563 |
+
W0702 19:39:58.794000 139898939537216 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1294308 closing signal SIGTERM
|
| 564 |
+
W0702 19:40:01.803000 139898939537216 torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1203] The node 'ip-26-0-162-233.ec2.internal_1294232_0' has failed to shutdown the rendezvous 'none' due to an error of type RendezvousConnectionError.
|
| 565 |
+
W0702 19:40:01.819000 139898939537216 torch/distributed/elastic/rendezvous/dynamic_rendezvous.py:1203] The node 'ip-26-0-162-233.ec2.internal_1294232_0' has failed to shutdown the rendezvous 'none' due to an error of type RendezvousConnectionError.
|
| 566 |
+
Traceback (most recent call last):
|
| 567 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 113, in _call_store
|
| 568 |
+
return getattr(self._store, store_op)(*args, **kwargs)
|
| 569 |
+
torch.distributed.DistNetworkError: Broken pipe
|
| 570 |
+
|
| 571 |
+
The above exception was the direct cause of the following exception:
|
| 572 |
+
|
| 573 |
+
Traceback (most recent call last):
|
| 574 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/torchrun", line 8, in <module>
|
| 575 |
+
sys.exit(main())
|
| 576 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 347, in wrapper
|
| 577 |
+
return f(*args, **kwargs)
|
| 578 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 879, in main
|
| 579 |
+
run(args)
|
| 580 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 870, in run
|
| 581 |
+
elastic_launch(
|
| 582 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
|
| 583 |
+
return launch_agent(self._config, self._entrypoint, list(args))
|
| 584 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 254, in launch_agent
|
| 585 |
+
result = agent.run()
|
| 586 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/metrics/api.py", line 123, in wrapper
|
| 587 |
+
result = f(*args, **kwargs)
|
| 588 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/agent/server/api.py", line 733, in run
|
| 589 |
+
result = self._invoke_run(role)
|
| 590 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/agent/server/api.py", line 908, in _invoke_run
|
| 591 |
+
num_nodes_waiting = rdzv_handler.num_nodes_waiting()
|
| 592 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 1174, in num_nodes_waiting
|
| 593 |
+
self._state_holder.sync()
|
| 594 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/rendezvous/dynamic_rendezvous.py", line 419, in sync
|
| 595 |
+
get_response = self._backend.get_state()
|
| 596 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 73, in get_state
|
| 597 |
+
base64_state: bytes = self._call_store("get", self._key)
|
| 598 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/rendezvous/c10d_rendezvous_backend.py", line 115, in _call_store
|
| 599 |
+
raise RendezvousConnectionError(
|
| 600 |
+
torch.distributed.elastic.rendezvous.api.RendezvousConnectionError: The connection to the C10d store has failed. See inner exception for details.
|
| 601 |
+
srun: error: ip-26-0-162-233: task 1: Exited with exit code 1
|
| 602 |
+
Consider using `hf_transfer` for faster uploads. This solution comes with some limitations. See https://huggingface.co/docs/huggingface_hub/hf_transfer for more details.
|
llama-1B/16_GPUS/dp-2_tp-8_pp-1_mbz-1/status.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
timeout
|