Upload llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64
Browse files
llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/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-1_tp-16_pp-1_mbz-64/log.out
|
| 13 |
+
#SBATCH --error=/fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/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-1_tp-16_pp-1_mbz-64/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-1_tp-16_pp-1_mbz-64/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-1_tp-16_pp-1_mbz-64/status.txt
|
| 84 |
+
else
|
| 85 |
+
if grep -q "OutOfMemoryError" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/log.out; then
|
| 86 |
+
printf "oom" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/status.txt
|
| 87 |
+
elif grep -q " CUDA error: an illegal memory access" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/log.out; then
|
| 88 |
+
printf "oom" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/status.txt
|
| 89 |
+
elif grep -q "Timeout at NCCL" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/log.out; then
|
| 90 |
+
printf "timeout" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/status.txt
|
| 91 |
+
else
|
| 92 |
+
printf "fail" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64/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-1_tp-16_pp-1_mbz-64 --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-1_tp-16_pp-1_mbz-64 --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-1_tp-16_pp-1_mbz-64 llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64 --commit-message "Upload llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64"
|
| 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-1_tp-16_pp-1_mbz-64/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: 1
|
| 49 |
+
expert_parallel_size: 1
|
| 50 |
+
pp: 1
|
| 51 |
+
pp_engine: 1f1b
|
| 52 |
+
tp: 16
|
| 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-1_tp-16_pp-1_mbz-64
|
| 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: 16
|
| 79 |
+
limit_test_batches: 0
|
| 80 |
+
limit_val_batches: 0
|
| 81 |
+
micro_batch_size: 64
|
| 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-1_tp-16_pp-1_mbz-64/log.out
ADDED
|
@@ -0,0 +1,723 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
========================
|
| 2 |
+
START TIME: Tue Jul 2 16:32:11 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 16:32:13.712000 140209290786624 torch/distributed/run.py:757]
|
| 18 |
+
W0702 16:32:13.712000 140209290786624 torch/distributed/run.py:757] *****************************************
|
| 19 |
+
W0702 16:32:13.712000 140209290786624 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 16:32:13.712000 140209290786624 torch/distributed/run.py:757] *****************************************
|
| 21 |
+
W0702 16:32:13.732000 139646960318272 torch/distributed/run.py:757]
|
| 22 |
+
W0702 16:32:13.732000 139646960318272 torch/distributed/run.py:757] *****************************************
|
| 23 |
+
W0702 16:32:13.732000 139646960318272 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 16:32:13.732000 139646960318272 torch/distributed/run.py:757] *****************************************
|
| 25 |
+
[default0]:07/02/2024 16:32:32 [WARNING|DP=0|PP=0|TP=0|ip-26-0-160-192]: [Vocab Size Padding] Padded vocab (size: 50257) with 15 dummy tokens (new size: 50272)
|
| 26 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Config:
|
| 27 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Config(general=GeneralArgs(project='bench_cluster',
|
| 28 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: run='%date_%jobid',
|
| 29 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: seed=42,
|
| 30 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: step=None,
|
| 31 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: consumed_train_samples=None,
|
| 32 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: benchmark_csv_path=None,
|
| 33 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: ignore_sanity_checks=True),
|
| 34 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: parallelism=ParallelismArgs(dp=1,
|
| 35 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: pp=1,
|
| 36 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tp=16,
|
| 37 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: pp_engine=<nanotron.parallel.pipeline_parallel.engine.OneForwardOneBackwardPipelineEngine object at 0x7f87ab5f8910>,
|
| 38 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tp_mode=<TensorParallelLinearMode.REDUCE_SCATTER: 2>,
|
| 39 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tp_linear_async_communication=False,
|
| 40 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: expert_parallel_size=1),
|
| 41 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: model=ModelArgs(model_config=LlamaConfig(bos_token_id=1,
|
| 42 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: eos_token_id=2,
|
| 43 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: hidden_act='silu',
|
| 44 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: hidden_size=2048,
|
| 45 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: initializer_range=0.02,
|
| 46 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: intermediate_size=4096,
|
| 47 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: is_llama_config=True,
|
| 48 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: max_position_embeddings=4096,
|
| 49 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: num_attention_heads=32,
|
| 50 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: num_hidden_layers=24,
|
| 51 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: num_key_value_heads=32,
|
| 52 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: pad_token_id=None,
|
| 53 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: pretraining_tp=1,
|
| 54 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: rms_norm_eps=1e-05,
|
| 55 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: rope_scaling=None,
|
| 56 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: rope_theta=10000.0,
|
| 57 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tie_word_embeddings=True,
|
| 58 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: use_cache=True,
|
| 59 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: vocab_size=50272),
|
| 60 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: init_method=RandomInit(std=0.025),
|
| 61 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: dtype=torch.bfloat16,
|
| 62 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: make_vocab_size_divisible_by=1,
|
| 63 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: ddp_bucket_cap_mb=25),
|
| 64 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tokenizer=TokenizerArgs(tokenizer_name_or_path='openai-community/gpt2',
|
| 65 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tokenizer_revision=None,
|
| 66 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tokenizer_max_length=None),
|
| 67 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: checkpoints=CheckpointsArgs(checkpoints_path=Path('/dev/null'),
|
| 68 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: checkpoint_interval=100000,
|
| 69 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: save_initial_state=False,
|
| 70 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: resume_checkpoint_path=None,
|
| 71 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: checkpoints_path_is_shared_file_system=False),
|
| 72 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: logging=LoggingArgs(log_level='info',
|
| 73 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: log_level_replica='info',
|
| 74 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: iteration_step_info_interval=1),
|
| 75 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tokens=TokensArgs(sequence_length=4096,
|
| 76 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: train_steps=20,
|
| 77 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: micro_batch_size=64,
|
| 78 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: batch_accumulation_per_replica=16,
|
| 79 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: val_check_interval=-1,
|
| 80 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: limit_val_batches=0,
|
| 81 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: limit_test_batches=0),
|
| 82 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: optimizer=OptimizerArgs(optimizer_factory=AdamWOptimizerArgs(adam_eps=1e-08,
|
| 83 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: adam_beta1=0.9,
|
| 84 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: adam_beta2=0.95,
|
| 85 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: torch_adam_is_fused=True,
|
| 86 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: name='adamW'),
|
| 87 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: zero_stage=1,
|
| 88 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: weight_decay=0.01,
|
| 89 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: clip_grad=1.0,
|
| 90 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: accumulate_grad_in_fp32=True,
|
| 91 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: learning_rate_scheduler=LRSchedulerArgs(learning_rate=0.0001,
|
| 92 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: lr_warmup_steps=1,
|
| 93 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: lr_warmup_style='linear',
|
| 94 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: lr_decay_style='linear',
|
| 95 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: lr_decay_steps=19,
|
| 96 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: lr_decay_starting_step=None,
|
| 97 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: min_decay_lr=1e-05)),
|
| 98 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: data_stages=[DatasetStageArgs(name='Training Stage',
|
| 99 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: start_training_step=1,
|
| 100 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: data=DataArgs(dataset=PretrainDatasetsArgs(hf_dataset_or_datasets='roneneldan/TinyStories',
|
| 101 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: hf_dataset_splits='train',
|
| 102 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: hf_dataset_config_name=None,
|
| 103 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: dataset_processing_num_proc_per_process=64,
|
| 104 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: dataset_overwrite_cache=False,
|
| 105 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: text_column_name='text'),
|
| 106 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: seed=42,
|
| 107 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: num_loading_workers=32))],
|
| 108 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: profiler=ProfilerArgs(profiler_export_path=Path('/fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-1_tp-16_pp-1_mbz-64')),
|
| 109 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: lighteval=None)
|
| 110 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Model Config:
|
| 111 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: LlamaConfig(bos_token_id=1,
|
| 112 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: eos_token_id=2,
|
| 113 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: hidden_act='silu',
|
| 114 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: hidden_size=2048,
|
| 115 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: initializer_range=0.02,
|
| 116 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: intermediate_size=4096,
|
| 117 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: is_llama_config=True,
|
| 118 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: max_position_embeddings=4096,
|
| 119 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: num_attention_heads=32,
|
| 120 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: num_hidden_layers=24,
|
| 121 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: num_key_value_heads=32,
|
| 122 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: pad_token_id=None,
|
| 123 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: pretraining_tp=1,
|
| 124 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: rms_norm_eps=1e-05,
|
| 125 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: rope_scaling=None,
|
| 126 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: rope_theta=10000.0,
|
| 127 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: tie_word_embeddings=True,
|
| 128 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: use_cache=True,
|
| 129 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: vocab_size=50272)
|
| 130 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Building model..
|
| 131 |
+
[default0]:07/02/2024 16:32:32 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Setting PP block ranks...
|
| 132 |
+
[default3]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=3|ip-26-0-160-192]: Local number of parameters: 69.4M (132.46MiB)
|
| 133 |
+
[default3]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=3|ip-26-0-160-192]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 134 |
+
[default3]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=3|ip-26-0-160-192]: No checkpoint path provided.
|
| 135 |
+
[default7]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=7|ip-26-0-160-192]: Local number of parameters: 69.4M (132.46MiB)
|
| 136 |
+
[default7]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=7|ip-26-0-160-192]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 137 |
+
[default7]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=7|ip-26-0-160-192]: No checkpoint path provided.
|
| 138 |
+
[default5]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=5|ip-26-0-160-192]: Local number of parameters: 69.4M (132.46MiB)
|
| 139 |
+
[default5]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=5|ip-26-0-160-192]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 140 |
+
[default5]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=5|ip-26-0-160-192]: No checkpoint path provided.
|
| 141 |
+
[default3]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=11|ip-26-0-162-233]: Local number of parameters: 69.4M (132.46MiB)
|
| 142 |
+
[default4]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=12|ip-26-0-162-233]: Local number of parameters: 69.4M (132.46MiB)
|
| 143 |
+
[default4]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=12|ip-26-0-162-233]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 144 |
+
[default3]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=11|ip-26-0-162-233]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 145 |
+
[default3]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=11|ip-26-0-162-233]: No checkpoint path provided.
|
| 146 |
+
[default4]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=12|ip-26-0-162-233]: No checkpoint path provided.
|
| 147 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Total number of parameters: 1.11G (2119.44MiB)
|
| 148 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Local number of parameters: 69.4M (132.46MiB)
|
| 149 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 150 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: No checkpoint path provided.
|
| 151 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Parametrizing model parameters using StandardParametrizator
|
| 152 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: [Optimizer Building] Using LearningRateForSP as learning rate
|
| 153 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: [ZeRO sharding] Size of optimizer params per rank:
|
| 154 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: [ZeRO sharding] DP Rank 0 has 69.4M out of 69.4M (100.00%) params' optimizer states
|
| 155 |
+
[default2]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=10|ip-26-0-162-233]: Local number of parameters: 69.4M (132.46MiB)
|
| 156 |
+
[default2]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=10|ip-26-0-162-233]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 157 |
+
[default2]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=10|ip-26-0-162-233]: No checkpoint path provided.
|
| 158 |
+
[default7]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=15|ip-26-0-162-233]: Local number of parameters: 69.4M (132.46MiB)
|
| 159 |
+
[default7]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=15|ip-26-0-162-233]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 160 |
+
[default7]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=15|ip-26-0-162-233]: No checkpoint path provided.
|
| 161 |
+
[default1]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=9|ip-26-0-162-233]: Local number of parameters: 69.4M (132.46MiB)
|
| 162 |
+
[default1]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=9|ip-26-0-162-233]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 163 |
+
[default1]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=9|ip-26-0-162-233]: No checkpoint path provided.
|
| 164 |
+
[default2]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=2|ip-26-0-160-192]: Local number of parameters: 69.4M (132.46MiB)
|
| 165 |
+
[default1]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=1|ip-26-0-160-192]: Local number of parameters: 69.4M (132.46MiB)
|
| 166 |
+
[default1]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=1|ip-26-0-160-192]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 167 |
+
[default6]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=6|ip-26-0-160-192]: Local number of parameters: 69.4M (132.46MiB)
|
| 168 |
+
[default6]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=6|ip-26-0-160-192]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 169 |
+
[default6]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=6|ip-26-0-160-192]: No checkpoint path provided.
|
| 170 |
+
[default4]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=4|ip-26-0-160-192]: Local number of parameters: 69.4M (132.46MiB)
|
| 171 |
+
[default4]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=4|ip-26-0-160-192]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 172 |
+
[default1]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=1|ip-26-0-160-192]: No checkpoint path provided.
|
| 173 |
+
[default4]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=4|ip-26-0-160-192]: No checkpoint path provided.
|
| 174 |
+
[default2]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=2|ip-26-0-160-192]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 175 |
+
[default2]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=2|ip-26-0-160-192]: No checkpoint path provided.
|
| 176 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=8|ip-26-0-162-233]: Local number of parameters: 69.4M (132.46MiB)
|
| 177 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=8|ip-26-0-162-233]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 178 |
+
[default0]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=8|ip-26-0-162-233]: No checkpoint path provided.
|
| 179 |
+
[default5]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=13|ip-26-0-162-233]: Local number of parameters: 69.4M (132.46MiB)
|
| 180 |
+
[default5]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=13|ip-26-0-162-233]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 181 |
+
[default6]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=14|ip-26-0-162-233]: Local number of parameters: 69.4M (132.46MiB)
|
| 182 |
+
[default5]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=13|ip-26-0-162-233]: No checkpoint path provided.
|
| 183 |
+
[default6]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=14|ip-26-0-162-233]: [After model building] Memory usage: 159.71MiB. Peak allocated: 174.02MiB Peak reserved: 178.00MiB
|
| 184 |
+
[default6]:07/02/2024 16:32:49 [INFO|DP=0|PP=0|TP=14|ip-26-0-162-233]: No checkpoint path provided.
|
| 185 |
+
[default0]:07/02/2024 16:32:50 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: [Training Plan] Stage Training Stage has 19 remaining training steps and has consumed 0 samples
|
| 186 |
+
[default0]:07/02/2024 16:32:50 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Using `datasets` library
|
| 187 |
+
[default0]:07/02/2024 16:32:50 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Loading tokenizer from openai-community/gpt2 and transformers/hf_hub versions ('4.41.2', '0.23.4')
|
| 188 |
+
[default0]:07/02/2024 16:32:50 [WARNING|DP=0|PP=0|TP=0|ip-26-0-160-192]: 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 |
+
[default0]:07/02/2024 16:32:51 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: [Training Plan] There are 1 training stages
|
| 191 |
+
[default0]:07/02/2024 16:32:51 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: [Stage Training Stage] start from step 1
|
| 192 |
+
[default0]:07/02/2024 16:32:51 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]:
|
| 193 |
+
[default0]:07/02/2024 16:32:51 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: [Start training] datetime: 2024-07-02 16:32:51.517913 | mbs: 64 | grad_accum: 16 | global_batch_size: 1024 | sequence_length: 4096 | train_steps: 20 | start_iteration_step: 0 | consumed_train_samples: 0
|
| 194 |
+
[default0]:07/02/2024 16:32:51 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Resuming training from stage Training Stage, it has trained for 0 samples and has 19 remaining train steps
|
| 195 |
+
[default0]:07/02/2024 16:32:51 [INFO|DP=0|PP=0|TP=0|ip-26-0-160-192]: Memory usage: 689.57MiB. Peak allocated 689.57MiB. Peak reserved: 710.00MiB
|
| 196 |
+
[default5]:Repo card metadata block was not found. Setting CardData to empty.
|
| 197 |
+
[default2]:Repo card metadata block was not found. Setting CardData to empty.
|
| 198 |
+
[default5]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=5|ip-26-0-160-192]: Repo card metadata block was not found. Setting CardData to empty.
|
| 199 |
+
[default0]:Repo card metadata block was not found. Setting CardData to empty.
|
| 200 |
+
[default3]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=11|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 201 |
+
[default4]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=12|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 202 |
+
[default4]:Repo card metadata block was not found. Setting CardData to empty.
|
| 203 |
+
[default2]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=10|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 204 |
+
[default1]:Repo card metadata block was not found. Setting CardData to empty.
|
| 205 |
+
[default7]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=15|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 206 |
+
[default1]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=9|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 207 |
+
[default1]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=1|ip-26-0-160-192]: Repo card metadata block was not found. Setting CardData to empty.
|
| 208 |
+
[default2]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=2|ip-26-0-160-192]: Repo card metadata block was not found. Setting CardData to empty.
|
| 209 |
+
[default0]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=8|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 210 |
+
[default3]:Repo card metadata block was not found. Setting CardData to empty.
|
| 211 |
+
[default5]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=13|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 212 |
+
[default7]:Repo card metadata block was not found. Setting CardData to empty.
|
| 213 |
+
[default2]:Repo card metadata block was not found. Setting CardData to empty.
|
| 214 |
+
[default1]:Repo card metadata block was not found. Setting CardData to empty.
|
| 215 |
+
[default5]:Repo card metadata block was not found. Setting CardData to empty.
|
| 216 |
+
[default6]:Repo card metadata block was not found. Setting CardData to empty.
|
| 217 |
+
[default4]:Repo card metadata block was not found. Setting CardData to empty.
|
| 218 |
+
[default7]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=7|ip-26-0-160-192]: Repo card metadata block was not found. Setting CardData to empty.
|
| 219 |
+
[default3]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=3|ip-26-0-160-192]: Repo card metadata block was not found. Setting CardData to empty.
|
| 220 |
+
[default7]:Repo card metadata block was not found. Setting CardData to empty.
|
| 221 |
+
[default6]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=6|ip-26-0-160-192]: Repo card metadata block was not found. Setting CardData to empty.
|
| 222 |
+
[default4]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=4|ip-26-0-160-192]: Repo card metadata block was not found. Setting CardData to empty.
|
| 223 |
+
[default6]:07/02/2024 16:32:51 [WARNING|DP=0|PP=0|TP=14|ip-26-0-162-233]: Repo card metadata block was not found. Setting CardData to empty.
|
| 224 |
+
[default6]:Repo card metadata block was not found. Setting CardData to empty.
|
| 225 |
+
[default3]:Repo card metadata block was not found. Setting CardData to empty.
|
| 226 |
+
[default7]:[rank7]: OSError: [Errno 122] Disk quota exceeded
|
| 227 |
+
[default7]:
|
| 228 |
+
[default7]:[rank7]: During handling of the above exception, another exception occurred:
|
| 229 |
+
[default7]:
|
| 230 |
+
[default7]:[rank7]: Traceback (most recent call last):
|
| 231 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 232 |
+
[default7]:[rank7]: trainer.train(dataloader)
|
| 233 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 234 |
+
[default7]:[rank7]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 235 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 236 |
+
[default7]:[rank7]: outputs = self.pipeline_engine.train_batch_iter(
|
| 237 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 238 |
+
[default7]:[rank7]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 239 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 240 |
+
[default7]:[rank7]: output = model(**micro_batch)
|
| 241 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 242 |
+
[default7]:[rank7]: return self._call_impl(*args, **kwargs)
|
| 243 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 244 |
+
[default7]:[rank7]: return forward_call(*args, **kwargs)
|
| 245 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 246 |
+
[default7]:[rank7]: sharded_logits = self.model(
|
| 247 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 248 |
+
[default7]:[rank7]: return self._call_impl(*args, **kwargs)
|
| 249 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 250 |
+
[default7]:[rank7]: return forward_call(*args, **kwargs)
|
| 251 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 252 |
+
[default7]:[rank7]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 253 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 254 |
+
[default7]:[rank7]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 255 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 256 |
+
[default7]:[rank7]: return self._call_impl(*args, **kwargs)
|
| 257 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 258 |
+
[default7]:[rank7]: return forward_call(*args, **kwargs)
|
| 259 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 260 |
+
[default7]:[rank7]: output = self.pp_block(**new_kwargs)
|
| 261 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 262 |
+
[default7]:[rank7]: return self._call_impl(*args, **kwargs)
|
| 263 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 264 |
+
[default7]:[rank7]: return forward_call(*args, **kwargs)
|
| 265 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 629, in forward
|
| 266 |
+
[default7]:[rank7]: hidden_states = self.input_layernorm(hidden_states)
|
| 267 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 268 |
+
[default7]:[rank7]: return self._call_impl(*args, **kwargs)
|
| 269 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 270 |
+
[default7]:[rank7]: return forward_call(*args, **kwargs)
|
| 271 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/nn/layer_norm.py", line 42, in forward
|
| 272 |
+
[default7]:[rank7]: return layer_norm_fn(
|
| 273 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 875, in layer_norm_fn
|
| 274 |
+
[default7]:[rank7]: return LayerNormFn.apply(
|
| 275 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 276 |
+
[default7]:[rank7]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 277 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 748, in forward
|
| 278 |
+
[default7]:[rank7]: y, y1, mean, rstd, residual_out, seeds, dropout_mask, dropout_mask1 = _layer_norm_fwd(
|
| 279 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 335, in _layer_norm_fwd
|
| 280 |
+
[default7]:[rank7]: _layer_norm_fwd_1pass_kernel[(M,)](
|
| 281 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 282 |
+
[default7]:[rank7]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 283 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in run
|
| 284 |
+
[default7]:[rank7]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 285 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in <dictcomp>
|
| 286 |
+
[default7]:[rank7]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 287 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 122, in _bench
|
| 288 |
+
[default7]:[rank7]: return do_bench(kernel_call, warmup=self.warmup, rep=self.rep, quantiles=(0.5, 0.2, 0.8))
|
| 289 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/testing.py", line 102, in do_bench
|
| 290 |
+
[default7]:[rank7]: fn()
|
| 291 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 110, in kernel_call
|
| 292 |
+
[default7]:[rank7]: self.fn.run(
|
| 293 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 294 |
+
[default7]:[rank7]: return self.fn.run(*args, **kwargs)
|
| 295 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 296 |
+
[default7]:[rank7]: return self.fn.run(*args, **kwargs)
|
| 297 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 298 |
+
[default7]:[rank7]: return self.fn.run(*args, **kwargs)
|
| 299 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 300 |
+
[default7]:[rank7]: self.cache[device][key] = compile(
|
| 301 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 302 |
+
[default7]:[rank7]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 303 |
+
[default7]:[rank7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 304 |
+
[default7]:[rank7]: with open(temp_path, mode) as f:
|
| 305 |
+
[default7]:[rank7]: OSError: [Errno 122] Disk quota exceeded
|
| 306 |
+
[default7]:Exception in thread Thread-2 (_pin_memory_loop):
|
| 307 |
+
[default7]:Traceback (most recent call last):
|
| 308 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
|
| 309 |
+
[default7]: self.run()
|
| 310 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/threading.py", line 953, in run
|
| 311 |
+
[default7]: self._target(*self._args, **self._kwargs)
|
| 312 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/utils/data/_utils/pin_memory.py", line 54, in _pin_memory_loop
|
| 313 |
+
[default7]: do_one_step()
|
| 314 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/utils/data/_utils/pin_memory.py", line 31, in do_one_step
|
| 315 |
+
[default7]: r = in_queue.get(timeout=MP_STATUS_CHECK_INTERVAL)
|
| 316 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/queues.py", line 122, in get
|
| 317 |
+
[default7]: return _ForkingPickler.loads(res)
|
| 318 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/multiprocessing/reductions.py", line 495, in rebuild_storage_fd
|
| 319 |
+
[default7]: fd = df.detach()
|
| 320 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/resource_sharer.py", line 57, in detach
|
| 321 |
+
[default7]: with _resource_sharer.get_connection(self._id) as conn:
|
| 322 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/resource_sharer.py", line 86, in get_connection
|
| 323 |
+
[default7]: c = Client(address, authkey=process.current_process().authkey)
|
| 324 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 508, in Client
|
| 325 |
+
[default7]: answer_challenge(c, authkey)
|
| 326 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 752, in answer_challenge
|
| 327 |
+
[default7]: message = connection.recv_bytes(256) # reject large message
|
| 328 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 216, in recv_bytes
|
| 329 |
+
[default7]: buf = self._recv_bytes(maxlength)
|
| 330 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 414, in _recv_bytes
|
| 331 |
+
[default7]: buf = self._recv(4)
|
| 332 |
+
[default7]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 379, in _recv
|
| 333 |
+
[default7]: chunk = read(handle, remaining)
|
| 334 |
+
[default7]:ConnectionResetError: [Errno 104] Connection reset by peer
|
| 335 |
+
[default4]:[rank4]: OSError: [Errno 122] Disk quota exceeded
|
| 336 |
+
[default4]:
|
| 337 |
+
[default4]:[rank4]: During handling of the above exception, another exception occurred:
|
| 338 |
+
[default4]:
|
| 339 |
+
[default4]:[rank4]: Traceback (most recent call last):
|
| 340 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 341 |
+
[default4]:[rank4]: trainer.train(dataloader)
|
| 342 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 343 |
+
[default4]:[rank4]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 344 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 345 |
+
[default4]:[rank4]: outputs = self.pipeline_engine.train_batch_iter(
|
| 346 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 347 |
+
[default4]:[rank4]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 348 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 349 |
+
[default4]:[rank4]: output = model(**micro_batch)
|
| 350 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 351 |
+
[default4]:[rank4]: return self._call_impl(*args, **kwargs)
|
| 352 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 353 |
+
[default4]:[rank4]: return forward_call(*args, **kwargs)
|
| 354 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 355 |
+
[default4]:[rank4]: sharded_logits = self.model(
|
| 356 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 357 |
+
[default4]:[rank4]: return self._call_impl(*args, **kwargs)
|
| 358 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 359 |
+
[default4]:[rank4]: return forward_call(*args, **kwargs)
|
| 360 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 361 |
+
[default4]:[rank4]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 362 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 363 |
+
[default4]:[rank4]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 364 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 365 |
+
[default4]:[rank4]: return self._call_impl(*args, **kwargs)
|
| 366 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 367 |
+
[default4]:[rank4]: return forward_call(*args, **kwargs)
|
| 368 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 369 |
+
[default4]:[rank4]: output = self.pp_block(**new_kwargs)
|
| 370 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 371 |
+
[default4]:[rank4]: return self._call_impl(*args, **kwargs)
|
| 372 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 373 |
+
[default4]:[rank4]: return forward_call(*args, **kwargs)
|
| 374 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 629, in forward
|
| 375 |
+
[default4]:[rank4]: hidden_states = self.input_layernorm(hidden_states)
|
| 376 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 377 |
+
[default4]:[rank4]: return self._call_impl(*args, **kwargs)
|
| 378 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 379 |
+
[default4]:[rank4]: return forward_call(*args, **kwargs)
|
| 380 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/nn/layer_norm.py", line 42, in forward
|
| 381 |
+
[default4]:[rank4]: return layer_norm_fn(
|
| 382 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 875, in layer_norm_fn
|
| 383 |
+
[default4]:[rank4]: return LayerNormFn.apply(
|
| 384 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 385 |
+
[default4]:[rank4]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 386 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 748, in forward
|
| 387 |
+
[default4]:[rank4]: y, y1, mean, rstd, residual_out, seeds, dropout_mask, dropout_mask1 = _layer_norm_fwd(
|
| 388 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 335, in _layer_norm_fwd
|
| 389 |
+
[default4]:[rank4]: _layer_norm_fwd_1pass_kernel[(M,)](
|
| 390 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 391 |
+
[default4]:[rank4]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 392 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in run
|
| 393 |
+
[default4]:[rank4]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 394 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in <dictcomp>
|
| 395 |
+
[default4]:[rank4]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 396 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 122, in _bench
|
| 397 |
+
[default4]:[rank4]: return do_bench(kernel_call, warmup=self.warmup, rep=self.rep, quantiles=(0.5, 0.2, 0.8))
|
| 398 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/testing.py", line 102, in do_bench
|
| 399 |
+
[default4]:[rank4]: fn()
|
| 400 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 110, in kernel_call
|
| 401 |
+
[default4]:[rank4]: self.fn.run(
|
| 402 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 403 |
+
[default4]:[rank4]: return self.fn.run(*args, **kwargs)
|
| 404 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 405 |
+
[default4]:[rank4]: return self.fn.run(*args, **kwargs)
|
| 406 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 407 |
+
[default4]:[rank4]: return self.fn.run(*args, **kwargs)
|
| 408 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 409 |
+
[default4]:[rank4]: self.cache[device][key] = compile(
|
| 410 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 411 |
+
[default4]:[rank4]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 412 |
+
[default4]:[rank4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 413 |
+
[default4]:[rank4]: with open(temp_path, mode) as f:
|
| 414 |
+
[default4]:[rank4]: OSError: [Errno 122] Disk quota exceeded
|
| 415 |
+
[default4]:Exception in thread Thread-2 (_pin_memory_loop):
|
| 416 |
+
[default4]:Traceback (most recent call last):
|
| 417 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
|
| 418 |
+
[default4]: self.run()
|
| 419 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/threading.py", line 953, in run
|
| 420 |
+
[default4]: self._target(*self._args, **self._kwargs)
|
| 421 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/utils/data/_utils/pin_memory.py", line 54, in _pin_memory_loop
|
| 422 |
+
[default4]: do_one_step()
|
| 423 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/utils/data/_utils/pin_memory.py", line 31, in do_one_step
|
| 424 |
+
[default4]: r = in_queue.get(timeout=MP_STATUS_CHECK_INTERVAL)
|
| 425 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/queues.py", line 122, in get
|
| 426 |
+
[default4]: return _ForkingPickler.loads(res)
|
| 427 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/multiprocessing/reductions.py", line 495, in rebuild_storage_fd
|
| 428 |
+
[default4]: fd = df.detach()
|
| 429 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/resource_sharer.py", line 57, in detach
|
| 430 |
+
[default4]: with _resource_sharer.get_connection(self._id) as conn:
|
| 431 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/resource_sharer.py", line 86, in get_connection
|
| 432 |
+
[default4]: c = Client(address, authkey=process.current_process().authkey)
|
| 433 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 508, in Client
|
| 434 |
+
[default4]: answer_challenge(c, authkey)
|
| 435 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 752, in answer_challenge
|
| 436 |
+
[default4]: message = connection.recv_bytes(256) # reject large message
|
| 437 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 216, in recv_bytes
|
| 438 |
+
[default4]: buf = self._recv_bytes(maxlength)
|
| 439 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 414, in _recv_bytes
|
| 440 |
+
[default4]: buf = self._recv(4)
|
| 441 |
+
[default4]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 379, in _recv
|
| 442 |
+
[default4]: chunk = read(handle, remaining)
|
| 443 |
+
[default4]:ConnectionResetError: [Errno 104] Connection reset by peer
|
| 444 |
+
[default1]:[rank1]: OSError: [Errno 122] Disk quota exceeded
|
| 445 |
+
[default1]:
|
| 446 |
+
[default1]:[rank1]: During handling of the above exception, another exception occurred:
|
| 447 |
+
[default1]:
|
| 448 |
+
[default1]:[rank1]: Traceback (most recent call last):
|
| 449 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 450 |
+
[default1]:[rank1]: trainer.train(dataloader)
|
| 451 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 452 |
+
[default1]:[rank1]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 453 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 454 |
+
[default1]:[rank1]: outputs = self.pipeline_engine.train_batch_iter(
|
| 455 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 456 |
+
[default1]:[rank1]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 457 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 458 |
+
[default1]:[rank1]: output = model(**micro_batch)
|
| 459 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 460 |
+
[default1]:[rank1]: return self._call_impl(*args, **kwargs)
|
| 461 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 462 |
+
[default1]:[rank1]: return forward_call(*args, **kwargs)
|
| 463 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 464 |
+
[default1]:[rank1]: sharded_logits = self.model(
|
| 465 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 466 |
+
[default1]:[rank1]: return self._call_impl(*args, **kwargs)
|
| 467 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 468 |
+
[default1]:[rank1]: return forward_call(*args, **kwargs)
|
| 469 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 470 |
+
[default1]:[rank1]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 471 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 472 |
+
[default1]:[rank1]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 473 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 474 |
+
[default1]:[rank1]: return self._call_impl(*args, **kwargs)
|
| 475 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 476 |
+
[default1]:[rank1]: return forward_call(*args, **kwargs)
|
| 477 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 478 |
+
[default1]:[rank1]: output = self.pp_block(**new_kwargs)
|
| 479 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 480 |
+
[default1]:[rank1]: return self._call_impl(*args, **kwargs)
|
| 481 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 482 |
+
[default1]:[rank1]: return forward_call(*args, **kwargs)
|
| 483 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 629, in forward
|
| 484 |
+
[default1]:[rank1]: hidden_states = self.input_layernorm(hidden_states)
|
| 485 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 486 |
+
[default1]:[rank1]: return self._call_impl(*args, **kwargs)
|
| 487 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 488 |
+
[default1]:[rank1]: return forward_call(*args, **kwargs)
|
| 489 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/nn/layer_norm.py", line 42, in forward
|
| 490 |
+
[default1]:[rank1]: return layer_norm_fn(
|
| 491 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 875, in layer_norm_fn
|
| 492 |
+
[default1]:[rank1]: return LayerNormFn.apply(
|
| 493 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 494 |
+
[default1]:[rank1]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 495 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 748, in forward
|
| 496 |
+
[default1]:[rank1]: y, y1, mean, rstd, residual_out, seeds, dropout_mask, dropout_mask1 = _layer_norm_fwd(
|
| 497 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 335, in _layer_norm_fwd
|
| 498 |
+
[default1]:[rank1]: _layer_norm_fwd_1pass_kernel[(M,)](
|
| 499 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 500 |
+
[default1]:[rank1]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 501 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in run
|
| 502 |
+
[default1]:[rank1]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 503 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in <dictcomp>
|
| 504 |
+
[default1]:[rank1]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 505 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 122, in _bench
|
| 506 |
+
[default1]:[rank1]: return do_bench(kernel_call, warmup=self.warmup, rep=self.rep, quantiles=(0.5, 0.2, 0.8))
|
| 507 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/testing.py", line 102, in do_bench
|
| 508 |
+
[default1]:[rank1]: fn()
|
| 509 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 110, in kernel_call
|
| 510 |
+
[default1]:[rank1]: self.fn.run(
|
| 511 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 512 |
+
[default1]:[rank1]: return self.fn.run(*args, **kwargs)
|
| 513 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 514 |
+
[default1]:[rank1]: return self.fn.run(*args, **kwargs)
|
| 515 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 516 |
+
[default1]:[rank1]: return self.fn.run(*args, **kwargs)
|
| 517 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 518 |
+
[default1]:[rank1]: self.cache[device][key] = compile(
|
| 519 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 520 |
+
[default1]:[rank1]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 521 |
+
[default1]:[rank1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 522 |
+
[default1]:[rank1]: with open(temp_path, mode) as f:
|
| 523 |
+
[default1]:[rank1]: OSError: [Errno 122] Disk quota exceeded
|
| 524 |
+
[default1]:Exception in thread Thread-2 (_pin_memory_loop):
|
| 525 |
+
[default1]:Traceback (most recent call last):
|
| 526 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
|
| 527 |
+
[default1]: self.run()
|
| 528 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/threading.py", line 953, in run
|
| 529 |
+
[default1]: self._target(*self._args, **self._kwargs)
|
| 530 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/utils/data/_utils/pin_memory.py", line 54, in _pin_memory_loop
|
| 531 |
+
[default1]: do_one_step()
|
| 532 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/utils/data/_utils/pin_memory.py", line 31, in do_one_step
|
| 533 |
+
[default1]: r = in_queue.get(timeout=MP_STATUS_CHECK_INTERVAL)
|
| 534 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/queues.py", line 122, in get
|
| 535 |
+
[default1]: return _ForkingPickler.loads(res)
|
| 536 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/multiprocessing/reductions.py", line 495, in rebuild_storage_fd
|
| 537 |
+
[default1]: fd = df.detach()
|
| 538 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/resource_sharer.py", line 57, in detach
|
| 539 |
+
[default1]: with _resource_sharer.get_connection(self._id) as conn:
|
| 540 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/resource_sharer.py", line 86, in get_connection
|
| 541 |
+
[default1]: c = Client(address, authkey=process.current_process().authkey)
|
| 542 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 508, in Client
|
| 543 |
+
[default1]: answer_challenge(c, authkey)
|
| 544 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 752, in answer_challenge
|
| 545 |
+
[default1]: message = connection.recv_bytes(256) # reject large message
|
| 546 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 216, in recv_bytes
|
| 547 |
+
[default1]: buf = self._recv_bytes(maxlength)
|
| 548 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 414, in _recv_bytes
|
| 549 |
+
[default1]: buf = self._recv(4)
|
| 550 |
+
[default1]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/multiprocessing/connection.py", line 379, in _recv
|
| 551 |
+
[default1]: chunk = read(handle, remaining)
|
| 552 |
+
[default1]:ConnectionResetError: [Errno 104] Connection reset by peer
|
| 553 |
+
[default0]:[rank8]: OSError: [Errno 122] Disk quota exceeded
|
| 554 |
+
[default0]:
|
| 555 |
+
[default0]:[rank8]: During handling of the above exception, another exception occurred:
|
| 556 |
+
[default0]:
|
| 557 |
+
[default0]:[rank8]: Traceback (most recent call last):
|
| 558 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 559 |
+
[default0]:[rank8]: trainer.train(dataloader)
|
| 560 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 561 |
+
[default0]:[rank8]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 562 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 563 |
+
[default0]:[rank8]: outputs = self.pipeline_engine.train_batch_iter(
|
| 564 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 565 |
+
[default0]:[rank8]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 566 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 567 |
+
[default0]:[rank8]: output = model(**micro_batch)
|
| 568 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 569 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 570 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 571 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 572 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 573 |
+
[default0]:[rank8]: sharded_logits = self.model(
|
| 574 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 575 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 576 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 577 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 578 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 579 |
+
[default0]:[rank8]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 580 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 581 |
+
[default0]:[rank8]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 582 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 583 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 584 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 585 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 586 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 587 |
+
[default0]:[rank8]: output = self.pp_block(**new_kwargs)
|
| 588 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 589 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 590 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 591 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 592 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 629, in forward
|
| 593 |
+
[default0]:[rank8]: hidden_states = self.input_layernorm(hidden_states)
|
| 594 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 595 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 596 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 597 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 598 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/nn/layer_norm.py", line 42, in forward
|
| 599 |
+
[default0]:[rank8]: return layer_norm_fn(
|
| 600 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 875, in layer_norm_fn
|
| 601 |
+
[default0]:[rank8]: return LayerNormFn.apply(
|
| 602 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 603 |
+
[default0]:[rank8]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 604 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 748, in forward
|
| 605 |
+
[default0]:[rank8]: y, y1, mean, rstd, residual_out, seeds, dropout_mask, dropout_mask1 = _layer_norm_fwd(
|
| 606 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 335, in _layer_norm_fwd
|
| 607 |
+
[default0]:[rank8]: _layer_norm_fwd_1pass_kernel[(M,)](
|
| 608 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 609 |
+
[default0]:[rank8]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 610 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in run
|
| 611 |
+
[default0]:[rank8]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 612 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in <dictcomp>
|
| 613 |
+
[default0]:[rank8]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 614 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 122, in _bench
|
| 615 |
+
[default0]:[rank8]: return do_bench(kernel_call, warmup=self.warmup, rep=self.rep, quantiles=(0.5, 0.2, 0.8))
|
| 616 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/testing.py", line 102, in do_bench
|
| 617 |
+
[default0]:[rank8]: fn()
|
| 618 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 110, in kernel_call
|
| 619 |
+
[default0]:[rank8]: self.fn.run(
|
| 620 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 621 |
+
[default0]:[rank8]: return self.fn.run(*args, **kwargs)
|
| 622 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 623 |
+
[default0]:[rank8]: return self.fn.run(*args, **kwargs)
|
| 624 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 625 |
+
[default0]:[rank8]: return self.fn.run(*args, **kwargs)
|
| 626 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 627 |
+
[default0]:[rank8]: self.cache[device][key] = compile(
|
| 628 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 629 |
+
[default0]:[rank8]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 630 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 631 |
+
[default0]:[rank8]: with open(temp_path, mode) as f:
|
| 632 |
+
[default0]:[rank8]: OSError: [Errno 122] Disk quota exceeded
|
| 633 |
+
[default0]:Exception in thread Thread-2 (_pin_memory_loop):
|
| 634 |
+
[default0]:Traceback (most recent call last):
|
| 635 |
+
[default0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
|
| 636 |
+
W0702 16:33:04.893000 140209290786624 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 891756 closing signal SIGTERM
|
| 637 |
+
W0702 16:33:04.898000 140209290786624 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 891758 closing signal SIGTERM
|
| 638 |
+
W0702 16:33:04.896000 139646960318272 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1239957 closing signal SIGTERM
|
| 639 |
+
W0702 16:33:04.903000 140209290786624 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 891759 closing signal SIGTERM
|
| 640 |
+
W0702 16:33:04.900000 139646960318272 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1239958 closing signal SIGTERM
|
| 641 |
+
W0702 16:33:04.903000 139646960318272 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1239959 closing signal SIGTERM
|
| 642 |
+
W0702 16:33:04.908000 140209290786624 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 891761 closing signal SIGTERM
|
| 643 |
+
W0702 16:33:04.908000 139646960318272 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1239960 closing signal SIGTERM
|
| 644 |
+
W0702 16:33:04.927000 140209290786624 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 891762 closing signal SIGTERM
|
| 645 |
+
W0702 16:33:04.940000 139646960318272 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1239961 closing signal SIGTERM
|
| 646 |
+
W0702 16:33:04.942000 139646960318272 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1239962 closing signal SIGTERM
|
| 647 |
+
W0702 16:33:04.948000 139646960318272 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 1239963 closing signal SIGTERM
|
| 648 |
+
E0702 16:33:07.383000 140209290786624 torch/distributed/elastic/multiprocessing/api.py:826] failed (exitcode: 1) local_rank: 1 (pid: 891757) of binary: /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/python3.10
|
| 649 |
+
E0702 16:33:07.446000 139646960318272 torch/distributed/elastic/multiprocessing/api.py:826] failed (exitcode: 1) local_rank: 0 (pid: 1239956) of binary: /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/python3.10
|
| 650 |
+
Traceback (most recent call last):
|
| 651 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/torchrun", line 8, in <module>
|
| 652 |
+
sys.exit(main())
|
| 653 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 347, in wrapper
|
| 654 |
+
return f(*args, **kwargs)
|
| 655 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 879, in main
|
| 656 |
+
run(args)
|
| 657 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 870, in run
|
| 658 |
+
elastic_launch(
|
| 659 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
|
| 660 |
+
return launch_agent(self._config, self._entrypoint, list(args))
|
| 661 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 263, in launch_agent
|
| 662 |
+
raise ChildFailedError(
|
| 663 |
+
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
|
| 664 |
+
============================================================
|
| 665 |
+
/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py FAILED
|
| 666 |
+
------------------------------------------------------------
|
| 667 |
+
Failures:
|
| 668 |
+
[1]:
|
| 669 |
+
time : 2024-07-02_16:33:04
|
| 670 |
+
host : ip-26-0-160-192.ec2.internal
|
| 671 |
+
rank : 4 (local_rank: 4)
|
| 672 |
+
exitcode : 1 (pid: 891760)
|
| 673 |
+
error_file: <N/A>
|
| 674 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 675 |
+
[2]:
|
| 676 |
+
time : 2024-07-02_16:33:04
|
| 677 |
+
host : ip-26-0-160-192.ec2.internal
|
| 678 |
+
rank : 7 (local_rank: 7)
|
| 679 |
+
exitcode : 1 (pid: 891763)
|
| 680 |
+
error_file: <N/A>
|
| 681 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 682 |
+
------------------------------------------------------------
|
| 683 |
+
Root Cause (first observed failure):
|
| 684 |
+
[0]:
|
| 685 |
+
time : 2024-07-02_16:33:04
|
| 686 |
+
host : ip-26-0-160-192.ec2.internal
|
| 687 |
+
rank : 1 (local_rank: 1)
|
| 688 |
+
exitcode : 1 (pid: 891757)
|
| 689 |
+
error_file: <N/A>
|
| 690 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 691 |
+
============================================================
|
| 692 |
+
Traceback (most recent call last):
|
| 693 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/torchrun", line 8, in <module>
|
| 694 |
+
sys.exit(main())
|
| 695 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 347, in wrapper
|
| 696 |
+
return f(*args, **kwargs)
|
| 697 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 879, in main
|
| 698 |
+
run(args)
|
| 699 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 870, in run
|
| 700 |
+
elastic_launch(
|
| 701 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
|
| 702 |
+
return launch_agent(self._config, self._entrypoint, list(args))
|
| 703 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 263, in launch_agent
|
| 704 |
+
raise ChildFailedError(
|
| 705 |
+
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
|
| 706 |
+
============================================================
|
| 707 |
+
/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py FAILED
|
| 708 |
+
------------------------------------------------------------
|
| 709 |
+
Failures:
|
| 710 |
+
<NO_OTHER_FAILURES>
|
| 711 |
+
------------------------------------------------------------
|
| 712 |
+
Root Cause (first observed failure):
|
| 713 |
+
[0]:
|
| 714 |
+
time : 2024-07-02_16:33:04
|
| 715 |
+
host : ip-26-0-162-233.ec2.internal
|
| 716 |
+
rank : 8 (local_rank: 0)
|
| 717 |
+
exitcode : 1 (pid: 1239956)
|
| 718 |
+
error_file: <N/A>
|
| 719 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 720 |
+
============================================================
|
| 721 |
+
srun: error: ip-26-0-160-192: task 0: Exited with exit code 1
|
| 722 |
+
srun: error: ip-26-0-162-233: task 1: Exited with exit code 1
|
| 723 |
+
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-1_tp-16_pp-1_mbz-64/status.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
fail
|