Upload llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4
Browse files
llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/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-8_tp-2_pp-1_mbz-4/log.out
|
| 13 |
+
#SBATCH --error=/fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/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-8_tp-2_pp-1_mbz-4/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-8_tp-2_pp-1_mbz-4/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-8_tp-2_pp-1_mbz-4/status.txt
|
| 84 |
+
else
|
| 85 |
+
if grep -q "OutOfMemoryError" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/log.out; then
|
| 86 |
+
printf "oom" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/status.txt
|
| 87 |
+
elif grep -q " CUDA error: an illegal memory access" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/log.out; then
|
| 88 |
+
printf "oom" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/status.txt
|
| 89 |
+
elif grep -q "Timeout at NCCL" /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/log.out; then
|
| 90 |
+
printf "timeout" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/status.txt
|
| 91 |
+
else
|
| 92 |
+
printf "fail" > /fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4/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-8_tp-2_pp-1_mbz-4 --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-8_tp-2_pp-1_mbz-4 --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-8_tp-2_pp-1_mbz-4 llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4 --commit-message "Upload llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4"
|
| 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-8_tp-2_pp-1_mbz-4/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: 8
|
| 49 |
+
expert_parallel_size: 1
|
| 50 |
+
pp: 1
|
| 51 |
+
pp_engine: 1f1b
|
| 52 |
+
tp: 2
|
| 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-8_tp-2_pp-1_mbz-4
|
| 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: 32
|
| 79 |
+
limit_test_batches: 0
|
| 80 |
+
limit_val_batches: 0
|
| 81 |
+
micro_batch_size: 4
|
| 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-8_tp-2_pp-1_mbz-4/log.out
ADDED
|
@@ -0,0 +1,930 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
========================
|
| 2 |
+
START TIME: Tue Jul 2 16:31:21 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 16:31:24.185000 140446723327808 torch/distributed/run.py:757]
|
| 18 |
+
W0702 16:31:24.185000 140446723327808 torch/distributed/run.py:757] *****************************************
|
| 19 |
+
W0702 16:31:24.185000 140446723327808 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 16:31:24.185000 140446723327808 torch/distributed/run.py:757] *****************************************
|
| 21 |
+
W0702 16:31:24.204000 139668155955008 torch/distributed/run.py:757]
|
| 22 |
+
W0702 16:31:24.204000 139668155955008 torch/distributed/run.py:757] *****************************************
|
| 23 |
+
W0702 16:31:24.204000 139668155955008 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 16:31:24.204000 139668155955008 torch/distributed/run.py:757] *****************************************
|
| 25 |
+
[default0]:07/02/2024 16:31:42 [WARNING|DP=0|PP=0|TP=0|ip-26-0-171-62]: [Vocab Size Padding] Padded vocab (size: 50257) with 1 dummy tokens (new size: 50258)
|
| 26 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Config:
|
| 27 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Config(general=GeneralArgs(project='bench_cluster',
|
| 28 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: run='%date_%jobid',
|
| 29 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: seed=42,
|
| 30 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: step=None,
|
| 31 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: consumed_train_samples=None,
|
| 32 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: benchmark_csv_path=None,
|
| 33 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: ignore_sanity_checks=True),
|
| 34 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: parallelism=ParallelismArgs(dp=8,
|
| 35 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: pp=1,
|
| 36 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tp=2,
|
| 37 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: pp_engine=<nanotron.parallel.pipeline_parallel.engine.OneForwardOneBackwardPipelineEngine object at 0x7f3c75348790>,
|
| 38 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tp_mode=<TensorParallelLinearMode.REDUCE_SCATTER: 2>,
|
| 39 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tp_linear_async_communication=False,
|
| 40 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: expert_parallel_size=1),
|
| 41 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: model=ModelArgs(model_config=LlamaConfig(bos_token_id=1,
|
| 42 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: eos_token_id=2,
|
| 43 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: hidden_act='silu',
|
| 44 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: hidden_size=2048,
|
| 45 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: initializer_range=0.02,
|
| 46 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: intermediate_size=4096,
|
| 47 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: is_llama_config=True,
|
| 48 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: max_position_embeddings=4096,
|
| 49 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: num_attention_heads=32,
|
| 50 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: num_hidden_layers=24,
|
| 51 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: num_key_value_heads=32,
|
| 52 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: pad_token_id=None,
|
| 53 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: pretraining_tp=1,
|
| 54 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: rms_norm_eps=1e-05,
|
| 55 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: rope_scaling=None,
|
| 56 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: rope_theta=10000.0,
|
| 57 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tie_word_embeddings=True,
|
| 58 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: use_cache=True,
|
| 59 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: vocab_size=50258),
|
| 60 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: init_method=RandomInit(std=0.025),
|
| 61 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: dtype=torch.bfloat16,
|
| 62 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: make_vocab_size_divisible_by=1,
|
| 63 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: ddp_bucket_cap_mb=25),
|
| 64 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tokenizer=TokenizerArgs(tokenizer_name_or_path='openai-community/gpt2',
|
| 65 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tokenizer_revision=None,
|
| 66 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tokenizer_max_length=None),
|
| 67 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: checkpoints=CheckpointsArgs(checkpoints_path=Path('/dev/null'),
|
| 68 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: checkpoint_interval=100000,
|
| 69 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: save_initial_state=False,
|
| 70 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: resume_checkpoint_path=None,
|
| 71 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: checkpoints_path_is_shared_file_system=False),
|
| 72 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: logging=LoggingArgs(log_level='info',
|
| 73 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: log_level_replica='info',
|
| 74 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: iteration_step_info_interval=1),
|
| 75 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tokens=TokensArgs(sequence_length=4096,
|
| 76 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: train_steps=20,
|
| 77 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: micro_batch_size=4,
|
| 78 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: batch_accumulation_per_replica=32,
|
| 79 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: val_check_interval=-1,
|
| 80 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: limit_val_batches=0,
|
| 81 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: limit_test_batches=0),
|
| 82 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: optimizer=OptimizerArgs(optimizer_factory=AdamWOptimizerArgs(adam_eps=1e-08,
|
| 83 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: adam_beta1=0.9,
|
| 84 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: adam_beta2=0.95,
|
| 85 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: torch_adam_is_fused=True,
|
| 86 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: name='adamW'),
|
| 87 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: zero_stage=1,
|
| 88 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: weight_decay=0.01,
|
| 89 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: clip_grad=1.0,
|
| 90 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: accumulate_grad_in_fp32=True,
|
| 91 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: learning_rate_scheduler=LRSchedulerArgs(learning_rate=0.0001,
|
| 92 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: lr_warmup_steps=1,
|
| 93 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: lr_warmup_style='linear',
|
| 94 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: lr_decay_style='linear',
|
| 95 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: lr_decay_steps=19,
|
| 96 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: lr_decay_starting_step=None,
|
| 97 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: min_decay_lr=1e-05)),
|
| 98 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: data_stages=[DatasetStageArgs(name='Training Stage',
|
| 99 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: start_training_step=1,
|
| 100 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: data=DataArgs(dataset=PretrainDatasetsArgs(hf_dataset_or_datasets='roneneldan/TinyStories',
|
| 101 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: hf_dataset_splits='train',
|
| 102 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: hf_dataset_config_name=None,
|
| 103 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: dataset_processing_num_proc_per_process=64,
|
| 104 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: dataset_overwrite_cache=False,
|
| 105 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: text_column_name='text'),
|
| 106 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: seed=42,
|
| 107 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: num_loading_workers=32))],
|
| 108 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: profiler=ProfilerArgs(profiler_export_path=Path('/fsx/ferdinandmom/ferdinand-hf/bench_cluster/results/llama-1B/16_GPUS/dp-8_tp-2_pp-1_mbz-4')),
|
| 109 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: lighteval=None)
|
| 110 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Model Config:
|
| 111 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: LlamaConfig(bos_token_id=1,
|
| 112 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: eos_token_id=2,
|
| 113 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: hidden_act='silu',
|
| 114 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: hidden_size=2048,
|
| 115 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: initializer_range=0.02,
|
| 116 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: intermediate_size=4096,
|
| 117 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: is_llama_config=True,
|
| 118 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: max_position_embeddings=4096,
|
| 119 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: num_attention_heads=32,
|
| 120 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: num_hidden_layers=24,
|
| 121 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: num_key_value_heads=32,
|
| 122 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: pad_token_id=None,
|
| 123 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: pretraining_tp=1,
|
| 124 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: rms_norm_eps=1e-05,
|
| 125 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: rope_scaling=None,
|
| 126 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: rope_theta=10000.0,
|
| 127 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: tie_word_embeddings=True,
|
| 128 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: use_cache=True,
|
| 129 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: vocab_size=50258)
|
| 130 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Building model..
|
| 131 |
+
[default0]:07/02/2024 16:31:42 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Setting PP block ranks...
|
| 132 |
+
[default1]:07/02/2024 16:31:53 [INFO|DP=0|PP=0|TP=1|ip-26-0-171-62]: Local number of parameters: 555M (1058.35MiB)
|
| 133 |
+
[default1]:07/02/2024 16:31:53 [INFO|DP=0|PP=0|TP=1|ip-26-0-171-62]: [After model building] Memory usage: 1082.37MiB. Peak allocated: 1182.56MiB Peak reserved: 1200.00MiB
|
| 134 |
+
[default1]:07/02/2024 16:31:53 [INFO|DP=0|PP=0|TP=1|ip-26-0-171-62]: No checkpoint path provided.
|
| 135 |
+
[default0]:07/02/2024 16:31:53 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Total number of parameters: 1.11G (2116.70MiB)
|
| 136 |
+
[default0]:07/02/2024 16:31:53 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Local number of parameters: 555M (1058.35MiB)
|
| 137 |
+
[default0]:07/02/2024 16:31:53 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [After model building] Memory usage: 1082.37MiB. Peak allocated: 1182.56MiB Peak reserved: 1200.00MiB
|
| 138 |
+
[default0]:07/02/2024 16:31:53 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: No checkpoint path provided.
|
| 139 |
+
[default0]:07/02/2024 16:31:53 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Parametrizing model parameters using StandardParametrizator
|
| 140 |
+
[default2]:07/02/2024 16:31:53 [INFO|DP=5|PP=0|TP=0|ip-26-0-171-88]: No checkpoint path provided.
|
| 141 |
+
[default3]:07/02/2024 16:31:53 [INFO|DP=5|PP=0|TP=1|ip-26-0-171-88]: No checkpoint path provided.
|
| 142 |
+
[default0]:07/02/2024 16:31:53 [INFO|DP=4|PP=0|TP=0|ip-26-0-171-88]: No checkpoint path provided.
|
| 143 |
+
[default1]:07/02/2024 16:31:53 [INFO|DP=4|PP=0|TP=1|ip-26-0-171-88]: No checkpoint path provided.
|
| 144 |
+
[default5]:07/02/2024 16:31:53 [INFO|DP=2|PP=0|TP=1|ip-26-0-171-62]: No checkpoint path provided.
|
| 145 |
+
[default3]:07/02/2024 16:31:53 [INFO|DP=1|PP=0|TP=1|ip-26-0-171-62]: No checkpoint path provided.
|
| 146 |
+
[default2]:07/02/2024 16:31:53 [INFO|DP=1|PP=0|TP=0|ip-26-0-171-62]: No checkpoint path provided.
|
| 147 |
+
[default4]:07/02/2024 16:31:53 [INFO|DP=2|PP=0|TP=0|ip-26-0-171-62]: No checkpoint path provided.
|
| 148 |
+
[default7]:07/02/2024 16:31:53 [INFO|DP=7|PP=0|TP=1|ip-26-0-171-88]: No checkpoint path provided.
|
| 149 |
+
[default6]:07/02/2024 16:31:53 [INFO|DP=7|PP=0|TP=0|ip-26-0-171-88]: No checkpoint path provided.
|
| 150 |
+
[default5]:07/02/2024 16:31:53 [INFO|DP=6|PP=0|TP=1|ip-26-0-171-88]: No checkpoint path provided.
|
| 151 |
+
[default4]:07/02/2024 16:31:53 [INFO|DP=6|PP=0|TP=0|ip-26-0-171-88]: No checkpoint path provided.
|
| 152 |
+
[default7]:07/02/2024 16:31:53 [INFO|DP=3|PP=0|TP=1|ip-26-0-171-62]: No checkpoint path provided.
|
| 153 |
+
[default6]:07/02/2024 16:31:53 [INFO|DP=3|PP=0|TP=0|ip-26-0-171-62]: No checkpoint path provided.
|
| 154 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [Optimizer Building] Using LearningRateForSP as learning rate
|
| 155 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] Size of optimizer params per rank:
|
| 156 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] DP Rank 0 has 69.4M out of 555M (12.50%) params' optimizer states
|
| 157 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] DP Rank 1 has 69.4M out of 555M (12.50%) params' optimizer states
|
| 158 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] DP Rank 2 has 69.4M out of 555M (12.50%) params' optimizer states
|
| 159 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] DP Rank 3 has 69.4M out of 555M (12.50%) params' optimizer states
|
| 160 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] DP Rank 4 has 69.4M out of 555M (12.50%) params' optimizer states
|
| 161 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] DP Rank 5 has 69.4M out of 555M (12.50%) params' optimizer states
|
| 162 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] DP Rank 6 has 69.4M out of 555M (12.50%) params' optimizer states
|
| 163 |
+
[default0]:07/02/2024 16:31:58 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [ZeRO sharding] DP Rank 7 has 69.4M out of 555M (12.50%) params' optimizer states
|
| 164 |
+
[default0]:07/02/2024 16:32:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [Training Plan] Stage Training Stage has 19 remaining training steps and has consumed 0 samples
|
| 165 |
+
[default0]:07/02/2024 16:32:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Using `datasets` library
|
| 166 |
+
[default0]:07/02/2024 16:32:00 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Loading tokenizer from openai-community/gpt2 and transformers/hf_hub versions ('4.41.2', '0.23.4')
|
| 167 |
+
[default0]:Repo card metadata block was not found. Setting CardData to empty.
|
| 168 |
+
[default0]:07/02/2024 16:32:01 [WARNING|DP=0|PP=0|TP=0|ip-26-0-171-62]: Repo card metadata block was not found. Setting CardData to empty.
|
| 169 |
+
[default0]:07/02/2024 16:32:01 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [Training Plan] There are 1 training stages
|
| 170 |
+
[default0]:07/02/2024 16:32:01 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [Stage Training Stage] start from step 1
|
| 171 |
+
[default0]:07/02/2024 16:32:01 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]:
|
| 172 |
+
[default0]:07/02/2024 16:32:01 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: [Start training] datetime: 2024-07-02 16:32:01.872683 | mbs: 4 | grad_accum: 32 | global_batch_size: 1024 | sequence_length: 4096 | train_steps: 20 | start_iteration_step: 0 | consumed_train_samples: 0
|
| 173 |
+
[default0]:07/02/2024 16:32:01 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Resuming training from stage Training Stage, it has trained for 0 samples and has 19 remaining train steps
|
| 174 |
+
[default0]:07/02/2024 16:32:01 [INFO|DP=0|PP=0|TP=0|ip-26-0-171-62]: Memory usage: 3463.66MiB. Peak allocated 3463.66MiB. Peak reserved: 3584.00MiB
|
| 175 |
+
[default4]:07/02/2024 16:32:02 [WARNING|DP=2|PP=0|TP=0|ip-26-0-171-62]: Repo card metadata block was not found. Setting CardData to empty.
|
| 176 |
+
[default5]:07/02/2024 16:32:02 [WARNING|DP=2|PP=0|TP=1|ip-26-0-171-62]: Repo card metadata block was not found. Setting CardData to empty.
|
| 177 |
+
[default0]:07/02/2024 16:32:02 [WARNING|DP=4|PP=0|TP=0|ip-26-0-171-88]: Repo card metadata block was not found. Setting CardData to empty.
|
| 178 |
+
[default1]:07/02/2024 16:32:02 [WARNING|DP=0|PP=0|TP=1|ip-26-0-171-62]: Repo card metadata block was not found. Setting CardData to empty.
|
| 179 |
+
[default6]:07/02/2024 16:32:02 [WARNING|DP=7|PP=0|TP=0|ip-26-0-171-88]: Repo card metadata block was not found. Setting CardData to empty.
|
| 180 |
+
[default2]:07/02/2024 16:32:02 [WARNING|DP=5|PP=0|TP=0|ip-26-0-171-88]: Repo card metadata block was not found. Setting CardData to empty.
|
| 181 |
+
[default4]:Repo card metadata block was not found. Setting CardData to empty.
|
| 182 |
+
[default2]:Repo card metadata block was not found. Setting CardData to empty.
|
| 183 |
+
[default5]:Repo card metadata block was not found. Setting CardData to empty.
|
| 184 |
+
[default0]:Repo card metadata block was not found. Setting CardData to empty.
|
| 185 |
+
[default4]:07/02/2024 16:32:02 [WARNING|DP=6|PP=0|TP=0|ip-26-0-171-88]: Repo card metadata block was not found. Setting CardData to empty.
|
| 186 |
+
[default5]:07/02/2024 16:32:02 [WARNING|DP=6|PP=0|TP=1|ip-26-0-171-88]: Repo card metadata block was not found. Setting CardData to empty.
|
| 187 |
+
[default6]:07/02/2024 16:32:02 [WARNING|DP=3|PP=0|TP=0|ip-26-0-171-62]: Repo card metadata block was not found. Setting CardData to empty.
|
| 188 |
+
[default6]:Repo card metadata block was not found. Setting CardData to empty.
|
| 189 |
+
[default6]:Repo card metadata block was not found. Setting CardData to empty.
|
| 190 |
+
[default5]:Repo card metadata block was not found. Setting CardData to empty.
|
| 191 |
+
[default4]:Repo card metadata block was not found. Setting CardData to empty.
|
| 192 |
+
[default1]:Repo card metadata block was not found. Setting CardData to empty.
|
| 193 |
+
[default3]:07/02/2024 16:32:02 [WARNING|DP=1|PP=0|TP=1|ip-26-0-171-62]: Repo card metadata block was not found. Setting CardData to empty.
|
| 194 |
+
[default2]:07/02/2024 16:32:02 [WARNING|DP=1|PP=0|TP=0|ip-26-0-171-62]: Repo card metadata block was not found. Setting CardData to empty.
|
| 195 |
+
[default1]:07/02/2024 16:32:02 [WARNING|DP=4|PP=0|TP=1|ip-26-0-171-88]: Repo card metadata block was not found. Setting CardData to empty.
|
| 196 |
+
[default3]:Repo card metadata block was not found. Setting CardData to empty.
|
| 197 |
+
[default3]:07/02/2024 16:32:02 [WARNING|DP=5|PP=0|TP=1|ip-26-0-171-88]: Repo card metadata block was not found. Setting CardData to empty.
|
| 198 |
+
[default3]:Repo card metadata block was not found. Setting CardData to empty.
|
| 199 |
+
[default7]:Repo card metadata block was not found. Setting CardData to empty.
|
| 200 |
+
[default1]:Repo card metadata block was not found. Setting CardData to empty.
|
| 201 |
+
[default2]:Repo card metadata block was not found. Setting CardData to empty.
|
| 202 |
+
[default7]:07/02/2024 16:32:02 [WARNING|DP=7|PP=0|TP=1|ip-26-0-171-88]: Repo card metadata block was not found. Setting CardData to empty.
|
| 203 |
+
[default7]:Repo card metadata block was not found. Setting CardData to empty.
|
| 204 |
+
[default7]:07/02/2024 16:32:02 [WARNING|DP=3|PP=0|TP=1|ip-26-0-171-62]: Repo card metadata block was not found. Setting CardData to empty.
|
| 205 |
+
[default0]:[rank0]: OSError: [Errno 122] Disk quota exceeded
|
| 206 |
+
[default0]:
|
| 207 |
+
[default0]:[rank0]: During handling of the above exception, another exception occurred:
|
| 208 |
+
[default0]:
|
| 209 |
+
[default0]:[rank0]: Traceback (most recent call last):
|
| 210 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 211 |
+
[default0]:[rank0]: trainer.train(dataloader)
|
| 212 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 213 |
+
[default0]:[rank0]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 214 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 215 |
+
[default0]:[rank0]: outputs = self.pipeline_engine.train_batch_iter(
|
| 216 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 217 |
+
[default0]:[rank0]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 218 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 219 |
+
[default0]:[rank0]: output = model(**micro_batch)
|
| 220 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 221 |
+
[default0]:[rank0]: return self._call_impl(*args, **kwargs)
|
| 222 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 223 |
+
[default0]:[rank0]: return forward_call(*args, **kwargs)
|
| 224 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 225 |
+
[default0]:[rank0]: sharded_logits = self.model(
|
| 226 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 227 |
+
[default0]:[rank0]: return self._call_impl(*args, **kwargs)
|
| 228 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 229 |
+
[default0]:[rank0]: return forward_call(*args, **kwargs)
|
| 230 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 231 |
+
[default0]:[rank0]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 232 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 233 |
+
[default0]:[rank0]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 234 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 235 |
+
[default0]:[rank0]: return self._call_impl(*args, **kwargs)
|
| 236 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 237 |
+
[default0]:[rank0]: return forward_call(*args, **kwargs)
|
| 238 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 239 |
+
[default0]:[rank0]: output = self.pp_block(**new_kwargs)
|
| 240 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 241 |
+
[default0]:[rank0]: return self._call_impl(*args, **kwargs)
|
| 242 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 243 |
+
[default0]:[rank0]: return forward_call(*args, **kwargs)
|
| 244 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 629, in forward
|
| 245 |
+
[default0]:[rank0]: hidden_states = self.input_layernorm(hidden_states)
|
| 246 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 247 |
+
[default0]:[rank0]: return self._call_impl(*args, **kwargs)
|
| 248 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 249 |
+
[default0]:[rank0]: return forward_call(*args, **kwargs)
|
| 250 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/nn/layer_norm.py", line 42, in forward
|
| 251 |
+
[default0]:[rank0]: return layer_norm_fn(
|
| 252 |
+
[default0]:[rank0]: 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
|
| 253 |
+
[default0]:[rank0]: return LayerNormFn.apply(
|
| 254 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 255 |
+
[default0]:[rank0]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 256 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 748, in forward
|
| 257 |
+
[default0]:[rank0]: y, y1, mean, rstd, residual_out, seeds, dropout_mask, dropout_mask1 = _layer_norm_fwd(
|
| 258 |
+
[default0]:[rank0]: 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
|
| 259 |
+
[default0]:[rank0]: _layer_norm_fwd_1pass_kernel[(M,)](
|
| 260 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 261 |
+
[default0]:[rank0]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 262 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in run
|
| 263 |
+
[default0]:[rank0]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 264 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in <dictcomp>
|
| 265 |
+
[default0]:[rank0]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 266 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 122, in _bench
|
| 267 |
+
[default0]:[rank0]: return do_bench(kernel_call, warmup=self.warmup, rep=self.rep, quantiles=(0.5, 0.2, 0.8))
|
| 268 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/testing.py", line 102, in do_bench
|
| 269 |
+
[default0]:[rank0]: fn()
|
| 270 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 110, in kernel_call
|
| 271 |
+
[default0]:[rank0]: self.fn.run(
|
| 272 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 273 |
+
[default0]:[rank0]: return self.fn.run(*args, **kwargs)
|
| 274 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 275 |
+
[default0]:[rank0]: return self.fn.run(*args, **kwargs)
|
| 276 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 277 |
+
[default0]:[rank0]: return self.fn.run(*args, **kwargs)
|
| 278 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 279 |
+
[default0]:[rank0]: self.cache[device][key] = compile(
|
| 280 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 281 |
+
[default0]:[rank0]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 282 |
+
[default0]:[rank0]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 283 |
+
[default0]:[rank0]: with open(temp_path, mode) as f:
|
| 284 |
+
[default0]:[rank0]: OSError: [Errno 122] Disk quota exceeded
|
| 285 |
+
[default5]:[rank5]: OSError: [Errno 122] Disk quota exceeded
|
| 286 |
+
[default5]:
|
| 287 |
+
[default5]:[rank5]: During handling of the above exception, another exception occurred:
|
| 288 |
+
[default5]:
|
| 289 |
+
[default5]:[rank5]: Traceback (most recent call last):
|
| 290 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 291 |
+
[default5]:[rank5]: trainer.train(dataloader)
|
| 292 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 293 |
+
[default5]:[rank5]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 294 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 295 |
+
[default5]:[rank5]: outputs = self.pipeline_engine.train_batch_iter(
|
| 296 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 297 |
+
[default5]:[rank5]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 298 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 299 |
+
[default5]:[rank5]: output = model(**micro_batch)
|
| 300 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 301 |
+
[default5]:[rank5]: return self._call_impl(*args, **kwargs)
|
| 302 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 303 |
+
[default5]:[rank5]: return forward_call(*args, **kwargs)
|
| 304 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 305 |
+
[default5]:[rank5]: sharded_logits = self.model(
|
| 306 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 307 |
+
[default5]:[rank5]: return self._call_impl(*args, **kwargs)
|
| 308 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 309 |
+
[default5]:[rank5]: return forward_call(*args, **kwargs)
|
| 310 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 311 |
+
[default5]:[rank5]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 312 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 313 |
+
[default5]:[rank5]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 314 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 315 |
+
[default5]:[rank5]: return self._call_impl(*args, **kwargs)
|
| 316 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 317 |
+
[default5]:[rank5]: return forward_call(*args, **kwargs)
|
| 318 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 319 |
+
[default5]:[rank5]: output = self.pp_block(**new_kwargs)
|
| 320 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 321 |
+
[default5]:[rank5]: return self._call_impl(*args, **kwargs)
|
| 322 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 323 |
+
[default5]:[rank5]: return forward_call(*args, **kwargs)
|
| 324 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 629, in forward
|
| 325 |
+
[default5]:[rank5]: hidden_states = self.input_layernorm(hidden_states)
|
| 326 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 327 |
+
[default5]:[rank5]: return self._call_impl(*args, **kwargs)
|
| 328 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 329 |
+
[default5]:[rank5]: return forward_call(*args, **kwargs)
|
| 330 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/nn/layer_norm.py", line 42, in forward
|
| 331 |
+
[default5]:[rank5]: return layer_norm_fn(
|
| 332 |
+
[default5]:[rank5]: 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
|
| 333 |
+
[default5]:[rank5]: return LayerNormFn.apply(
|
| 334 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 335 |
+
[default5]:[rank5]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 336 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 748, in forward
|
| 337 |
+
[default5]:[rank5]: y, y1, mean, rstd, residual_out, seeds, dropout_mask, dropout_mask1 = _layer_norm_fwd(
|
| 338 |
+
[default5]:[rank5]: 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
|
| 339 |
+
[default5]:[rank5]: _layer_norm_fwd_1pass_kernel[(M,)](
|
| 340 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 341 |
+
[default5]:[rank5]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 342 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in run
|
| 343 |
+
[default5]:[rank5]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 344 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in <dictcomp>
|
| 345 |
+
[default5]:[rank5]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 346 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 122, in _bench
|
| 347 |
+
[default5]:[rank5]: return do_bench(kernel_call, warmup=self.warmup, rep=self.rep, quantiles=(0.5, 0.2, 0.8))
|
| 348 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/testing.py", line 102, in do_bench
|
| 349 |
+
[default5]:[rank5]: fn()
|
| 350 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 110, in kernel_call
|
| 351 |
+
[default5]:[rank5]: self.fn.run(
|
| 352 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 353 |
+
[default5]:[rank5]: return self.fn.run(*args, **kwargs)
|
| 354 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 355 |
+
[default5]:[rank5]: return self.fn.run(*args, **kwargs)
|
| 356 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 357 |
+
[default5]:[rank5]: return self.fn.run(*args, **kwargs)
|
| 358 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 359 |
+
[default5]:[rank5]: self.cache[device][key] = compile(
|
| 360 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 361 |
+
[default5]:[rank5]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 362 |
+
[default5]:[rank5]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 363 |
+
[default5]:[rank5]: with open(temp_path, mode) as f:
|
| 364 |
+
[default5]:[rank5]: OSError: [Errno 122] Disk quota exceeded
|
| 365 |
+
[default3]:[rank3]: OSError: [Errno 122] Disk quota exceeded
|
| 366 |
+
[default3]:
|
| 367 |
+
[default3]:[rank3]: During handling of the above exception, another exception occurred:
|
| 368 |
+
[default3]:
|
| 369 |
+
[default3]:[rank3]: Traceback (most recent call last):
|
| 370 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 371 |
+
[default3]:[rank3]: trainer.train(dataloader)
|
| 372 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 373 |
+
[default3]:[rank3]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 374 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 375 |
+
[default3]:[rank3]: outputs = self.pipeline_engine.train_batch_iter(
|
| 376 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 377 |
+
[default3]:[rank3]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 378 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 379 |
+
[default3]:[rank3]: output = model(**micro_batch)
|
| 380 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 381 |
+
[default3]:[rank3]: return self._call_impl(*args, **kwargs)
|
| 382 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 383 |
+
[default3]:[rank3]: return forward_call(*args, **kwargs)
|
| 384 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 385 |
+
[default3]:[rank3]: sharded_logits = self.model(
|
| 386 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 387 |
+
[default3]:[rank3]: return self._call_impl(*args, **kwargs)
|
| 388 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 389 |
+
[default3]:[rank3]: return forward_call(*args, **kwargs)
|
| 390 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 391 |
+
[default3]:[rank3]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 392 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 393 |
+
[default3]:[rank3]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 394 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 395 |
+
[default3]:[rank3]: return self._call_impl(*args, **kwargs)
|
| 396 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 397 |
+
[default3]:[rank3]: return forward_call(*args, **kwargs)
|
| 398 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 399 |
+
[default3]:[rank3]: output = self.pp_block(**new_kwargs)
|
| 400 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 401 |
+
[default3]:[rank3]: return self._call_impl(*args, **kwargs)
|
| 402 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 403 |
+
[default3]:[rank3]: return forward_call(*args, **kwargs)
|
| 404 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 629, in forward
|
| 405 |
+
[default3]:[rank3]: hidden_states = self.input_layernorm(hidden_states)
|
| 406 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 407 |
+
[default3]:[rank3]: return self._call_impl(*args, **kwargs)
|
| 408 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 409 |
+
[default3]:[rank3]: return forward_call(*args, **kwargs)
|
| 410 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/nn/layer_norm.py", line 42, in forward
|
| 411 |
+
[default3]:[rank3]: return layer_norm_fn(
|
| 412 |
+
[default3]:[rank3]: 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
|
| 413 |
+
[default3]:[rank3]: return LayerNormFn.apply(
|
| 414 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 415 |
+
[default3]:[rank3]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 416 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 748, in forward
|
| 417 |
+
[default3]:[rank3]: y, y1, mean, rstd, residual_out, seeds, dropout_mask, dropout_mask1 = _layer_norm_fwd(
|
| 418 |
+
[default3]:[rank3]: 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
|
| 419 |
+
[default3]:[rank3]: _layer_norm_fwd_1pass_kernel[(M,)](
|
| 420 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 421 |
+
[default3]:[rank3]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 422 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in run
|
| 423 |
+
[default3]:[rank3]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 424 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in <dictcomp>
|
| 425 |
+
[default3]:[rank3]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 426 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 122, in _bench
|
| 427 |
+
[default3]:[rank3]: return do_bench(kernel_call, warmup=self.warmup, rep=self.rep, quantiles=(0.5, 0.2, 0.8))
|
| 428 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/testing.py", line 102, in do_bench
|
| 429 |
+
[default3]:[rank3]: fn()
|
| 430 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 110, in kernel_call
|
| 431 |
+
[default3]:[rank3]: self.fn.run(
|
| 432 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 433 |
+
[default3]:[rank3]: return self.fn.run(*args, **kwargs)
|
| 434 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 435 |
+
[default3]:[rank3]: return self.fn.run(*args, **kwargs)
|
| 436 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 437 |
+
[default3]:[rank3]: return self.fn.run(*args, **kwargs)
|
| 438 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 439 |
+
[default3]:[rank3]: self.cache[device][key] = compile(
|
| 440 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 441 |
+
[default3]:[rank3]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 442 |
+
[default3]:[rank3]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 443 |
+
[default3]:[rank3]: with open(temp_path, mode) as f:
|
| 444 |
+
[default3]:[rank3]: OSError: [Errno 122] Disk quota exceeded
|
| 445 |
+
[default0]:[rank8]: OSError: [Errno 122] Disk quota exceeded
|
| 446 |
+
[default0]:
|
| 447 |
+
[default0]:[rank8]: During handling of the above exception, another exception occurred:
|
| 448 |
+
[default0]:
|
| 449 |
+
[default0]:[rank8]: Traceback (most recent call last):
|
| 450 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 451 |
+
[default0]:[rank8]: trainer.train(dataloader)
|
| 452 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 453 |
+
[default0]:[rank8]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 454 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 455 |
+
[default0]:[rank8]: outputs = self.pipeline_engine.train_batch_iter(
|
| 456 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 457 |
+
[default0]:[rank8]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 458 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 459 |
+
[default0]:[rank8]: output = model(**micro_batch)
|
| 460 |
+
[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
|
| 461 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 462 |
+
[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
|
| 463 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 464 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 465 |
+
[default0]:[rank8]: sharded_logits = self.model(
|
| 466 |
+
[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
|
| 467 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 468 |
+
[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
|
| 469 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 470 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 471 |
+
[default0]:[rank8]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 472 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 473 |
+
[default0]:[rank8]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 474 |
+
[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
|
| 475 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 476 |
+
[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
|
| 477 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 478 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 479 |
+
[default0]:[rank8]: output = self.pp_block(**new_kwargs)
|
| 480 |
+
[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
|
| 481 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 482 |
+
[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
|
| 483 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 484 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 631, in forward
|
| 485 |
+
[default0]:[rank8]: output = self.attn(hidden_states=hidden_states, sequence_mask=sequence_mask)
|
| 486 |
+
[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
|
| 487 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 488 |
+
[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
|
| 489 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 490 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 566, in forward
|
| 491 |
+
[default0]:[rank8]: query_states, key_value_states = self.flash_rotary_embedding(query_states, kv=key_value_states)
|
| 492 |
+
[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
|
| 493 |
+
[default0]:[rank8]: return self._call_impl(*args, **kwargs)
|
| 494 |
+
[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
|
| 495 |
+
[default0]:[rank8]: return forward_call(*args, **kwargs)
|
| 496 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 457, in forward
|
| 497 |
+
[default0]:[rank8]: q = apply_rotary_emb_func(
|
| 498 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 122, in apply_rotary_emb
|
| 499 |
+
[default0]:[rank8]: return ApplyRotaryEmb.apply(
|
| 500 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 501 |
+
[default0]:[rank8]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 502 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 48, in forward
|
| 503 |
+
[default0]:[rank8]: out = apply_rotary(
|
| 504 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/rotary.py", line 202, in apply_rotary
|
| 505 |
+
[default0]:[rank8]: rotary_kernel[grid](
|
| 506 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 507 |
+
[default0]:[rank8]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 508 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 509 |
+
[default0]:[rank8]: self.cache[device][key] = compile(
|
| 510 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 511 |
+
[default0]:[rank8]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 512 |
+
[default0]:[rank8]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 513 |
+
[default0]:[rank8]: with open(temp_path, mode) as f:
|
| 514 |
+
[default0]:[rank8]: OSError: [Errno 122] Disk quota exceeded
|
| 515 |
+
[default7]:[rank15]: OSError: [Errno 122] Disk quota exceeded
|
| 516 |
+
[default7]:
|
| 517 |
+
[default7]:[rank15]: During handling of the above exception, another exception occurred:
|
| 518 |
+
[default7]:
|
| 519 |
+
[default7]:[rank15]: Traceback (most recent call last):
|
| 520 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 521 |
+
[default6]:[rank14]: OSError: [Errno 122] Disk quota exceeded
|
| 522 |
+
[default6]:
|
| 523 |
+
[default7]:[rank15]: trainer.train(dataloader)
|
| 524 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 525 |
+
[default7]:[rank15]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 526 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 527 |
+
[default7]:[rank15]: outputs = self.pipeline_engine.train_batch_iter(
|
| 528 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 529 |
+
[default7]:[rank15]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 530 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 531 |
+
[default7]:[rank15]: output = model(**micro_batch)
|
| 532 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 533 |
+
[default7]:[rank15]: return self._call_impl(*args, **kwargs)
|
| 534 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 535 |
+
[default7]:[rank15]: return forward_call(*args, **kwargs)
|
| 536 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 537 |
+
[default7]:[rank15]: sharded_logits = self.model(
|
| 538 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 539 |
+
[default7]:[rank15]: return self._call_impl(*args, **kwargs)
|
| 540 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 541 |
+
[default7]:[rank15]: return forward_call(*args, **kwargs)
|
| 542 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 543 |
+
[default7]:[rank15]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 544 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 545 |
+
[default7]:[rank15]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 546 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 547 |
+
[default6]:[rank14]: During handling of the above exception, another exception occurred:
|
| 548 |
+
[default6]:
|
| 549 |
+
[default6]:[rank14]: Traceback (most recent call last):
|
| 550 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 551 |
+
[default6]:[rank14]: trainer.train(dataloader)
|
| 552 |
+
[default7]:[rank15]: return self._call_impl(*args, **kwargs)
|
| 553 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 554 |
+
[default7]:[rank15]: return forward_call(*args, **kwargs)
|
| 555 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 556 |
+
[default7]:[rank15]: output = self.pp_block(**new_kwargs)
|
| 557 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 558 |
+
[default7]:[rank15]: return self._call_impl(*args, **kwargs)
|
| 559 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 560 |
+
[default7]:[rank15]: return forward_call(*args, **kwargs)
|
| 561 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 631, in forward
|
| 562 |
+
[default7]:[rank15]: output = self.attn(hidden_states=hidden_states, sequence_mask=sequence_mask)
|
| 563 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 564 |
+
[default7]:[rank15]: return self._call_impl(*args, **kwargs)
|
| 565 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 566 |
+
[default7]:[rank15]: return forward_call(*args, **kwargs)
|
| 567 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 566, in forward
|
| 568 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 569 |
+
[default6]:[rank14]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 570 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 571 |
+
[default7]:[rank15]: query_states, key_value_states = self.flash_rotary_embedding(query_states, kv=key_value_states)
|
| 572 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 573 |
+
[default6]:[rank14]: outputs = self.pipeline_engine.train_batch_iter(
|
| 574 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 575 |
+
[default6]:[rank14]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 576 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 577 |
+
[default6]:[rank14]: output = model(**micro_batch)
|
| 578 |
+
[default7]:[rank15]: return self._call_impl(*args, **kwargs)
|
| 579 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 580 |
+
[default6]:[rank14]: return self._call_impl(*args, **kwargs)
|
| 581 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 582 |
+
[default6]:[rank14]: return forward_call(*args, **kwargs)
|
| 583 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 584 |
+
[default6]:[rank14]: sharded_logits = self.model(
|
| 585 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 586 |
+
[default6]:[rank14]: return self._call_impl(*args, **kwargs)
|
| 587 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 588 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 589 |
+
[default7]:[rank15]: return forward_call(*args, **kwargs)
|
| 590 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 457, in forward
|
| 591 |
+
[default6]:[rank14]: return forward_call(*args, **kwargs)
|
| 592 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 593 |
+
[default6]:[rank14]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 594 |
+
[default7]:[rank15]: q = apply_rotary_emb_func(
|
| 595 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 122, in apply_rotary_emb
|
| 596 |
+
[default7]:[rank15]: return ApplyRotaryEmb.apply(
|
| 597 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 598 |
+
[default7]:[rank15]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 599 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 600 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 48, in forward
|
| 601 |
+
[default7]:[rank15]: out = apply_rotary(
|
| 602 |
+
[default6]:[rank14]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 603 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 604 |
+
[default6]:[rank14]: return self._call_impl(*args, **kwargs)
|
| 605 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 606 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/rotary.py", line 202, in apply_rotary
|
| 607 |
+
[default7]:[rank15]: rotary_kernel[grid](
|
| 608 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 609 |
+
[default6]:[rank14]: return forward_call(*args, **kwargs)
|
| 610 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 611 |
+
[default6]:[rank14]: output = self.pp_block(**new_kwargs)
|
| 612 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 613 |
+
[default6]:[rank14]: return self._call_impl(*args, **kwargs)
|
| 614 |
+
[default7]:[rank15]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 615 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 616 |
+
[default7]:[rank15]: self.cache[device][key] = compile(
|
| 617 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 618 |
+
[default7]:[rank15]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 619 |
+
[default7]:[rank15]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 620 |
+
[default7]:[rank15]: with open(temp_path, mode) as f:
|
| 621 |
+
[default7]:[rank15]: OSError: [Errno 122] Disk quota exceeded
|
| 622 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 623 |
+
[default6]:[rank14]: return forward_call(*args, **kwargs)
|
| 624 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 631, in forward
|
| 625 |
+
[default6]:[rank14]: output = self.attn(hidden_states=hidden_states, sequence_mask=sequence_mask)
|
| 626 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 627 |
+
[default6]:[rank14]: return self._call_impl(*args, **kwargs)
|
| 628 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 629 |
+
[default6]:[rank14]: return forward_call(*args, **kwargs)
|
| 630 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 566, in forward
|
| 631 |
+
[default6]:[rank14]: query_states, key_value_states = self.flash_rotary_embedding(query_states, kv=key_value_states)
|
| 632 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 633 |
+
[default6]:[rank14]: return self._call_impl(*args, **kwargs)
|
| 634 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 635 |
+
[default6]:[rank14]: return forward_call(*args, **kwargs)
|
| 636 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 457, in forward
|
| 637 |
+
[default6]:[rank14]: q = apply_rotary_emb_func(
|
| 638 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 122, in apply_rotary_emb
|
| 639 |
+
[default6]:[rank14]: return ApplyRotaryEmb.apply(
|
| 640 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 641 |
+
[default6]:[rank14]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 642 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 48, in forward
|
| 643 |
+
[default6]:[rank14]: out = apply_rotary(
|
| 644 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/rotary.py", line 202, in apply_rotary
|
| 645 |
+
[default6]:[rank14]: rotary_kernel[grid](
|
| 646 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 647 |
+
[default6]:[rank14]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 648 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 649 |
+
[default6]:[rank14]: self.cache[device][key] = compile(
|
| 650 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 651 |
+
[default6]:[rank14]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 652 |
+
[default6]:[rank14]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 653 |
+
[default6]:[rank14]: with open(temp_path, mode) as f:
|
| 654 |
+
[default6]:[rank14]: OSError: [Errno 122] Disk quota exceeded
|
| 655 |
+
[default3]:[rank11]: OSError: [Errno 122] Disk quota exceeded
|
| 656 |
+
[default3]:
|
| 657 |
+
[default3]:[rank11]: During handling of the above exception, another exception occurred:
|
| 658 |
+
[default3]:
|
| 659 |
+
[default3]:[rank11]: Traceback (most recent call last):
|
| 660 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 661 |
+
[default3]:[rank11]: trainer.train(dataloader)
|
| 662 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 663 |
+
[default3]:[rank11]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 664 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 665 |
+
[default3]:[rank11]: outputs = self.pipeline_engine.train_batch_iter(
|
| 666 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 295, in train_batch_iter
|
| 667 |
+
[default3]:[rank11]: self.backward(context=context, state=state, grad_accumulator=grad_accumulator)
|
| 668 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 86, in backward
|
| 669 |
+
[default3]:[rank11]: grad_accumulator.backward(sum(activations))
|
| 670 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/optim/gradient_accumulator.py", line 205, in backward
|
| 671 |
+
[default3]:[rank11]: result = loss.backward()
|
| 672 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/_tensor.py", line 525, in backward
|
| 673 |
+
[default3]:[rank11]: torch.autograd.backward(
|
| 674 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/__init__.py", line 267, in backward
|
| 675 |
+
[default3]:[rank11]: _engine_run_backward(
|
| 676 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py", line 744, in _engine_run_backward
|
| 677 |
+
[default3]:[rank11]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 678 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 301, in apply
|
| 679 |
+
[default3]:[rank11]: return user_fn(self, *args)
|
| 680 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 821, in backward
|
| 681 |
+
[default3]:[rank11]: dx, dw, db, dresidual_in, dx1, dw1, db1 = _layer_norm_bwd(
|
| 682 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/layer_norm.py", line 643, in _layer_norm_bwd
|
| 683 |
+
[default3]:[rank11]: _layer_norm_bwd_kernel[grid](
|
| 684 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 685 |
+
[default3]:[rank11]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 686 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in run
|
| 687 |
+
[default3]:[rank11]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 688 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 143, in <dictcomp>
|
| 689 |
+
[default3]:[rank11]: timings = {config: self._bench(*args, config=config, **kwargs) for config in pruned_configs}
|
| 690 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 122, in _bench
|
| 691 |
+
[default3]:[rank11]: return do_bench(kernel_call, warmup=self.warmup, rep=self.rep, quantiles=(0.5, 0.2, 0.8))
|
| 692 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/testing.py", line 102, in do_bench
|
| 693 |
+
[default3]:[rank11]: fn()
|
| 694 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 110, in kernel_call
|
| 695 |
+
[default3]:[rank11]: self.fn.run(
|
| 696 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 697 |
+
[default3]:[rank11]: return self.fn.run(*args, **kwargs)
|
| 698 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 699 |
+
[default3]:[rank11]: return self.fn.run(*args, **kwargs)
|
| 700 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/autotuner.py", line 305, in run
|
| 701 |
+
[default3]:[rank11]: return self.fn.run(*args, **kwargs)
|
| 702 |
+
[default3]:[rank11]: [Previous line repeated 2 more times]
|
| 703 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 704 |
+
[default3]:[rank11]: self.cache[device][key] = compile(
|
| 705 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 706 |
+
[default3]:[rank11]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 707 |
+
[default3]:[rank11]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 708 |
+
[default3]:[rank11]: with open(temp_path, mode) as f:
|
| 709 |
+
[default3]:[rank11]: OSError: [Errno 122] Disk quota exceeded
|
| 710 |
+
[default4]:/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py:744: UserWarning: c10d::allreduce_: an autograd kernel was not registered to the Autograd key(s) but we are trying to backprop through it. This may lead to silently incorrect behavior. This behavior is deprecated and will be removed in a future version of PyTorch. If your operator is differentiable, please ensure you have registered an autograd kernel to the correct Autograd key (e.g. DispatchKey::Autograd, DispatchKey::CompositeImplicitAutograd). If your operator is not differentiable, or to squash this warning and use the previous behavior, please register torch::CppFunction::makeFallthrough() to DispatchKey::Autograd. (Triggered internally at ../torch/csrc/autograd/autograd_not_implemented_fallback.cpp:63.)
|
| 711 |
+
[default4]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 712 |
+
[default4]:[rank12]: OSError: [Errno 122] Disk quota exceeded
|
| 713 |
+
[default4]:
|
| 714 |
+
[default4]:[rank12]: During handling of the above exception, another exception occurred:
|
| 715 |
+
[default4]:
|
| 716 |
+
[default4]:[rank12]: Traceback (most recent call last):
|
| 717 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 718 |
+
[default4]:[rank12]: trainer.train(dataloader)
|
| 719 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 720 |
+
[default4]:[rank12]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 721 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 722 |
+
[default4]:[rank12]: outputs = self.pipeline_engine.train_batch_iter(
|
| 723 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 295, in train_batch_iter
|
| 724 |
+
[default4]:[rank12]: self.backward(context=context, state=state, grad_accumulator=grad_accumulator)
|
| 725 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 86, in backward
|
| 726 |
+
[default4]:[rank12]: grad_accumulator.backward(sum(activations))
|
| 727 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/optim/gradient_accumulator.py", line 205, in backward
|
| 728 |
+
[default4]:[rank12]: result = loss.backward()
|
| 729 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/_tensor.py", line 525, in backward
|
| 730 |
+
[default4]:[rank12]: torch.autograd.backward(
|
| 731 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/__init__.py", line 267, in backward
|
| 732 |
+
[default4]:[rank12]: _engine_run_backward(
|
| 733 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/graph.py", line 744, in _engine_run_backward
|
| 734 |
+
[default4]:[rank12]: return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
|
| 735 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 301, in apply
|
| 736 |
+
[default4]:[rank12]: return user_fn(self, *args)
|
| 737 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 261, in backward
|
| 738 |
+
[default4]:[rank12]: apply_rotary(
|
| 739 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/rotary.py", line 202, in apply_rotary
|
| 740 |
+
[default4]:[rank12]: rotary_kernel[grid](
|
| 741 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 742 |
+
[default4]:[rank12]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 743 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 744 |
+
[default4]:[rank12]: self.cache[device][key] = compile(
|
| 745 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 746 |
+
[default4]:[rank12]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 747 |
+
[default4]:[rank12]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 748 |
+
[default4]:[rank12]: with open(temp_path, mode) as f:
|
| 749 |
+
[default4]:[rank12]: OSError: [Errno 122] Disk quota exceeded
|
| 750 |
+
[default6]:[rank6]: OSError: [Errno 122] Disk quota exceeded
|
| 751 |
+
[default6]:
|
| 752 |
+
[default6]:[rank6]: During handling of the above exception, another exception occurred:
|
| 753 |
+
[default6]:
|
| 754 |
+
[default6]:[rank6]: Traceback (most recent call last):
|
| 755 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py", line 237, in <module>
|
| 756 |
+
[default6]:[rank6]: trainer.train(dataloader)
|
| 757 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 429, in train
|
| 758 |
+
[default6]:[rank6]: outputs, loss_avg = self.training_step(dataloader=self.current_dataloader)
|
| 759 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/trainer.py", line 462, in training_step
|
| 760 |
+
[default6]:[rank6]: outputs = self.pipeline_engine.train_batch_iter(
|
| 761 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 278, in train_batch_iter
|
| 762 |
+
[default6]:[rank6]: output = self.forward(context=context, state=state, micro_batch=micro_batch, model=model)
|
| 763 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/engine.py", line 44, in forward
|
| 764 |
+
[default6]:[rank6]: output = model(**micro_batch)
|
| 765 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 766 |
+
[default6]:[rank6]: return self._call_impl(*args, **kwargs)
|
| 767 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 768 |
+
[default6]:[rank6]: return forward_call(*args, **kwargs)
|
| 769 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 891, in forward
|
| 770 |
+
[default6]:[rank6]: sharded_logits = self.model(
|
| 771 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 772 |
+
[default6]:[rank6]: return self._call_impl(*args, **kwargs)
|
| 773 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 774 |
+
[default6]:[rank6]: return forward_call(*args, **kwargs)
|
| 775 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 764, in forward
|
| 776 |
+
[default6]:[rank6]: return self.forward_with_hidden_states(input_ids=input_ids, input_mask=input_mask)[0]
|
| 777 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 780, in forward_with_hidden_states
|
| 778 |
+
[default6]:[rank6]: hidden_encoder_states = encoder_block(**hidden_encoder_states)
|
| 779 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 780 |
+
[default6]:[rank6]: return self._call_impl(*args, **kwargs)
|
| 781 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 782 |
+
[default6]:[rank6]: return forward_call(*args, **kwargs)
|
| 783 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/parallel/pipeline_parallel/block.py", line 151, in forward
|
| 784 |
+
[default6]:[rank6]: output = self.pp_block(**new_kwargs)
|
| 785 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 786 |
+
[default6]:[rank6]: return self._call_impl(*args, **kwargs)
|
| 787 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 788 |
+
[default6]:[rank6]: return forward_call(*args, **kwargs)
|
| 789 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 631, in forward
|
| 790 |
+
[default6]:[rank6]: output = self.attn(hidden_states=hidden_states, sequence_mask=sequence_mask)
|
| 791 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 792 |
+
[default6]:[rank6]: return self._call_impl(*args, **kwargs)
|
| 793 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 794 |
+
[default6]:[rank6]: return forward_call(*args, **kwargs)
|
| 795 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/src/nanotron/models/llama.py", line 566, in forward
|
| 796 |
+
[default6]:[rank6]: query_states, key_value_states = self.flash_rotary_embedding(query_states, kv=key_value_states)
|
| 797 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1532, in _wrapped_call_impl
|
| 798 |
+
[default6]:[rank6]: return self._call_impl(*args, **kwargs)
|
| 799 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1541, in _call_impl
|
| 800 |
+
[default6]:[rank6]: return forward_call(*args, **kwargs)
|
| 801 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 457, in forward
|
| 802 |
+
[default6]:[rank6]: q = apply_rotary_emb_func(
|
| 803 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 122, in apply_rotary_emb
|
| 804 |
+
[default6]:[rank6]: return ApplyRotaryEmb.apply(
|
| 805 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/autograd/function.py", line 598, in apply
|
| 806 |
+
[default6]:[rank6]: return super().apply(*args, **kwargs) # type: ignore[misc]
|
| 807 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/layers/rotary.py", line 48, in forward
|
| 808 |
+
[default6]:[rank6]: out = apply_rotary(
|
| 809 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/flash_attn/ops/triton/rotary.py", line 202, in apply_rotary
|
| 810 |
+
[default6]:[rank6]: rotary_kernel[grid](
|
| 811 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 167, in <lambda>
|
| 812 |
+
[default6]:[rank6]: return lambda *args, **kwargs: self.run(grid=grid, warmup=False, *args, **kwargs)
|
| 813 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/jit.py", line 416, in run
|
| 814 |
+
[default6]:[rank6]: self.cache[device][key] = compile(
|
| 815 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/compiler/compiler.py", line 194, in compile
|
| 816 |
+
[default6]:[rank6]: metadata_group[f"{src.name}.{ext}"] = fn_cache_manager.put(next_module, f"{src.name}.{ext}")
|
| 817 |
+
[default6]:[rank6]: File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/triton/runtime/cache.py", line 123, in put
|
| 818 |
+
[default6]:[rank6]: with open(temp_path, mode) as f:
|
| 819 |
+
[default6]:[rank6]: OSError: [Errno 122] Disk quota exceeded
|
| 820 |
+
W0702 16:32:10.570000 139668155955008 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 3409968 closing signal SIGTERM
|
| 821 |
+
W0702 16:32:10.574000 139668155955008 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 3409969 closing signal SIGTERM
|
| 822 |
+
W0702 16:32:10.576000 140446723327808 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 403560 closing signal SIGTERM
|
| 823 |
+
W0702 16:32:10.580000 140446723327808 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 403561 closing signal SIGTERM
|
| 824 |
+
W0702 16:32:10.581000 139668155955008 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 3409971 closing signal SIGTERM
|
| 825 |
+
W0702 16:32:10.586000 140446723327808 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 403564 closing signal SIGTERM
|
| 826 |
+
W0702 16:32:10.602000 139668155955008 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 3409973 closing signal SIGTERM
|
| 827 |
+
W0702 16:32:10.603000 139668155955008 torch/distributed/elastic/multiprocessing/api.py:851] Sending process 3409974 closing signal SIGTERM
|
| 828 |
+
E0702 16:32:12.295000 140446723327808 torch/distributed/elastic/multiprocessing/api.py:826] failed (exitcode: 1) local_rank: 0 (pid: 403559) of binary: /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/python3.10
|
| 829 |
+
Traceback (most recent call last):
|
| 830 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/torchrun", line 8, in <module>
|
| 831 |
+
sys.exit(main())
|
| 832 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 347, in wrapper
|
| 833 |
+
return f(*args, **kwargs)
|
| 834 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 879, in main
|
| 835 |
+
run(args)
|
| 836 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 870, in run
|
| 837 |
+
elastic_launch(
|
| 838 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
|
| 839 |
+
return launch_agent(self._config, self._entrypoint, list(args))
|
| 840 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 263, in launch_agent
|
| 841 |
+
raise ChildFailedError(
|
| 842 |
+
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
|
| 843 |
+
============================================================
|
| 844 |
+
/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py FAILED
|
| 845 |
+
------------------------------------------------------------
|
| 846 |
+
Failures:
|
| 847 |
+
[1]:
|
| 848 |
+
time : 2024-07-02_16:32:10
|
| 849 |
+
host : ip-26-0-171-88.ec2.internal
|
| 850 |
+
rank : 11 (local_rank: 3)
|
| 851 |
+
exitcode : 1 (pid: 403562)
|
| 852 |
+
error_file: <N/A>
|
| 853 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 854 |
+
[2]:
|
| 855 |
+
time : 2024-07-02_16:32:10
|
| 856 |
+
host : ip-26-0-171-88.ec2.internal
|
| 857 |
+
rank : 12 (local_rank: 4)
|
| 858 |
+
exitcode : 1 (pid: 403563)
|
| 859 |
+
error_file: <N/A>
|
| 860 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 861 |
+
[3]:
|
| 862 |
+
time : 2024-07-02_16:32:10
|
| 863 |
+
host : ip-26-0-171-88.ec2.internal
|
| 864 |
+
rank : 14 (local_rank: 6)
|
| 865 |
+
exitcode : 1 (pid: 403565)
|
| 866 |
+
error_file: <N/A>
|
| 867 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 868 |
+
[4]:
|
| 869 |
+
time : 2024-07-02_16:32:10
|
| 870 |
+
host : ip-26-0-171-88.ec2.internal
|
| 871 |
+
rank : 15 (local_rank: 7)
|
| 872 |
+
exitcode : 1 (pid: 403566)
|
| 873 |
+
error_file: <N/A>
|
| 874 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 875 |
+
------------------------------------------------------------
|
| 876 |
+
Root Cause (first observed failure):
|
| 877 |
+
[0]:
|
| 878 |
+
time : 2024-07-02_16:32:10
|
| 879 |
+
host : ip-26-0-171-88.ec2.internal
|
| 880 |
+
rank : 8 (local_rank: 0)
|
| 881 |
+
exitcode : 1 (pid: 403559)
|
| 882 |
+
error_file: <N/A>
|
| 883 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 884 |
+
============================================================
|
| 885 |
+
E0702 16:32:12.506000 139668155955008 torch/distributed/elastic/multiprocessing/api.py:826] failed (exitcode: 1) local_rank: 0 (pid: 3409967) of binary: /fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/python3.10
|
| 886 |
+
Traceback (most recent call last):
|
| 887 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/bin/torchrun", line 8, in <module>
|
| 888 |
+
sys.exit(main())
|
| 889 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/elastic/multiprocessing/errors/__init__.py", line 347, in wrapper
|
| 890 |
+
return f(*args, **kwargs)
|
| 891 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 879, in main
|
| 892 |
+
run(args)
|
| 893 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/run.py", line 870, in run
|
| 894 |
+
elastic_launch(
|
| 895 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 132, in __call__
|
| 896 |
+
return launch_agent(self._config, self._entrypoint, list(args))
|
| 897 |
+
File "/fsx/ferdinandmom/miniforge3/envs/env-bench-cluster/lib/python3.10/site-packages/torch/distributed/launcher/api.py", line 263, in launch_agent
|
| 898 |
+
raise ChildFailedError(
|
| 899 |
+
torch.distributed.elastic.multiprocessing.errors.ChildFailedError:
|
| 900 |
+
============================================================
|
| 901 |
+
/fsx/ferdinandmom/ferdinand-hf/bench_cluster/nanotron/run_train.py FAILED
|
| 902 |
+
------------------------------------------------------------
|
| 903 |
+
Failures:
|
| 904 |
+
[1]:
|
| 905 |
+
time : 2024-07-02_16:32:10
|
| 906 |
+
host : ip-26-0-171-62.ec2.internal
|
| 907 |
+
rank : 3 (local_rank: 3)
|
| 908 |
+
exitcode : 1 (pid: 3409970)
|
| 909 |
+
error_file: <N/A>
|
| 910 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 911 |
+
[2]:
|
| 912 |
+
time : 2024-07-02_16:32:10
|
| 913 |
+
host : ip-26-0-171-62.ec2.internal
|
| 914 |
+
rank : 5 (local_rank: 5)
|
| 915 |
+
exitcode : 1 (pid: 3409972)
|
| 916 |
+
error_file: <N/A>
|
| 917 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 918 |
+
------------------------------------------------------------
|
| 919 |
+
Root Cause (first observed failure):
|
| 920 |
+
[0]:
|
| 921 |
+
time : 2024-07-02_16:32:10
|
| 922 |
+
host : ip-26-0-171-62.ec2.internal
|
| 923 |
+
rank : 0 (local_rank: 0)
|
| 924 |
+
exitcode : 1 (pid: 3409967)
|
| 925 |
+
error_file: <N/A>
|
| 926 |
+
traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
|
| 927 |
+
============================================================
|
| 928 |
+
srun: error: ip-26-0-171-88: task 1: Exited with exit code 1
|
| 929 |
+
srun: error: ip-26-0-171-62: task 0: Exited with exit code 1
|
| 930 |
+
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-8_tp-2_pp-1_mbz-4/status.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
fail
|