[2025-04-13 14:56:48,290] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:50,327] [WARNING] [runner.py:215:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only.
Detected VISIBLE_DEVICES=0,1,2,3,4,5,6,7: setting --include=localhost:0,1,2,3,4,5,6,7
[2025-04-13 14:56:50,327] [INFO] [runner.py:605:main] cmd = /usr/bin/python3 -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgMywgNCwgNSwgNiwgN119 --master_addr=127.0.0.1 --master_port=29500 --enable_each_rank_log=None train.py --deepspeed scripts/newzero3.json --seed 42 --model_name_or_path /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct --train_tokenized_file /home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl --output_dir /home/stern/GRPO/offline_rl_v2/output --per_device_train_batch_size 1 --gradient_accumulation_steps 4 --evaluation_strategy no --save_strategy no --learning_rate 2e-6 --lr_scheduler_type cosine --save_only_model True --remove_unused_columns False --warmup_ratio 0.03 --num_train_epochs 4 --logging_steps 1 --report_to tensorboard --gradient_checkpointing True --overwrite_output_dir --bf16 True
[2025-04-13 14:56:51,788] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:53,706] [INFO] [launch.py:146:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]}
[2025-04-13 14:56:53,707] [INFO] [launch.py:152:main] nnodes=1, num_local_procs=8, node_rank=0
[2025-04-13 14:56:53,707] [INFO] [launch.py:163:main] global_rank_mapping=defaultdict(<class 'list'>, {'localhost': [0, 1, 2, 3, 4, 5, 6, 7]})
[2025-04-13 14:56:53,707] [INFO] [launch.py:164:main] dist_world_size=8
[2025-04-13 14:56:53,707] [INFO] [launch.py:168:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
[2025-04-13 14:56:53,707] [INFO] [launch.py:256:main] process 1025161 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=0', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '4', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-13 14:56:53,708] [INFO] [launch.py:256:main] process 1025162 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=1', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '4', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-13 14:56:53,708] [INFO] [launch.py:256:main] process 1025163 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=2', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '4', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-13 14:56:53,708] [INFO] [launch.py:256:main] process 1025164 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=3', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '4', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-13 14:56:53,709] [INFO] [launch.py:256:main] process 1025165 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=4', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '4', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-13 14:56:53,709] [INFO] [launch.py:256:main] process 1025166 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=5', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '4', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-13 14:56:53,710] [INFO] [launch.py:256:main] process 1025167 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=6', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '4', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-13 14:56:53,710] [INFO] [launch.py:256:main] process 1025168 spawned with command: ['/usr/bin/python3', '-u', 'train.py', '--local_rank=7', '--deepspeed', 'scripts/newzero3.json', '--seed', '42', '--model_name_or_path', '/home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct', '--train_tokenized_file', '/home/stern/GRPO/offline_rl_v2/data/32K_neg_tokenized.jsonl', '--output_dir', '/home/stern/GRPO/offline_rl_v2/output', '--per_device_train_batch_size', '1', '--gradient_accumulation_steps', '4', '--evaluation_strategy', 'no', '--save_strategy', 'no', '--learning_rate', '2e-6', '--lr_scheduler_type', 'cosine', '--save_only_model', 'True', '--remove_unused_columns', 'False', '--warmup_ratio', '0.03', '--num_train_epochs', '4', '--logging_steps', '1', '--report_to', 'tensorboard', '--gradient_checkpointing', 'True', '--overwrite_output_dir', '--bf16', 'True']
[2025-04-13 14:56:57,613] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:58,352] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:58,555] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:58,618] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:58,666] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:58,692] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:58,696] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
[2025-04-13 14:56:58,702] [INFO] [real_accelerator.py:239:get_accelerator] Setting ds_accelerator to cuda (auto detect)
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
  warnings.warn(
[2025-04-13 14:57:00,288] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
  warnings.warn(
[2025-04-13 14:57:00,342] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
  warnings.warn(
[2025-04-13 14:57:00,660] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
  warnings.warn(
[2025-04-13 14:57:00,718] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
  warnings.warn(
[2025-04-13 14:57:00,764] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
  warnings.warn(
[2025-04-13 14:57:00,773] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
  warnings.warn(
[2025-04-13 14:57:00,778] [INFO] [comm.py:658:init_distributed] cdb=None
/home/stern/.local/lib/python3.10/site-packages/transformers/training_args.py:1611: FutureWarning: `evaluation_strategy` is deprecated and will be removed in version 4.46 of 🤗 Transformers. Use `eval_strategy` instead
  warnings.warn(
[2025-04-13 14:57:00,814] [INFO] [comm.py:658:init_distributed] cdb=None
[2025-04-13 14:57:00,815] [INFO] [comm.py:689:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl
WARNING:__main__:Process rank: 7, device: cuda:7, n_gpu: 1
WARNING:__main__:Process rank: 1, device: cuda:1, n_gpu: 1
[2025-04-13 14:57:02,243] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-13 14:57:02,246 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-13 14:57:02,258] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-13 14:57:02,260 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
WARNING:__main__:Process rank: 0, device: cuda:0, n_gpu: 1
INFO:__main__:Training parameters CustomTrainingArguments(
_n_gpu=1,
accelerator_config={'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None, 'use_configured_state': False},
adafactor=False,
adam_beta1=0.9,
adam_beta2=0.999,
adam_epsilon=1e-08,
auto_find_batch_size=False,
average_tokens_across_devices=False,
batch_eval_metrics=False,
bf16=True,
bf16_full_eval=False,
data_seed=None,
dataloader_drop_last=False,
dataloader_num_workers=0,
dataloader_persistent_workers=False,
dataloader_pin_memory=True,
dataloader_prefetch_factor=None,
ddp_backend=None,
ddp_broadcast_buffers=None,
ddp_bucket_cap_mb=None,
ddp_find_unused_parameters=None,
ddp_timeout=1800,
debug=[],
deepspeed=scripts/newzero3.json,
disable_tqdm=False,
dispatch_batches=None,
do_eval=False,
do_predict=False,
do_train=False,
eval_accumulation_steps=None,
eval_delay=0,
eval_do_concat_batches=True,
eval_on_start=False,
eval_steps=None,
eval_strategy=no,
eval_use_gather_object=False,
evaluation_strategy=no,
fp16=False,
fp16_backend=auto,
fp16_full_eval=False,
fp16_opt_level=O1,
fsdp=[],
fsdp_config={'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False},
fsdp_min_num_params=0,
fsdp_transformer_layer_cls_to_wrap=None,
full_determinism=False,
gradient_accumulation_steps=4,
gradient_checkpointing=True,
gradient_checkpointing_kwargs=None,
greater_is_better=None,
group_by_length=False,
half_precision_backend=auto,
hub_always_push=False,
hub_model_id=None,
hub_private_repo=None,
hub_strategy=every_save,
hub_token=<HUB_TOKEN>,
ignore_data_skip=False,
include_for_metrics=[],
include_inputs_for_metrics=False,
include_num_input_tokens_seen=False,
include_tokens_per_second=False,
jit_mode_eval=False,
kl_coeff=0.0,
label_names=None,
label_smoothing_factor=0.0,
learning_rate=2e-06,
length_column_name=length,
load_best_model_at_end=False,
local_rank=0,
log_level=passive,
log_level_replica=warning,
log_on_each_node=True,
logging_dir=/home/stern/GRPO/offline_rl_v2/output/runs/Apr13_14-57-00_nacamontrealdc1-p2r203n1.enovum.hivecloud.com,
logging_first_step=False,
logging_nan_inf_filter=True,
logging_steps=1.0,
logging_strategy=steps,
lr_scheduler_kwargs={},
lr_scheduler_type=cosine,
max_grad_norm=1.0,
max_steps=-1,
metric_for_best_model=None,
mp_parameters=,
neftune_noise_alpha=None,
no_cuda=False,
num_train_epochs=4.0,
optim=adamw_torch,
optim_args=None,
optim_target_modules=None,
output_dir=/home/stern/GRPO/offline_rl_v2/output,
overwrite_output_dir=True,
past_index=-1,
per_device_eval_batch_size=8,
per_device_train_batch_size=1,
prediction_loss_only=False,
push_to_hub=False,
push_to_hub_model_id=None,
push_to_hub_organization=None,
push_to_hub_token=<PUSH_TO_HUB_TOKEN>,
ray_scope=last,
remove_unused_columns=False,
report_to=['tensorboard'],
restore_callback_states_from_checkpoint=False,
resume_from_checkpoint=None,
run_name=/home/stern/GRPO/offline_rl_v2/output,
save_on_each_node=False,
save_only_model=True,
save_safetensors=True,
save_steps=500,
save_strategy=no,
save_total_limit=None,
seed=42,
skip_memory_metrics=True,
split_batches=None,
tf32=None,
torch_compile=False,
torch_compile_backend=None,
torch_compile_mode=None,
torch_empty_cache_steps=None,
torchdynamo=None,
tp_size=0,
tpu_metrics_debug=False,
tpu_num_cores=None,
use_cpu=False,
use_ipex=False,
use_legacy_prediction_loop=False,
use_liger_kernel=False,
use_mps_device=False,
warmup_ratio=0.03,
warmup_steps=0,
weight_decay=0.0,
)
[INFO|tokenization_utils_base.py:2058] 2025-04-13 14:57:02,409 >> loading file vocab.json
[INFO|tokenization_utils_base.py:2058] 2025-04-13 14:57:02,409 >> loading file merges.txt
[INFO|tokenization_utils_base.py:2058] 2025-04-13 14:57:02,409 >> loading file tokenizer.json
[INFO|tokenization_utils_base.py:2058] 2025-04-13 14:57:02,409 >> loading file added_tokens.json
[INFO|tokenization_utils_base.py:2058] 2025-04-13 14:57:02,409 >> loading file special_tokens_map.json
[INFO|tokenization_utils_base.py:2058] 2025-04-13 14:57:02,409 >> loading file tokenizer_config.json
[INFO|tokenization_utils_base.py:2058] 2025-04-13 14:57:02,409 >> loading file chat_template.jinja
WARNING:__main__:Process rank: 2, device: cuda:2, n_gpu: 1
WARNING:__main__:Process rank: 5, device: cuda:5, n_gpu: 1
WARNING:__main__:Process rank: 6, device: cuda:6, n_gpu: 1
WARNING:__main__:Process rank: 4, device: cuda:4, n_gpu: 1
WARNING:__main__:Process rank: 3, device: cuda:3, n_gpu: 1
[INFO|tokenization_utils_base.py:2323] 2025-04-13 14:57:02,684 >> Special tokens have been added in the vocabulary, make sure the associated word embeddings are fine-tuned or trained.
[INFO|configuration_utils.py:697] 2025-04-13 14:57:02,684 >> loading configuration file /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct/config.json
[INFO|configuration_utils.py:771] 2025-04-13 14:57:02,686 >> Model config Qwen2Config {
  "architectures": [
    "Qwen2ForCausalLM"
  ],
  "attention_dropout": 0.0,
  "bos_token_id": 151643,
  "eos_token_id": 151645,
  "hidden_act": "silu",
  "hidden_size": 5120,
  "initializer_range": 0.02,
  "intermediate_size": 27648,
  "max_position_embeddings": 32768,
  "max_window_layers": 70,
  "model_type": "qwen2",
  "num_attention_heads": 40,
  "num_hidden_layers": 64,
  "num_key_value_heads": 8,
  "rms_norm_eps": 1e-06,
  "rope_scaling": null,
  "rope_theta": 1000000.0,
  "sliding_window": 131072,
  "tie_word_embeddings": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.50.3",
  "use_cache": true,
  "use_sliding_window": false,
  "vocab_size": 152064
}

[INFO|modeling_utils.py:1151] 2025-04-13 14:57:02,716 >> loading weights file /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct/model.safetensors.index.json
[INFO|modeling_utils.py:1225] 2025-04-13 14:57:02,717 >> Will use torch_dtype=torch.bfloat16 as defined in model's config object
[INFO|modeling_utils.py:2170] 2025-04-13 14:57:02,717 >> Instantiating Qwen2ForCausalLM model under default dtype torch.bfloat16.
[INFO|modeling_utils.py:3747] 2025-04-13 14:57:02,717 >> Detected DeepSpeed ZeRO-3: activating zero.init() for this model
[2025-04-13 14:57:02,717] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-13 14:57:02,720 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[INFO|configuration_utils.py:1139] 2025-04-13 14:57:02,726 >> Generate config GenerationConfig {
  "bos_token_id": 151643,
  "eos_token_id": 151645
}

[2025-04-13 14:57:02,862] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-13 14:57:02,865 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-13 14:57:02,870] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-13 14:57:02,872 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-13 14:57:02,910] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-13 14:57:02,912 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-13 14:57:03,005] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-13 14:57:03,008 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-13 14:57:03,009] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
[WARNING|logging.py:329] 2025-04-13 14:57:03,011 >> You are attempting to use Flash Attention 2.0 with a model not initialized on GPU. Make sure to move the model to GPU after initializing it on CPU with `model.to('cuda')`.
[2025-04-13 14:57:24,786] [INFO] [partition_parameters.py:348:__exit__] finished initializing model - num_params = 771, num_elems = 32.76B

Loading checkpoint shards:   0%|          | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards:   0%|          | 0/17 [00:00<?, ?it/s]
Loading checkpoint shards:   6%|▌         | 1/17 [00:00<00:11,  1.42it/s]
Loading checkpoint shards:   6%|▌         | 1/17 [00:00<00:11,  1.33it/s]
Loading checkpoint shards:   6%|▌         | 1/17 [00:00<00:11,  1.35it/s]
Loading checkpoint shards:   6%|▌         | 1/17 [00:00<00:12,  1.31it/s]
Loading checkpoint shards:   6%|▌         | 1/17 [00:00<00:11,  1.37it/s]
Loading checkpoint shards:   6%|▌         | 1/17 [00:00<00:12,  1.24it/s]
Loading checkpoint shards:   6%|▌         | 1/17 [00:00<00:12,  1.29it/s]
Loading checkpoint shards:   6%|▌         | 1/17 [00:00<00:14,  1.13it/s]
Loading checkpoint shards:  12%|█▏        | 2/17 [00:01<00:11,  1.35it/s]
Loading checkpoint shards:  12%|█▏        | 2/17 [00:01<00:11,  1.30it/s]
Loading checkpoint shards:  12%|█▏        | 2/17 [00:01<00:11,  1.31it/s]
Loading checkpoint shards:  12%|█▏        | 2/17 [00:01<00:11,  1.28it/s]
Loading checkpoint shards:  12%|█▏        | 2/17 [00:01<00:11,  1.32it/s]
Loading checkpoint shards:  12%|█▏        | 2/17 [00:01<00:11,  1.33it/s]
Loading checkpoint shards:  12%|█▏        | 2/17 [00:01<00:11,  1.30it/s]
Loading checkpoint shards:  12%|█▏        | 2/17 [00:01<00:12,  1.22it/s]
Loading checkpoint shards:  18%|█▊        | 3/17 [00:02<00:10,  1.30it/s]
Loading checkpoint shards:  18%|█▊        | 3/17 [00:02<00:10,  1.31it/s]
Loading checkpoint shards:  18%|█▊        | 3/17 [00:02<00:10,  1.31it/s]
Loading checkpoint shards:  18%|█▊        | 3/17 [00:02<00:10,  1.32it/s]
Loading checkpoint shards:  18%|█▊        | 3/17 [00:02<00:10,  1.30it/s]
Loading checkpoint shards:  18%|█▊        | 3/17 [00:02<00:10,  1.29it/s]
Loading checkpoint shards:  18%|█▊        | 3/17 [00:02<00:10,  1.28it/s]
Loading checkpoint shards:  18%|█▊        | 3/17 [00:02<00:11,  1.25it/s]
Loading checkpoint shards:  24%|██▎       | 4/17 [00:03<00:10,  1.30it/s]
Loading checkpoint shards:  24%|██▎       | 4/17 [00:03<00:09,  1.31it/s]
Loading checkpoint shards:  24%|██▎       | 4/17 [00:03<00:10,  1.29it/s]
Loading checkpoint shards:  24%|██▎       | 4/17 [00:03<00:09,  1.30it/s]
Loading checkpoint shards:  24%|██▎       | 4/17 [00:03<00:10,  1.29it/s]
Loading checkpoint shards:  24%|██▎       | 4/17 [00:03<00:10,  1.29it/s]
Loading checkpoint shards:  24%|██▎       | 4/17 [00:03<00:09,  1.30it/s]
Loading checkpoint shards:  24%|██▎       | 4/17 [00:03<00:10,  1.25it/s]
Loading checkpoint shards:  29%|██▉       | 5/17 [00:03<00:09,  1.29it/s]
Loading checkpoint shards:  29%|██▉       | 5/17 [00:03<00:09,  1.30it/s]
Loading checkpoint shards:  29%|██▉       | 5/17 [00:03<00:09,  1.29it/s]
Loading checkpoint shards:  29%|██▉       | 5/17 [00:03<00:09,  1.30it/s]
Loading checkpoint shards:  29%|██▉       | 5/17 [00:03<00:09,  1.30it/s]
Loading checkpoint shards:  29%|██▉       | 5/17 [00:03<00:09,  1.29it/s]
Loading checkpoint shards:  29%|██▉       | 5/17 [00:03<00:09,  1.29it/s]
Loading checkpoint shards:  29%|██▉       | 5/17 [00:03<00:09,  1.28it/s]
Loading checkpoint shards:  35%|███▌      | 6/17 [00:04<00:08,  1.30it/s]
Loading checkpoint shards:  35%|███▌      | 6/17 [00:04<00:08,  1.30it/s]
Loading checkpoint shards:  35%|███▌      | 6/17 [00:04<00:08,  1.30it/s]
Loading checkpoint shards:  35%|███▌      | 6/17 [00:04<00:08,  1.30it/s]
Loading checkpoint shards:  35%|███▌      | 6/17 [00:04<00:08,  1.29it/s]
Loading checkpoint shards:  35%|███▌      | 6/17 [00:04<00:08,  1.29it/s]
Loading checkpoint shards:  35%|███▌      | 6/17 [00:04<00:08,  1.29it/s]
Loading checkpoint shards:  35%|███▌      | 6/17 [00:04<00:08,  1.28it/s]
Loading checkpoint shards:  41%|████      | 7/17 [00:05<00:07,  1.29it/s]
Loading checkpoint shards:  41%|████      | 7/17 [00:05<00:07,  1.29it/s]
Loading checkpoint shards:  41%|████      | 7/17 [00:05<00:07,  1.29it/s]
Loading checkpoint shards:  41%|████      | 7/17 [00:05<00:07,  1.29it/s]
Loading checkpoint shards:  41%|████      | 7/17 [00:05<00:07,  1.30it/s]
Loading checkpoint shards:  41%|████      | 7/17 [00:05<00:07,  1.29it/s]
Loading checkpoint shards:  41%|████      | 7/17 [00:05<00:07,  1.29it/s]
Loading checkpoint shards:  41%|████      | 7/17 [00:05<00:07,  1.29it/s]
Loading checkpoint shards:  47%|████▋     | 8/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  47%|████▋     | 8/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  47%|████▋     | 8/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  47%|████▋     | 8/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  47%|████▋     | 8/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  47%|████▋     | 8/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  47%|████▋     | 8/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  47%|████▋     | 8/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  53%|█████▎    | 9/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  53%|█████▎    | 9/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  53%|█████▎    | 9/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  53%|█████▎    | 9/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  53%|█████▎    | 9/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  53%|█████▎    | 9/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  53%|█████▎    | 9/17 [00:06<00:06,  1.30it/s]
Loading checkpoint shards:  53%|█████▎    | 9/17 [00:07<00:06,  1.30it/s]
Loading checkpoint shards:  59%|█████▉    | 10/17 [00:07<00:05,  1.31it/s]
Loading checkpoint shards:  59%|█████▉    | 10/17 [00:07<00:05,  1.31it/s]
Loading checkpoint shards:  59%|█████▉    | 10/17 [00:07<00:05,  1.31it/s]
Loading checkpoint shards:  59%|█████▉    | 10/17 [00:07<00:05,  1.31it/s]
Loading checkpoint shards:  59%|█████▉    | 10/17 [00:07<00:05,  1.31it/s]
Loading checkpoint shards:  59%|█████▉    | 10/17 [00:07<00:05,  1.31it/s]
Loading checkpoint shards:  59%|█████▉    | 10/17 [00:07<00:05,  1.31it/s]
Loading checkpoint shards:  59%|█████▉    | 10/17 [00:07<00:05,  1.31it/s]
Loading checkpoint shards:  65%|██████▍   | 11/17 [00:08<00:04,  1.32it/s]
Loading checkpoint shards:  65%|██████▍   | 11/17 [00:08<00:04,  1.32it/s]
Loading checkpoint shards:  65%|██████▍   | 11/17 [00:08<00:04,  1.32it/s]
Loading checkpoint shards:  65%|██████▍   | 11/17 [00:08<00:04,  1.32it/s]
Loading checkpoint shards:  65%|██████▍   | 11/17 [00:08<00:04,  1.32it/s]
Loading checkpoint shards:  65%|██████▍   | 11/17 [00:08<00:04,  1.32it/s]
Loading checkpoint shards:  65%|██████▍   | 11/17 [00:08<00:04,  1.32it/s]
Loading checkpoint shards:  65%|██████▍   | 11/17 [00:08<00:04,  1.32it/s]
Loading checkpoint shards:  71%|███████   | 12/17 [00:09<00:03,  1.33it/s]
Loading checkpoint shards:  71%|███████   | 12/17 [00:09<00:03,  1.33it/s]
Loading checkpoint shards:  71%|███████   | 12/17 [00:09<00:03,  1.33it/s]
Loading checkpoint shards:  71%|███████   | 12/17 [00:09<00:03,  1.33it/s]
Loading checkpoint shards:  71%|███████   | 12/17 [00:09<00:03,  1.33it/s]
Loading checkpoint shards:  71%|███████   | 12/17 [00:09<00:03,  1.33it/s]
Loading checkpoint shards:  71%|███████   | 12/17 [00:09<00:03,  1.33it/s]
Loading checkpoint shards:  71%|███████   | 12/17 [00:09<00:03,  1.32it/s]
Loading checkpoint shards:  76%|███████▋  | 13/17 [00:09<00:03,  1.32it/s]
Loading checkpoint shards:  76%|███████▋  | 13/17 [00:09<00:03,  1.32it/s]
Loading checkpoint shards:  76%|███████▋  | 13/17 [00:09<00:03,  1.32it/s]
Loading checkpoint shards:  76%|███████▋  | 13/17 [00:09<00:03,  1.32it/s]
Loading checkpoint shards:  76%|███████▋  | 13/17 [00:09<00:03,  1.32it/s]
Loading checkpoint shards:  76%|███████▋  | 13/17 [00:09<00:03,  1.32it/s]
Loading checkpoint shards:  76%|███████▋  | 13/17 [00:09<00:03,  1.32it/s]
Loading checkpoint shards:  76%|███████▋  | 13/17 [00:10<00:03,  1.32it/s]
Loading checkpoint shards:  82%|████████▏ | 14/17 [00:10<00:02,  1.32it/s]
Loading checkpoint shards:  82%|████████▏ | 14/17 [00:10<00:02,  1.32it/s]
Loading checkpoint shards:  82%|████████▏ | 14/17 [00:10<00:02,  1.32it/s]
Loading checkpoint shards:  82%|████████▏ | 14/17 [00:10<00:02,  1.32it/s]
Loading checkpoint shards:  82%|████████▏ | 14/17 [00:10<00:02,  1.32it/s]
Loading checkpoint shards:  82%|████████▏ | 14/17 [00:10<00:02,  1.32it/s]
Loading checkpoint shards:  82%|████████▏ | 14/17 [00:10<00:02,  1.32it/s]
Loading checkpoint shards:  82%|████████▏ | 14/17 [00:10<00:02,  1.31it/s]
Loading checkpoint shards:  88%|████████▊ | 15/17 [00:11<00:01,  1.30it/s]
Loading checkpoint shards:  88%|████████▊ | 15/17 [00:11<00:01,  1.30it/s]
Loading checkpoint shards:  88%|████████▊ | 15/17 [00:11<00:01,  1.30it/s]
Loading checkpoint shards:  88%|████████▊ | 15/17 [00:11<00:01,  1.30it/s]
Loading checkpoint shards:  88%|████████▊ | 15/17 [00:11<00:01,  1.30it/s]
Loading checkpoint shards:  88%|████████▊ | 15/17 [00:11<00:01,  1.30it/s]
Loading checkpoint shards:  88%|████████▊ | 15/17 [00:11<00:01,  1.30it/s]
Loading checkpoint shards:  88%|████████▊ | 15/17 [00:11<00:01,  1.29it/s]
Loading checkpoint shards:  94%|█████████▍| 16/17 [00:12<00:00,  1.30it/s]
Loading checkpoint shards:  94%|█████████▍| 16/17 [00:12<00:00,  1.30it/s]
Loading checkpoint shards:  94%|█████████▍| 16/17 [00:12<00:00,  1.30it/s]
Loading checkpoint shards:  94%|█████████▍| 16/17 [00:12<00:00,  1.30it/s]
Loading checkpoint shards:  94%|█████████▍| 16/17 [00:12<00:00,  1.30it/s]
Loading checkpoint shards:  94%|█████████▍| 16/17 [00:12<00:00,  1.30it/s]
Loading checkpoint shards:  94%|█████████▍| 16/17 [00:12<00:00,  1.30it/s]
Loading checkpoint shards:  94%|█████████▍| 16/17 [00:12<00:00,  1.31it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.39it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.32it/s]

Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.39it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.39it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.32it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.32it/s]


Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.39it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.33it/s]

Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.39it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.32it/s]

Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.39it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.39it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.32it/s]

Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.32it/s]

Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.39it/s]
Loading checkpoint shards: 100%|██████████| 17/17 [00:12<00:00,  1.31it/s]
[INFO|modeling_utils.py:4987] 2025-04-13 14:57:37,816 >> All model checkpoint weights were used when initializing Qwen2ForCausalLM.

[INFO|modeling_utils.py:4995] 2025-04-13 14:57:37,816 >> All the weights of Qwen2ForCausalLM were initialized from the model checkpoint at /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct.
If your task is similar to the task the model of the checkpoint was trained on, you can already use Qwen2ForCausalLM for predictions without further training.
[INFO|configuration_utils.py:1092] 2025-04-13 14:57:37,821 >> loading configuration file /home/stern/GRPO/saved_models/Qwen2.5-32B-Instruct/generation_config.json
[INFO|configuration_utils.py:1139] 2025-04-13 14:57:37,821 >> Generate config GenerationConfig {
  "bos_token_id": 151643,
  "do_sample": true,
  "eos_token_id": [
    151645,
    151643
  ],
  "pad_token_id": 151643,
  "repetition_penalty": 1.05,
  "temperature": 0.7,
  "top_k": 20,
  "top_p": 0.8
}

/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
  trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
  trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
  trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
  trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
  trainer = OfflineREINFORCETrainer(
Using custom data configuration default-e96436450119f8e9
INFO:datasets.builder:Using custom data configuration default-e96436450119f8e9
Loading Dataset Infos from /home/stern/.local/lib/python3.10/site-packages/datasets/packaged_modules/json
INFO:datasets.info:Loading Dataset Infos from /home/stern/.local/lib/python3.10/site-packages/datasets/packaged_modules/json
Overwrite dataset info from restored data version if exists.
INFO:datasets.builder:Overwrite dataset info from restored data version if exists.
Loading Dataset info from /home/stern/.cache/huggingface/datasets/json/default-e96436450119f8e9/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092
INFO:datasets.info:Loading Dataset info from /home/stern/.cache/huggingface/datasets/json/default-e96436450119f8e9/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
  trainer = OfflineREINFORCETrainer(
Found cached dataset json (/home/stern/.cache/huggingface/datasets/json/default-e96436450119f8e9/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092)
INFO:datasets.builder:Found cached dataset json (/home/stern/.cache/huggingface/datasets/json/default-e96436450119f8e9/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092)
Loading Dataset info from /home/stern/.cache/huggingface/datasets/json/default-e96436450119f8e9/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092
INFO:datasets.info:Loading Dataset info from /home/stern/.cache/huggingface/datasets/json/default-e96436450119f8e9/0.0.0/f4e89e8750d5d5ffbef2c078bf0ddfedef29dc2faff52a6255cf513c05eb1092
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
  trainer = OfflineREINFORCETrainer(
/home/stern/GRPO/offline_rl_v2/train.py:274: FutureWarning: `tokenizer` is deprecated and will be removed in version 5.0.0 for `OfflineREINFORCETrainer.__init__`. Use `processing_class` instead.
  trainer = OfflineREINFORCETrainer(
[INFO|trainer.py:748] 2025-04-13 14:57:38,146 >> Using auto half precision backend
INFO:__main__:*** Train ***
[INFO|deepspeed.py:386] 2025-04-13 14:57:38,377 >> Detected ZeRO Offload and non-DeepSpeed optimizers: This combination should work as long as the custom optimizer has both CPU and GPU implementation (except LAMB)
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/cpu_adam/build.ninja...
Building extension module cpu_adam...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.529125928878784 seconds
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/cpu_adam/build.ninja...
Building extension module cpu_adam...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.3708255290985107 seconds
Adam Optimizer #0 is created with AVX512 arithmetic capability.
Config: alpha=0.000002, betas=(0.900000, 0.999000), weight_decay=0.010000, adam_w=1
[2025-04-13 14:57:42,026] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed info: version=0.16.5, git-hash=unknown, git-branch=unknown
[2025-04-13 14:57:42,026] [INFO] [config.py:734:__init__] Config mesh_device None world_size = 8
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/cpu_adam/build.ninja...
Building extension module cpu_adam...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.3084402084350586 seconds
[2025-04-13 14:57:42,068] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed Flops Profiler Enabled: False
[2025-04-13 14:57:42,071] [INFO] [logging.py:107:log_dist] [Rank 0] Using client Optimizer as basic optimizer
[2025-04-13 14:57:42,071] [INFO] [logging.py:107:log_dist] [Rank 0] Removing param_group that has no 'params' in the basic Optimizer
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.6082324981689453 seconds
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.629338502883911 seconds
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/cpu_adam/build.ninja...
Building extension module cpu_adam...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.327012538909912 seconds
[2025-04-13 14:57:42,124] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed Basic Optimizer = DeepSpeedCPUAdam
[2025-04-13 14:57:42,124] [INFO] [utils.py:59:is_zero_supported_optimizer] Checking ZeRO support for optimizer=DeepSpeedCPUAdam type=<class 'deepspeed.ops.adam.cpu_adam.DeepSpeedCPUAdam'>
[2025-04-13 14:57:42,124] [INFO] [logging.py:107:log_dist] [Rank 0] Creating fp16 ZeRO stage 3 optimizer, MiCS is enabled False, Hierarchical params gather False
[2025-04-13 14:57:42,124] [INFO] [logging.py:107:log_dist] [Rank 0] Creating torch.bfloat16 ZeRO stage 3 optimizer
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Installed CUDA version 12.4 does not match the version torch was compiled with 12.1 but since the APIs are compatible, accepting this combination
Using /home/stern/.cache/torch_extensions/py310_cu121 as PyTorch extensions root...
Emitting ninja build file /home/stern/.cache/torch_extensions/py310_cu121/cpu_adam/build.ninja...
Building extension module cpu_adam...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
ninja: no work to do.
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.439788341522217 seconds
[2025-04-13 14:57:42,256] [INFO] [utils.py:781:see_memory_usage] Stage 3 initialize beginning
[2025-04-13 14:57:42,257] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 2.9 GB         CA 0.0 GB         Max_CA 3 GB 
[2025-04-13 14:57:42,257] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 140.14 GB, percent = 13.9%
[2025-04-13 14:57:42,259] [INFO] [stage3.py:170:__init__] Reduce bucket size 100000000
[2025-04-13 14:57:42,259] [INFO] [stage3.py:171:__init__] Prefetch bucket size 100000000
Loading extension module cpu_adam...
Time to load cpu_adam op: 2.5354325771331787 seconds
[2025-04-13 14:57:42,358] [INFO] [utils.py:781:see_memory_usage] DeepSpeedZeRoOffload initialize [begin]
[2025-04-13 14:57:42,359] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
[2025-04-13 14:57:42,359] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 140.16 GB, percent = 13.9%
Parameter Offload: Total persistent parameters: 1119232 in 321 params
[2025-04-13 14:57:42,535] [INFO] [utils.py:781:see_memory_usage] DeepSpeedZeRoOffload initialize [end]
[2025-04-13 14:57:42,536] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
[2025-04-13 14:57:42,536] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 140.18 GB, percent = 13.9%
[2025-04-13 14:57:42,642] [INFO] [utils.py:781:see_memory_usage] Before creating fp16 partitions
[2025-04-13 14:57:42,643] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
[2025-04-13 14:57:42,643] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 140.18 GB, percent = 13.9%
[2025-04-13 14:58:01,980] [INFO] [utils.py:781:see_memory_usage] After creating fp16 partitions: 41
[2025-04-13 14:58:01,981] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
[2025-04-13 14:58:01,981] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 205.39 GB, percent = 20.4%
[2025-04-13 14:58:02,134] [INFO] [utils.py:781:see_memory_usage] Before creating fp32 partitions
[2025-04-13 14:58:02,135] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
[2025-04-13 14:58:02,135] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 205.38 GB, percent = 20.4%
[2025-04-13 14:58:06,255] [INFO] [utils.py:781:see_memory_usage] After creating fp32 partitions
[2025-04-13 14:58:06,256] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
[2025-04-13 14:58:06,256] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 285.93 GB, percent = 28.4%
[2025-04-13 14:58:08,607] [INFO] [utils.py:781:see_memory_usage] Before initializing optimizer states
[2025-04-13 14:58:08,608] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
[2025-04-13 14:58:08,608] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 341.25 GB, percent = 33.9%
[2025-04-13 14:58:21,496] [INFO] [utils.py:781:see_memory_usage] After initializing optimizer states
[2025-04-13 14:58:21,496] [INFO] [utils.py:782:see_memory_usage] MA 0.0 GB         Max_MA 0.0 GB         CA 0.0 GB         Max_CA 0 GB 
[2025-04-13 14:58:21,497] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 515.49 GB, percent = 51.2%
[2025-04-13 14:58:21,497] [INFO] [stage3.py:534:_setup_for_real_optimizer] optimizer state initialized
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
  batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
  batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
  batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
  batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
  batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
  batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
  batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
[WARNING|logging.py:329] 2025-04-13 14:58:27,659 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-13 14:58:27,660 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-13 14:58:27,662 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-13 14:58:27,663 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-13 14:58:27,663 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-13 14:58:27,664 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[WARNING|logging.py:329] 2025-04-13 14:58:27,666 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
[2025-04-13 14:58:27,751] [INFO] [utils.py:781:see_memory_usage] After initializing ZeRO optimizer
[2025-04-13 14:58:27,751] [INFO] [utils.py:782:see_memory_usage] MA 0.19 GB         Max_MA 3.09 GB         CA 3.09 GB         Max_CA 3 GB 
[2025-04-13 14:58:27,752] [INFO] [utils.py:789:see_memory_usage] CPU Virtual Memory:  used = 566.38 GB, percent = 56.2%
[2025-04-13 14:58:27,752] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed Final Optimizer = DeepSpeedZeroOptimizer_Stage3
[2025-04-13 14:58:27,752] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed using configured LR scheduler = None
[2025-04-13 14:58:27,752] [INFO] [logging.py:107:log_dist] [Rank 0] DeepSpeed LR Scheduler = None
[2025-04-13 14:58:27,752] [INFO] [logging.py:107:log_dist] [Rank 0] step=0, skipped=0, lr=[0.0, 0.0], mom=[(0.9, 0.999), (0.9, 0.999)]
[2025-04-13 14:58:27,753] [INFO] [config.py:1000:print] DeepSpeedEngine configuration:
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   activation_checkpointing_config  {
    "partition_activations": false, 
    "contiguous_memory_optimization": false, 
    "cpu_checkpointing": false, 
    "number_checkpoints": null, 
    "synchronize_checkpoint_boundary": false, 
    "profile": false
}
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   aio_config ................... {'block_size': 1048576, 'queue_depth': 8, 'intra_op_parallelism': 1, 'single_submit': False, 'overlap_events': True, 'use_gds': False}
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   amp_enabled .................. False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   amp_params ................... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   autotuning_config ............ {
    "enabled": false, 
    "start_step": null, 
    "end_step": null, 
    "metric_path": null, 
    "arg_mappings": null, 
    "metric": "throughput", 
    "model_info": null, 
    "results_dir": "autotuning_results", 
    "exps_dir": "autotuning_exps", 
    "overwrite": true, 
    "fast": true, 
    "start_profile_step": 3, 
    "end_profile_step": 5, 
    "tuner_type": "gridsearch", 
    "tuner_early_stopping": 5, 
    "tuner_num_trials": 50, 
    "model_info_path": null, 
    "mp_size": 1, 
    "max_train_batch_size": null, 
    "min_train_batch_size": 1, 
    "max_train_micro_batch_size_per_gpu": 1.024000e+03, 
    "min_train_micro_batch_size_per_gpu": 1, 
    "num_tuning_micro_batch_sizes": 3
}
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   bfloat16_enabled ............. True
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   bfloat16_immediate_grad_update  True
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   checkpoint_parallel_write_pipeline  False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   checkpoint_tag_validation_enabled  True
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   checkpoint_tag_validation_fail  False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   comms_config ................. <deepspeed.comm.config.DeepSpeedCommsConfig object at 0x7d648039b6d0>
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   communication_data_type ...... None
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   compression_config ........... {'weight_quantization': {'shared_parameters': {'enabled': False, 'quantizer_kernel': False, 'schedule_offset': 0, 'quantize_groups': 1, 'quantize_verbose': False, 'quantization_type': 'symmetric', 'quantize_weight_in_forward': False, 'rounding': 'nearest', 'fp16_mixed_quantize': False, 'quantize_change_ratio': 0.001}, 'different_groups': {}}, 'activation_quantization': {'shared_parameters': {'enabled': False, 'quantization_type': 'symmetric', 'range_calibration': 'dynamic', 'schedule_offset': 1000}, 'different_groups': {}}, 'sparse_pruning': {'shared_parameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'row_pruning': {'shared_parameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'head_pruning': {'shared_parameters': {'enabled': False, 'method': 'topk', 'schedule_offset': 1000}, 'different_groups': {}}, 'channel_pruning': {'shared_parameters': {'enabled': False, 'method': 'l1', 'schedule_offset': 1000}, 'different_groups': {}}, 'layer_reduction': {'enabled': False}}
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   curriculum_enabled_legacy .... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   curriculum_params_legacy ..... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   data_efficiency_config ....... {'enabled': False, 'seed': 1234, 'data_sampling': {'enabled': False, 'num_epochs': 1000, 'num_workers': 0, 'pin_memory': False, 'curriculum_learning': {'enabled': False}, 'dynamic_batching': {'enabled': False, 'lr_scaling_method': 'linear', 'min_batch_size': 1, 'max_batch_size': None, 'sequence_picking_order': 'dataloader', 'verbose': False}}, 'data_routing': {'enabled': False, 'random_ltd': {'enabled': False, 'layer_token_lr_schedule': {'enabled': False}}}}
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   data_efficiency_enabled ...... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   dataloader_drop_last ......... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   disable_allgather ............ False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   dump_state ................... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   dynamic_loss_scale_args ...... None
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   eigenvalue_enabled ........... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   eigenvalue_gas_boundary_resolution  1
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   eigenvalue_layer_name ........ bert.encoder.layer
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   eigenvalue_layer_num ......... 0
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   eigenvalue_max_iter .......... 100
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   eigenvalue_stability ......... 1e-06
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   eigenvalue_tol ............... 0.01
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   eigenvalue_verbose ........... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   elasticity_enabled ........... False
[2025-04-13 14:58:27,754] [INFO] [config.py:1004:print]   flops_profiler_config ........ {
    "enabled": false, 
    "recompute_fwd_factor": 0.0, 
    "profile_step": 1, 
    "module_depth": -1, 
    "top_modules": 1, 
    "detailed": true, 
    "output_file": null
}
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   fp16_auto_cast ............... None
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   fp16_enabled ................. False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   fp16_master_weights_and_gradients  False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   global_rank .................. 0
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   grad_accum_dtype ............. None
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   gradient_accumulation_steps .. 4
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   gradient_clipping ............ 1.0
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   gradient_predivide_factor .... 1.0
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   graph_harvesting ............. False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   hybrid_engine ................ enabled=False max_out_tokens=512 inference_tp_size=1 release_inference_cache=False pin_parameters=True tp_gather_partition_size=8
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   initial_dynamic_scale ........ 1
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   load_universal_checkpoint .... False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   loss_scale ................... 1.0
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   memory_breakdown ............. False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   mics_hierarchial_params_gather  False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   mics_shard_size .............. -1
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   monitor_config ............... tensorboard=TensorBoardConfig(enabled=False, output_path='', job_name='DeepSpeedJobName') comet=CometConfig(enabled=False, samples_log_interval=100, project=None, workspace=None, api_key=None, experiment_name=None, experiment_key=None, online=None, mode=None) wandb=WandbConfig(enabled=False, group=None, team=None, project='deepspeed') csv_monitor=CSVConfig(enabled=False, output_path='', job_name='DeepSpeedJobName')
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   nebula_config ................ {
    "enabled": false, 
    "persistent_storage_path": null, 
    "persistent_time_interval": 100, 
    "num_of_version_in_retention": 2, 
    "enable_nebula_load": true, 
    "load_path": null
}
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   optimizer_legacy_fusion ...... False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   optimizer_name ............... None
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   optimizer_params ............. None
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   pipeline ..................... {'stages': 'auto', 'partition': 'best', 'seed_layers': False, 'activation_checkpoint_interval': 0, 'pipe_partitioned': True, 'grad_partitioned': True}
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   pld_enabled .................. False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   pld_params ................... False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   prescale_gradients ........... False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   scheduler_name ............... None
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   scheduler_params ............. None
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   seq_parallel_communication_data_type  torch.float32
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   sparse_attention ............. None
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   sparse_gradients_enabled ..... False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   steps_per_print .............. inf
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   tensor_parallel_config ....... dtype=torch.float16 autotp_size=0 tensor_parallel=TPConfig(tp_size=1, tp_grain_size=1, mpu=None, tp_group=None) injection_policy_tuple=None keep_module_on_host=False replace_with_kernel_inject=False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   timers_config ................ enabled=True synchronized=True
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   train_batch_size ............. 32
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   train_micro_batch_size_per_gpu  1
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   use_data_before_expert_parallel_  False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   use_node_local_storage ....... False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   wall_clock_breakdown ......... False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   weight_quantization_config ... None
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   world_size ................... 8
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   zero_allow_untested_optimizer  True
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   zero_config .................. stage=3 contiguous_gradients=True reduce_scatter=True reduce_bucket_size=100000000 use_multi_rank_bucket_allreduce=True allgather_partitions=True allgather_bucket_size=500000000 overlap_comm=True load_from_fp32_weights=True elastic_checkpoint=False offload_param=DeepSpeedZeroOffloadParamConfig(device='cpu', nvme_path=None, buffer_count=5, buffer_size=100000000, max_in_cpu=1000000000, pin_memory=True) offload_optimizer=DeepSpeedZeroOffloadOptimizerConfig(device='cpu', nvme_path=None, buffer_count=4, pin_memory=True, pipeline_read=False, pipeline_write=False, fast_init=False, ratio=1.0) sub_group_size=100000000 cpu_offload_param=None cpu_offload_use_pin_memory=None cpu_offload=None prefetch_bucket_size=100000000 param_persistence_threshold=100000 model_persistence_threshold=9223372036854775807 max_live_parameters=100000000 max_reuse_distance=100000000 gather_16bit_weights_on_model_save=True module_granularity_threshold=0 use_all_reduce_for_fetch_params=False stage3_gather_fp16_weights_on_model_save=False ignore_unused_parameters=True legacy_stage1=False round_robin_gradients=False zero_hpz_partition_size=1 zero_quantized_weights=False zero_quantized_nontrainable_weights=False zero_quantized_gradients=False zeropp_loco_param=None mics_shard_size=-1 mics_hierarchical_params_gather=False memory_efficient_linear=True pipeline_loading_checkpoint=False override_module_apply=True log_trace_cache_warnings=False
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   zero_enabled ................. True
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   zero_force_ds_cpu_optimizer .. True
[2025-04-13 14:58:27,755] [INFO] [config.py:1004:print]   zero_optimization_stage ...... 3
[2025-04-13 14:58:27,755] [INFO] [config.py:990:print_user_config]   json = {
    "fp16": {
        "enabled": false
    }, 
    "bf16": {
        "enabled": true
    }, 
    "train_micro_batch_size_per_gpu": 1, 
    "gradient_accumulation_steps": 4, 
    "zero_optimization": {
        "stage": 3, 
        "offload_optimizer": {
            "device": "cpu", 
            "pin_memory": true
        }, 
        "offload_param": {
            "device": "cpu", 
            "pin_memory": true
        }, 
        "overlap_comm": true, 
        "contiguous_gradients": true, 
        "sub_group_size": 1.000000e+08, 
        "reduce_bucket_size": 1.000000e+08, 
        "stage3_prefetch_bucket_size": 1.000000e+08, 
        "stage3_param_persistence_threshold": 1.000000e+05, 
        "stage3_max_live_parameters": 1.000000e+08, 
        "stage3_max_reuse_distance": 1.000000e+08, 
        "stage3_gather_16bit_weights_on_model_save": true
    }, 
    "gradient_clipping": 1.0, 
    "wall_clock_breakdown": false, 
    "steps_per_print": inf, 
    "zero_allow_untested_optimizer": true
}
[INFO|trainer.py:2409] 2025-04-13 14:58:27,755 >> ***** Running training *****
[INFO|trainer.py:2410] 2025-04-13 14:58:27,756 >>   Num examples = 2,688
[INFO|trainer.py:2411] 2025-04-13 14:58:27,756 >>   Num Epochs = 4
[INFO|trainer.py:2412] 2025-04-13 14:58:27,756 >>   Instantaneous batch size per device = 1
[INFO|trainer.py:2415] 2025-04-13 14:58:27,756 >>   Total train batch size (w. parallel, distributed & accumulation) = 32
[INFO|trainer.py:2416] 2025-04-13 14:58:27,756 >>   Gradient Accumulation steps = 4
[INFO|trainer.py:2417] 2025-04-13 14:58:27,756 >>   Total optimization steps = 336
[INFO|trainer.py:2418] 2025-04-13 14:58:27,757 >>   Number of trainable parameters = 32,763,876,352

  0%|          | 0/336 [00:00<?, ?it/s]/home/stern/.local/lib/python3.10/site-packages/transformers/data/data_collator.py:741: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:278.)
  batch["labels"] = torch.tensor(batch["labels"], dtype=torch.int64)
[WARNING|logging.py:329] 2025-04-13 14:58:27,823 >> `use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]
/home/stern/.local/lib/python3.10/site-packages/torch/utils/checkpoint.py:295: FutureWarning: `torch.cpu.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cpu', args...)` instead.
  with torch.enable_grad(), device_autocast_ctx, torch.cpu.amp.autocast(**ctx.cpu_autocast_kwargs):  # type: ignore[attr-defined]

  0%|          | 1/336 [00:27<2:30:51, 27.02s/it]
                                                 
{'loss': 0.0336, 'grad_norm': 0.556594967842102, 'learning_rate': 1.818181818181818e-07, 'kl': -0.0009, 'entropy': -0.0017, 'ce_loss': 0.0271, 'epoch': 0.01}

  0%|          | 1/336 [00:27<2:30:51, 27.02s/it]
  1%|          | 2/336 [00:47<2:09:37, 23.29s/it]
                                                 
{'loss': 0.0306, 'grad_norm': 0.4636707007884979, 'learning_rate': 3.636363636363636e-07, 'kl': 0.0014, 'entropy': 0.0205, 'ce_loss': 0.01, 'epoch': 0.02}

  1%|          | 2/336 [00:47<2:09:37, 23.29s/it]
  1%|          | 3/336 [01:05<1:55:13, 20.76s/it]
                                                 
{'loss': 0.0263, 'grad_norm': 0.4290856719017029, 'learning_rate': 5.454545454545454e-07, 'kl': -0.0006, 'entropy': -0.0026, 'ce_loss': 0.0253, 'epoch': 0.04}

  1%|          | 3/336 [01:05<1:55:13, 20.76s/it]
  1%|          | 4/336 [01:27<1:58:21, 21.39s/it]
                                                 
{'loss': 0.0322, 'grad_norm': 0.5268592238426208, 'learning_rate': 7.272727272727272e-07, 'kl': 0.0004, 'entropy': -0.0366, 'ce_loss': 0.0298, 'epoch': 0.05}

  1%|          | 4/336 [01:27<1:58:21, 21.39s/it]
  1%|▏         | 5/336 [01:43<1:46:53, 19.38s/it]
                                                 
{'loss': 0.0364, 'grad_norm': 0.5142350196838379, 'learning_rate': 9.09090909090909e-07, 'kl': 0.0027, 'entropy': -0.0227, 'ce_loss': 0.0203, 'epoch': 0.06}

  1%|▏         | 5/336 [01:43<1:46:53, 19.38s/it]
  2%|▏         | 6/336 [02:02<1:46:20, 19.33s/it]
                                                 
{'loss': 0.0306, 'grad_norm': 0.3994697332382202, 'learning_rate': 1.0909090909090908e-06, 'kl': 0.0042, 'entropy': 0.0659, 'ce_loss': 0.0104, 'epoch': 0.07}

  2%|▏         | 6/336 [02:02<1:46:20, 19.33s/it]
  2%|▏         | 7/336 [02:21<1:43:55, 18.95s/it]
                                                 
{'loss': 0.0312, 'grad_norm': 0.29699787497520447, 'learning_rate': 1.2727272727272726e-06, 'kl': -0.0002, 'entropy': -0.0356, 'ce_loss': 0.0093, 'epoch': 0.08}

  2%|▏         | 7/336 [02:21<1:43:55, 18.95s/it]
  2%|▏         | 8/336 [02:40<1:44:02, 19.03s/it]
                                                 
{'loss': 0.037, 'grad_norm': 0.3551566004753113, 'learning_rate': 1.4545454545454544e-06, 'kl': 0.0039, 'entropy': 0.0383, 'ce_loss': 0.0162, 'epoch': 0.1}

  2%|▏         | 8/336 [02:40<1:44:02, 19.03s/it]
  3%|▎         | 9/336 [02:59<1:43:29, 18.99s/it]
                                                 
{'loss': 0.0253, 'grad_norm': 0.2004219889640808, 'learning_rate': 1.6363636363636365e-06, 'kl': 0.0058, 'entropy': -0.0649, 'ce_loss': 0.0215, 'epoch': 0.11}

  3%|▎         | 9/336 [02:59<1:43:29, 18.99s/it]
  3%|▎         | 10/336 [03:14<1:37:50, 18.01s/it]
                                                  
{'loss': 0.0322, 'grad_norm': 0.23521199822425842, 'learning_rate': 1.818181818181818e-06, 'kl': 0.0077, 'entropy': -0.0442, 'ce_loss': 0.0145, 'epoch': 0.12}

  3%|▎         | 10/336 [03:14<1:37:50, 18.01s/it]
  3%|▎         | 11/336 [03:35<1:42:21, 18.90s/it]
                                                  
{'loss': 0.0294, 'grad_norm': 0.2418721467256546, 'learning_rate': 2e-06, 'kl': 0.0026, 'entropy': -0.0408, 'ce_loss': 0.0082, 'epoch': 0.13}

  3%|▎         | 11/336 [03:35<1:42:21, 18.90s/it]
  4%|▎         | 12/336 [03:51<1:37:02, 17.97s/it]
                                                  
{'loss': 0.0351, 'grad_norm': 0.24574391543865204, 'learning_rate': 1.999953280342959e-06, 'kl': 0.0102, 'entropy': 0.0435, 'ce_loss': 0.0304, 'epoch': 0.14}

  4%|▎         | 12/336 [03:51<1:37:02, 17.97s/it]
  4%|▍         | 13/336 [04:11<1:39:40, 18.51s/it]
                                                  
{'loss': 0.0292, 'grad_norm': 0.24601855874061584, 'learning_rate': 1.9998131257372875e-06, 'kl': 0.0054, 'entropy': 0.0364, 'ce_loss': 0.0159, 'epoch': 0.15}

  4%|▍         | 13/336 [04:11<1:39:40, 18.51s/it]
  4%|▍         | 14/336 [04:30<1:40:12, 18.67s/it]
                                                  
{'loss': 0.038, 'grad_norm': 0.24783623218536377, 'learning_rate': 1.9995795492789365e-06, 'kl': 0.0111, 'entropy': -0.0146, 'ce_loss': 0.0152, 'epoch': 0.17}

  4%|▍         | 14/336 [04:30<1:40:12, 18.67s/it]
  4%|▍         | 15/336 [04:52<1:45:22, 19.70s/it]
                                                  
{'loss': 0.0335, 'grad_norm': 0.2631509006023407, 'learning_rate': 1.99925257279313e-06, 'kl': 0.0105, 'entropy': -0.0591, 'ce_loss': 0.0152, 'epoch': 0.18}

  4%|▍         | 15/336 [04:52<1:45:22, 19.70s/it]
  5%|▍         | 16/336 [05:10<1:41:34, 19.04s/it]
                                                  
{'loss': 0.0288, 'grad_norm': 0.22452250123023987, 'learning_rate': 1.9988322268323264e-06, 'kl': 0.0099, 'entropy': 0.0449, 'ce_loss': 0.0302, 'epoch': 0.19}

  5%|▍         | 16/336 [05:10<1:41:34, 19.04s/it]
  5%|▌         | 17/336 [05:27<1:38:56, 18.61s/it]
                                                  
{'loss': 0.0191, 'grad_norm': 0.17368917167186737, 'learning_rate': 1.998318550673364e-06, 'kl': 0.017, 'entropy': 0.0147, 'ce_loss': 0.0144, 'epoch': 0.2}

  5%|▌         | 17/336 [05:27<1:38:56, 18.61s/it]
  5%|▌         | 18/336 [05:43<1:34:10, 17.77s/it]
                                                  
{'loss': 0.0338, 'grad_norm': 0.32134443521499634, 'learning_rate': 1.997711592313791e-06, 'kl': 0.0036, 'entropy': -0.0304, 'ce_loss': 0.0167, 'epoch': 0.21}

  5%|▌         | 18/336 [05:43<1:34:10, 17.77s/it]
  6%|▌         | 19/336 [06:01<1:34:50, 17.95s/it]
                                                  
{'loss': 0.028, 'grad_norm': 0.2845238447189331, 'learning_rate': 1.9970114084673796e-06, 'kl': 0.0045, 'entropy': -0.0452, 'ce_loss': 0.03, 'epoch': 0.23}

  6%|▌         | 19/336 [06:01<1:34:50, 17.95s/it]
  6%|▌         | 20/336 [06:17<1:30:53, 17.26s/it]
                                                  
{'loss': 0.0335, 'grad_norm': 0.3595256507396698, 'learning_rate': 1.9962180645588286e-06, 'kl': 0.0123, 'entropy': -0.0933, 'ce_loss': 0.0147, 'epoch': 0.24}

  6%|▌         | 20/336 [06:17<1:30:53, 17.26s/it]
  6%|▋         | 21/336 [06:33<1:27:56, 16.75s/it]
                                                  
{'loss': 0.0417, 'grad_norm': 0.3472529947757721, 'learning_rate': 1.9953316347176486e-06, 'kl': 0.0129, 'entropy': -0.0261, 'ce_loss': 0.0183, 'epoch': 0.25}

  6%|▋         | 21/336 [06:33<1:27:56, 16.75s/it]
  7%|▋         | 22/336 [06:48<1:26:10, 16.47s/it]
                                                  
{'loss': 0.0332, 'grad_norm': 0.27452847361564636, 'learning_rate': 1.994352201771236e-06, 'kl': 0.0102, 'entropy': -0.0344, 'ce_loss': 0.02, 'epoch': 0.26}

  7%|▋         | 22/336 [06:48<1:26:10, 16.47s/it]
  7%|▋         | 23/336 [07:04<1:24:22, 16.18s/it]
                                                  
{'loss': 0.0279, 'grad_norm': 0.2443157136440277, 'learning_rate': 1.993279857237133e-06, 'kl': 0.0176, 'entropy': -0.0262, 'ce_loss': 0.0244, 'epoch': 0.27}

  7%|▋         | 23/336 [07:04<1:24:22, 16.18s/it]
  7%|▋         | 24/336 [07:20<1:23:16, 16.01s/it]
                                                  
{'loss': 0.0352, 'grad_norm': 0.27312204241752625, 'learning_rate': 1.9921147013144777e-06, 'kl': 0.0198, 'entropy': -0.0762, 'ce_loss': 0.0158, 'epoch': 0.29}

  7%|▋         | 24/336 [07:20<1:23:16, 16.01s/it]
  7%|▋         | 25/336 [07:37<1:24:34, 16.32s/it]
                                                  
{'loss': 0.0276, 'grad_norm': 0.22963714599609375, 'learning_rate': 1.9908568428746405e-06, 'kl': 0.0078, 'entropy': -0.0659, 'ce_loss': 0.0259, 'epoch': 0.3}

  7%|▋         | 25/336 [07:37<1:24:34, 16.32s/it]
  8%|▊         | 26/336 [07:52<1:23:38, 16.19s/it]
                                                  
{'loss': 0.0262, 'grad_norm': 0.17868177592754364, 'learning_rate': 1.989506399451051e-06, 'kl': 0.0114, 'entropy': -0.0557, 'ce_loss': 0.017, 'epoch': 0.31}

  8%|▊         | 26/336 [07:52<1:23:38, 16.19s/it]
  8%|▊         | 27/336 [08:09<1:23:10, 16.15s/it]
                                                  
{'loss': 0.0335, 'grad_norm': 0.2423129379749298, 'learning_rate': 1.9880634972282166e-06, 'kl': 0.0082, 'entropy': -0.0172, 'ce_loss': 0.025, 'epoch': 0.32}

  8%|▊         | 27/336 [08:09<1:23:10, 16.15s/it]
  8%|▊         | 28/336 [08:28<1:27:23, 17.02s/it]
                                                  
{'loss': 0.0289, 'grad_norm': 0.21936891973018646, 'learning_rate': 1.986528271029931e-06, 'kl': 0.0105, 'entropy': -0.019, 'ce_loss': 0.0328, 'epoch': 0.33}

  8%|▊         | 28/336 [08:28<1:27:23, 17.02s/it]
  9%|▊         | 29/336 [08:47<1:30:04, 17.60s/it]
                                                  
{'loss': 0.0287, 'grad_norm': 0.2251005619764328, 'learning_rate': 1.984900864306677e-06, 'kl': 0.0165, 'entropy': -0.0347, 'ce_loss': 0.0173, 'epoch': 0.35}

  9%|▊         | 29/336 [08:47<1:30:04, 17.60s/it]
  9%|▉         | 30/336 [09:02<1:26:24, 16.94s/it]
                                                  
{'loss': 0.0312, 'grad_norm': 0.2545115649700165, 'learning_rate': 1.9831814291222233e-06, 'kl': 0.0217, 'entropy': 0.028, 'ce_loss': 0.0162, 'epoch': 0.36}

  9%|▉         | 30/336 [09:02<1:26:24, 16.94s/it]
  9%|▉         | 31/336 [09:21<1:28:49, 17.47s/it]
                                                  
{'loss': 0.0306, 'grad_norm': 0.24097831547260284, 'learning_rate': 1.981370126139413e-06, 'kl': 0.0064, 'entropy': -0.0193, 'ce_loss': 0.0095, 'epoch': 0.37}

  9%|▉         | 31/336 [09:21<1:28:49, 17.47s/it]
 10%|▉         | 32/336 [09:42<1:34:47, 18.71s/it]
                                                  
{'loss': 0.021, 'grad_norm': 0.16445204615592957, 'learning_rate': 1.979467124605156e-06, 'kl': 0.0058, 'entropy': -0.0554, 'ce_loss': 0.0099, 'epoch': 0.38}

 10%|▉         | 32/336 [09:42<1:34:47, 18.71s/it]
 10%|▉         | 33/336 [09:59<1:31:19, 18.08s/it]
                                                  
{'loss': 0.0309, 'grad_norm': 0.325605571269989, 'learning_rate': 1.977472602334609e-06, 'kl': 0.0178, 'entropy': -0.0112, 'ce_loss': 0.0173, 'epoch': 0.39}

 10%|▉         | 33/336 [09:59<1:31:19, 18.08s/it]
 10%|█         | 34/336 [10:18<1:33:00, 18.48s/it]
                                                  
{'loss': 0.0272, 'grad_norm': 0.19901347160339355, 'learning_rate': 1.975386745694565e-06, 'kl': 0.0238, 'entropy': -0.03, 'ce_loss': 0.0364, 'epoch': 0.4}

 10%|█         | 34/336 [10:18<1:33:00, 18.48s/it]
 10%|█         | 35/336 [10:34<1:29:11, 17.78s/it]
                                                  
{'loss': 0.0256, 'grad_norm': 0.19283317029476166, 'learning_rate': 1.9732097495860385e-06, 'kl': 0.0128, 'entropy': -0.0212, 'ce_loss': 0.0127, 'epoch': 0.42}

 10%|█         | 35/336 [10:34<1:29:11, 17.78s/it]
 11%|█         | 36/336 [10:54<1:31:04, 18.21s/it]
                                                  
{'loss': 0.0316, 'grad_norm': 0.21884019672870636, 'learning_rate': 1.970941817426052e-06, 'kl': 0.0247, 'entropy': -0.0505, 'ce_loss': 0.0116, 'epoch': 0.43}

 11%|█         | 36/336 [10:54<1:31:04, 18.21s/it]
 11%|█         | 37/336 [11:10<1:27:48, 17.62s/it]
                                                  
{'loss': 0.0245, 'grad_norm': 0.19287759065628052, 'learning_rate': 1.968583161128631e-06, 'kl': 0.0091, 'entropy': -0.0093, 'ce_loss': 0.0204, 'epoch': 0.44}

 11%|█         | 37/336 [11:10<1:27:48, 17.62s/it]
 11%|█▏        | 38/336 [11:27<1:26:24, 17.40s/it]
                                                  
{'loss': 0.0235, 'grad_norm': 0.18665923178195953, 'learning_rate': 1.9661340010850024e-06, 'kl': 0.0187, 'entropy': -0.0312, 'ce_loss': 0.0144, 'epoch': 0.45}

 11%|█▏        | 38/336 [11:27<1:26:24, 17.40s/it]
 12%|█▏        | 39/336 [11:44<1:25:53, 17.35s/it]
                                                  
{'loss': 0.0296, 'grad_norm': 0.2117118537425995, 'learning_rate': 1.9635945661430005e-06, 'kl': 0.0104, 'entropy': -0.0732, 'ce_loss': 0.0147, 'epoch': 0.46}

 12%|█▏        | 39/336 [11:44<1:25:53, 17.35s/it]
 12%|█▏        | 40/336 [12:00<1:23:33, 16.94s/it]
                                                  
{'loss': 0.0331, 'grad_norm': 0.23639369010925293, 'learning_rate': 1.960965093585684e-06, 'kl': 0.0085, 'entropy': -0.0723, 'ce_loss': 0.0215, 'epoch': 0.48}

 12%|█▏        | 40/336 [12:00<1:23:33, 16.94s/it]
 12%|█▏        | 41/336 [12:19<1:26:32, 17.60s/it]
                                                  
{'loss': 0.0314, 'grad_norm': 0.2341509610414505, 'learning_rate': 1.9582458291091663e-06, 'kl': 0.0155, 'entropy': -0.0095, 'ce_loss': 0.0126, 'epoch': 0.49}

 12%|█▏        | 41/336 [12:19<1:26:32, 17.60s/it]
 12%|█▎        | 42/336 [12:38<1:28:42, 18.10s/it]
                                                  
{'loss': 0.0256, 'grad_norm': 0.20089305937290192, 'learning_rate': 1.9554370267996535e-06, 'kl': 0.0078, 'entropy': -0.0474, 'ce_loss': 0.0218, 'epoch': 0.5}

 12%|█▎        | 42/336 [12:38<1:28:42, 18.10s/it]
 13%|█▎        | 43/336 [12:54<1:25:06, 17.43s/it]
                                                  
{'loss': 0.03, 'grad_norm': 0.2235359102487564, 'learning_rate': 1.952538949109708e-06, 'kl': 0.0065, 'entropy': -0.0253, 'ce_loss': 0.0096, 'epoch': 0.51}

 13%|█▎        | 43/336 [12:54<1:25:06, 17.43s/it]
 13%|█▎        | 44/336 [13:13<1:27:23, 17.96s/it]
                                                  
{'loss': 0.0239, 'grad_norm': 0.16896812617778778, 'learning_rate': 1.94955186683372e-06, 'kl': 0.0215, 'entropy': -0.0181, 'ce_loss': 0.0085, 'epoch': 0.52}

 13%|█▎        | 44/336 [13:13<1:27:23, 17.96s/it]
 13%|█▎        | 45/336 [13:33<1:29:03, 18.36s/it]
                                                  
{'loss': 0.0293, 'grad_norm': 0.21316233277320862, 'learning_rate': 1.94647605908261e-06, 'kl': 0.0076, 'entropy': -0.0325, 'ce_loss': 0.0095, 'epoch': 0.54}

 13%|█▎        | 45/336 [13:33<1:29:03, 18.36s/it]
 14%|█▎        | 46/336 [13:49<1:25:22, 17.66s/it]
                                                  
{'loss': 0.0285, 'grad_norm': 0.25628453493118286, 'learning_rate': 1.943311813257743e-06, 'kl': 0.0093, 'entropy': -0.0471, 'ce_loss': 0.0218, 'epoch': 0.55}

 14%|█▎        | 46/336 [13:49<1:25:22, 17.66s/it]
 14%|█▍        | 47/336 [14:06<1:25:01, 17.65s/it]
                                                  
{'loss': 0.0337, 'grad_norm': 0.2560902535915375, 'learning_rate': 1.9400594250240794e-06, 'kl': 0.0126, 'entropy': -0.0515, 'ce_loss': 0.0283, 'epoch': 0.56}

 14%|█▍        | 47/336 [14:06<1:25:01, 17.65s/it]
 14%|█▍        | 48/336 [14:25<1:26:24, 18.00s/it]
                                                  
{'loss': 0.03, 'grad_norm': 0.22114038467407227, 'learning_rate': 1.9367191982825448e-06, 'kl': 0.0126, 'entropy': -0.064, 'ce_loss': 0.0145, 'epoch': 0.57}

 14%|█▍        | 48/336 [14:25<1:26:24, 18.00s/it]
 15%|█▍        | 49/336 [14:41<1:22:36, 17.27s/it]
                                                  
{'loss': 0.0264, 'grad_norm': 0.19219888746738434, 'learning_rate': 1.9332914451416345e-06, 'kl': 0.0208, 'entropy': 0.032, 'ce_loss': 0.0246, 'epoch': 0.58}

 15%|█▍        | 49/336 [14:41<1:22:36, 17.27s/it]
 15%|█▍        | 50/336 [15:03<1:29:27, 18.77s/it]
                                                  
{'loss': 0.0218, 'grad_norm': 0.16692093014717102, 'learning_rate': 1.929776485888251e-06, 'kl': 0.0085, 'entropy': -0.0498, 'ce_loss': 0.0061, 'epoch': 0.6}

 15%|█▍        | 50/336 [15:03<1:29:27, 18.77s/it]
 15%|█▌        | 51/336 [15:19<1:25:09, 17.93s/it]
                                                  
{'loss': 0.0351, 'grad_norm': 0.27574315667152405, 'learning_rate': 1.9261746489577764e-06, 'kl': 0.0063, 'entropy': -0.0728, 'ce_loss': 0.0241, 'epoch': 0.61}

 15%|█▌        | 51/336 [15:19<1:25:09, 17.93s/it]
 15%|█▌        | 52/336 [15:36<1:22:56, 17.52s/it]
                                                  
{'loss': 0.0316, 'grad_norm': 0.23243726789951324, 'learning_rate': 1.9224862709033824e-06, 'kl': 0.01, 'entropy': 0.0388, 'ce_loss': 0.0108, 'epoch': 0.62}

 15%|█▌        | 52/336 [15:36<1:22:56, 17.52s/it]
 16%|█▌        | 53/336 [15:52<1:20:29, 17.07s/it]
                                                  
{'loss': 0.039, 'grad_norm': 0.2668706774711609, 'learning_rate': 1.918711696364584e-06, 'kl': 0.0427, 'entropy': -0.2295, 'ce_loss': 0.0267, 'epoch': 0.63}

 16%|█▌        | 53/336 [15:52<1:20:29, 17.07s/it]
 16%|█▌        | 54/336 [16:07<1:18:02, 16.60s/it]
                                                  
{'loss': 0.0326, 'grad_norm': 0.25586774945259094, 'learning_rate': 1.914851278035038e-06, 'kl': 0.0059, 'entropy': -0.0284, 'ce_loss': 0.027, 'epoch': 0.64}

 16%|█▌        | 54/336 [16:07<1:18:02, 16.60s/it]
 16%|█▋        | 55/336 [16:23<1:16:27, 16.33s/it]
                                                  
{'loss': 0.0305, 'grad_norm': 0.23466119170188904, 'learning_rate': 1.910905376629585e-06, 'kl': 0.032, 'entropy': -0.014, 'ce_loss': 0.0443, 'epoch': 0.65}

 16%|█▋        | 55/336 [16:23<1:16:27, 16.33s/it]
 17%|█▋        | 56/336 [16:39<1:15:45, 16.23s/it]
                                                  
{'loss': 0.0302, 'grad_norm': 0.21334969997406006, 'learning_rate': 1.9068743608505452e-06, 'kl': 0.0036, 'entropy': -0.0138, 'ce_loss': 0.0075, 'epoch': 0.67}

 17%|█▋        | 56/336 [16:39<1:15:45, 16.23s/it]
 17%|█▋        | 57/336 [17:01<1:23:08, 17.88s/it]
                                                  
{'loss': 0.031, 'grad_norm': 0.25325629115104675, 'learning_rate': 1.902758607353269e-06, 'kl': 0.0186, 'entropy': 0.0009, 'ce_loss': 0.0097, 'epoch': 0.68}

 17%|█▋        | 57/336 [17:01<1:23:08, 17.88s/it]
 17%|█▋        | 58/336 [17:16<1:19:51, 17.24s/it]
                                                  
{'loss': 0.0286, 'grad_norm': 0.20371510088443756, 'learning_rate': 1.8985585007109388e-06, 'kl': 0.011, 'entropy': -0.0403, 'ce_loss': 0.0096, 'epoch': 0.69}

 17%|█▋        | 58/336 [17:16<1:19:51, 17.24s/it]
 18%|█▊        | 59/336 [17:35<1:22:13, 17.81s/it]
                                                  
{'loss': 0.0275, 'grad_norm': 0.21284231543540955, 'learning_rate': 1.8942744333786395e-06, 'kl': 0.0066, 'entropy': 0.0228, 'ce_loss': 0.0208, 'epoch': 0.7}

 18%|█▊        | 59/336 [17:35<1:22:13, 17.81s/it]
 18%|█▊        | 60/336 [17:51<1:18:59, 17.17s/it]
                                                  
{'loss': 0.037, 'grad_norm': 0.23954221606254578, 'learning_rate': 1.8899068056566838e-06, 'kl': 0.0226, 'entropy': 0.0033, 'ce_loss': 0.0262, 'epoch': 0.71}

 18%|█▊        | 60/336 [17:51<1:18:59, 17.17s/it]
 18%|█▊        | 61/336 [18:13<1:24:40, 18.47s/it]
                                                  
{'loss': 0.028, 'grad_norm': 0.2091853767633438, 'learning_rate': 1.8854560256532098e-06, 'kl': 0.0012, 'entropy': -0.0064, 'ce_loss': 0.0065, 'epoch': 0.73}

 18%|█▊        | 61/336 [18:13<1:24:40, 18.47s/it]
 18%|█▊        | 62/336 [18:29<1:21:50, 17.92s/it]
                                                  
{'loss': 0.0293, 'grad_norm': 0.21565210819244385, 'learning_rate': 1.8809225092460485e-06, 'kl': 0.0036, 'entropy': 0.013, 'ce_loss': 0.0099, 'epoch': 0.74}

 18%|█▊        | 62/336 [18:29<1:21:50, 17.92s/it]
 19%|█▉        | 63/336 [18:48<1:22:59, 18.24s/it]
                                                  
{'loss': 0.0264, 'grad_norm': 0.251737505197525, 'learning_rate': 1.8763066800438634e-06, 'kl': 0.0179, 'entropy': -0.0498, 'ce_loss': 0.0308, 'epoch': 0.75}

 19%|█▉        | 63/336 [18:48<1:22:59, 18.24s/it]
 19%|█▉        | 64/336 [19:04<1:18:50, 17.39s/it]
                                                  
{'loss': 0.0338, 'grad_norm': 0.2660210132598877, 'learning_rate': 1.8716089693465693e-06, 'kl': 0.0128, 'entropy': -0.0398, 'ce_loss': 0.0159, 'epoch': 0.76}

 19%|█▉        | 64/336 [19:04<1:18:50, 17.39s/it]
 19%|█▉        | 65/336 [19:19<1:16:02, 16.84s/it]
                                                  
{'loss': 0.0371, 'grad_norm': 0.2840781807899475, 'learning_rate': 1.8668298161050306e-06, 'kl': 0.0161, 'entropy': -0.0286, 'ce_loss': 0.0124, 'epoch': 0.77}

 19%|█▉        | 65/336 [19:19<1:16:02, 16.84s/it]
 20%|█▉        | 66/336 [19:35<1:14:24, 16.53s/it]
                                                  
{'loss': 0.03, 'grad_norm': 0.20204216241836548, 'learning_rate': 1.861969666880049e-06, 'kl': 0.0124, 'entropy': -0.042, 'ce_loss': 0.019, 'epoch': 0.79}

 20%|█▉        | 66/336 [19:35<1:14:24, 16.53s/it]
 20%|█▉        | 67/336 [19:51<1:13:19, 16.35s/it]
                                                  
{'loss': 0.0285, 'grad_norm': 0.19628241658210754, 'learning_rate': 1.8570289758006343e-06, 'kl': 0.0211, 'entropy': -0.0074, 'ce_loss': 0.0238, 'epoch': 0.8}

 20%|█▉        | 67/336 [19:51<1:13:19, 16.35s/it]
 20%|██        | 68/336 [20:07<1:12:15, 16.18s/it]
                                                  
{'loss': 0.0319, 'grad_norm': 0.23093868792057037, 'learning_rate': 1.8520082045215717e-06, 'kl': 0.0242, 'entropy': 0.02, 'ce_loss': 0.0311, 'epoch': 0.81}

 20%|██        | 68/336 [20:07<1:12:15, 16.18s/it]
 21%|██        | 69/336 [20:25<1:14:46, 16.80s/it]
                                                  
{'loss': 0.0301, 'grad_norm': 0.2109888792037964, 'learning_rate': 1.846907822180286e-06, 'kl': 0.0103, 'entropy': -0.1167, 'ce_loss': 0.0137, 'epoch': 0.82}

 21%|██        | 69/336 [20:25<1:14:46, 16.80s/it]
 21%|██        | 70/336 [20:41<1:12:49, 16.43s/it]
                                                  
{'loss': 0.0263, 'grad_norm': 0.18644583225250244, 'learning_rate': 1.8417283053530043e-06, 'kl': 0.0085, 'entropy': 0.0069, 'ce_loss': 0.0126, 'epoch': 0.83}

 21%|██        | 70/336 [20:41<1:12:49, 16.43s/it]
 21%|██        | 71/336 [20:59<1:15:34, 17.11s/it]
                                                  
{'loss': 0.0255, 'grad_norm': 0.2035277783870697, 'learning_rate': 1.8364701380102264e-06, 'kl': 0.0135, 'entropy': -0.0184, 'ce_loss': 0.0113, 'epoch': 0.85}

 21%|██        | 71/336 [20:59<1:15:34, 17.11s/it]
 21%|██▏       | 72/336 [21:15<1:13:08, 16.62s/it]
                                                  
{'loss': 0.0318, 'grad_norm': 0.22000524401664734, 'learning_rate': 1.8311338114715027e-06, 'kl': 0.0022, 'entropy': -0.0266, 'ce_loss': 0.0074, 'epoch': 0.86}

 21%|██▏       | 72/336 [21:15<1:13:08, 16.62s/it]
 22%|██▏       | 73/336 [21:31<1:12:12, 16.47s/it]
                                                  
{'loss': 0.0326, 'grad_norm': 0.23698453605175018, 'learning_rate': 1.825719824359524e-06, 'kl': 0.0059, 'entropy': -0.0469, 'ce_loss': 0.0155, 'epoch': 0.87}

 22%|██▏       | 73/336 [21:31<1:12:12, 16.47s/it]
 22%|██▏       | 74/336 [21:49<1:14:04, 16.97s/it]
                                                  
{'loss': 0.027, 'grad_norm': 0.18671374022960663, 'learning_rate': 1.8202286825535329e-06, 'kl': 0.0046, 'entropy': -0.0625, 'ce_loss': 0.0096, 'epoch': 0.88}

 22%|██▏       | 74/336 [21:49<1:14:04, 16.97s/it]
 22%|██▏       | 75/336 [22:06<1:13:40, 16.94s/it]
                                                  
{'loss': 0.0276, 'grad_norm': 0.22594326734542847, 'learning_rate': 1.814660899142053e-06, 'kl': 0.0148, 'entropy': -0.0654, 'ce_loss': 0.0119, 'epoch': 0.89}

 22%|██▏       | 75/336 [22:06<1:13:40, 16.94s/it]
 23%|██▎       | 76/336 [22:22<1:11:52, 16.58s/it]
                                                  
{'loss': 0.0267, 'grad_norm': 0.2022831290960312, 'learning_rate': 1.8090169943749474e-06, 'kl': 0.0457, 'entropy': -0.0771, 'ce_loss': 0.0215, 'epoch': 0.9}

 23%|██▎       | 76/336 [22:22<1:11:52, 16.58s/it]
 23%|██▎       | 77/336 [22:44<1:18:36, 18.21s/it]
                                                  
{'loss': 0.0269, 'grad_norm': 0.18786337971687317, 'learning_rate': 1.8032974956148062e-06, 'kl': 0.013, 'entropy': 0.0186, 'ce_loss': 0.0141, 'epoch': 0.92}

 23%|██▎       | 77/336 [22:44<1:18:36, 18.21s/it]
 23%|██▎       | 78/336 [23:04<1:20:45, 18.78s/it]
                                                  
{'loss': 0.0247, 'grad_norm': 0.1901751309633255, 'learning_rate': 1.7975029372876705e-06, 'kl': 0.002, 'entropy': -0.0408, 'ce_loss': 0.0094, 'epoch': 0.93}

 23%|██▎       | 78/336 [23:04<1:20:45, 18.78s/it]
 24%|██▎       | 79/336 [23:19<1:16:28, 17.86s/it]
                                                  
{'loss': 0.0325, 'grad_norm': 0.23225203156471252, 'learning_rate': 1.7916338608330956e-06, 'kl': 0.0014, 'entropy': -0.0278, 'ce_loss': 0.0159, 'epoch': 0.94}

 24%|██▎       | 79/336 [23:19<1:16:28, 17.86s/it]
 24%|██▍       | 80/336 [23:38<1:17:19, 18.12s/it]
                                                  
{'loss': 0.0356, 'grad_norm': 0.24554163217544556, 'learning_rate': 1.78569081465356e-06, 'kl': 0.0081, 'entropy': -0.0449, 'ce_loss': 0.0146, 'epoch': 0.95}

 24%|██▍       | 80/336 [23:38<1:17:19, 18.12s/it]
 24%|██▍       | 81/336 [23:54<1:13:44, 17.35s/it]
                                                  
{'loss': 0.044, 'grad_norm': 0.308929443359375, 'learning_rate': 1.7796743540632221e-06, 'kl': 0.0096, 'entropy': -0.0469, 'ce_loss': 0.0275, 'epoch': 0.96}

 24%|██▍       | 81/336 [23:54<1:13:44, 17.35s/it]
 24%|██▍       | 82/336 [24:18<1:22:51, 19.57s/it]
                                                  
{'loss': 0.022, 'grad_norm': 0.1706738919019699, 'learning_rate': 1.7735850412360328e-06, 'kl': 0.0157, 'entropy': -0.0369, 'ce_loss': 0.0045, 'epoch': 0.98}

 24%|██▍       | 82/336 [24:18<1:22:51, 19.57s/it]
 25%|██▍       | 83/336 [24:37<1:21:39, 19.37s/it]
                                                  
{'loss': 0.0262, 'grad_norm': 0.1916491985321045, 'learning_rate': 1.7674234451532063e-06, 'kl': 0.027, 'entropy': 0.051, 'ce_loss': 0.0224, 'epoch': 0.99}

 25%|██▍       | 83/336 [24:37<1:21:39, 19.37s/it]
 25%|██▌       | 84/336 [24:53<1:16:50, 18.29s/it]
                                                  
{'loss': 0.0277, 'grad_norm': 0.22529131174087524, 'learning_rate': 1.7611901415500533e-06, 'kl': 0.0033, 'entropy': -0.0476, 'ce_loss': 0.0147, 'epoch': 1.0}

 25%|██▌       | 84/336 [24:53<1:16:50, 18.29s/it]
 25%|██▌       | 85/336 [25:10<1:14:41, 17.85s/it]
                                                  
{'loss': 0.0244, 'grad_norm': 0.19255438446998596, 'learning_rate': 1.7548857128621874e-06, 'kl': 0.0228, 'entropy': -0.0388, 'ce_loss': 0.0178, 'epoch': 1.01}

 25%|██▌       | 85/336 [25:10<1:14:41, 17.85s/it]
 26%|██▌       | 86/336 [25:25<1:11:26, 17.15s/it]
                                                  
{'loss': 0.0288, 'grad_norm': 0.2253047525882721, 'learning_rate': 1.748510748171101e-06, 'kl': 0.005, 'entropy': -0.0052, 'ce_loss': 0.017, 'epoch': 1.02}

 26%|██▌       | 86/336 [25:25<1:11:26, 17.15s/it]
 26%|██▌       | 87/336 [25:45<1:13:32, 17.72s/it]
                                                  
{'loss': 0.0226, 'grad_norm': 0.1952665150165558, 'learning_rate': 1.7420658431491222e-06, 'kl': 0.0122, 'entropy': -0.0566, 'ce_loss': 0.0052, 'epoch': 1.04}

 26%|██▌       | 87/336 [25:45<1:13:32, 17.72s/it]
 26%|██▌       | 88/336 [26:00<1:10:49, 17.13s/it]
                                                  
{'loss': 0.0249, 'grad_norm': 0.1836732178926468, 'learning_rate': 1.735551600003755e-06, 'kl': 0.0236, 'entropy': -0.0576, 'ce_loss': 0.0319, 'epoch': 1.05}

 26%|██▌       | 88/336 [26:00<1:10:49, 17.13s/it]
 26%|██▋       | 89/336 [26:16<1:08:54, 16.74s/it]
                                                  
{'loss': 0.0193, 'grad_norm': 0.17091220617294312, 'learning_rate': 1.7289686274214115e-06, 'kl': 0.0133, 'entropy': -0.0378, 'ce_loss': 0.0301, 'epoch': 1.06}

 26%|██▋       | 89/336 [26:16<1:08:54, 16.74s/it]
 27%|██▋       | 90/336 [26:37<1:13:46, 17.99s/it]
                                                  
{'loss': 0.017, 'grad_norm': 0.12661834061145782, 'learning_rate': 1.722317540510534e-06, 'kl': 0.0194, 'entropy': -0.1187, 'ce_loss': 0.0205, 'epoch': 1.07}

 27%|██▋       | 90/336 [26:37<1:13:46, 17.99s/it]
 27%|██▋       | 91/336 [26:54<1:11:56, 17.62s/it]
                                                  
{'loss': 0.0218, 'grad_norm': 0.16420775651931763, 'learning_rate': 1.715598960744121e-06, 'kl': 0.0197, 'entropy': -0.0281, 'ce_loss': 0.012, 'epoch': 1.08}

 27%|██▋       | 91/336 [26:54<1:11:56, 17.62s/it]
 27%|██▋       | 92/336 [27:13<1:13:19, 18.03s/it]
                                                  
{'loss': 0.0213, 'grad_norm': 0.1939077079296112, 'learning_rate': 1.7088135159016582e-06, 'kl': 0.051, 'entropy': -0.0342, 'ce_loss': 0.0058, 'epoch': 1.1}

 27%|██▋       | 92/336 [27:13<1:13:19, 18.03s/it]
 28%|██▊       | 93/336 [27:29<1:10:51, 17.50s/it]
                                                  
{'loss': 0.0209, 'grad_norm': 0.17347100377082825, 'learning_rate': 1.7019618400104569e-06, 'kl': 0.0172, 'entropy': -0.0483, 'ce_loss': 0.01, 'epoch': 1.11}

 28%|██▊       | 93/336 [27:29<1:10:51, 17.50s/it]
 28%|██▊       | 94/336 [27:47<1:11:14, 17.66s/it]
                                                  
{'loss': 0.0203, 'grad_norm': 0.17469847202301025, 'learning_rate': 1.6950445732864126e-06, 'kl': 0.0179, 'entropy': -0.0645, 'ce_loss': 0.0186, 'epoch': 1.12}

 28%|██▊       | 94/336 [27:47<1:11:14, 17.66s/it]
 28%|██▊       | 95/336 [28:06<1:12:08, 17.96s/it]
                                                  
{'loss': 0.0287, 'grad_norm': 0.21660856902599335, 'learning_rate': 1.688062362074184e-06, 'kl': 0.0034, 'entropy': -0.019, 'ce_loss': 0.0076, 'epoch': 1.13}

 28%|██▊       | 95/336 [28:06<1:12:08, 17.96s/it]
 29%|██▊       | 96/336 [28:22<1:09:41, 17.42s/it]
                                                  
{'loss': 0.025, 'grad_norm': 0.2265072762966156, 'learning_rate': 1.681015858786797e-06, 'kl': 0.0155, 'entropy': -0.0737, 'ce_loss': 0.026, 'epoch': 1.14}

 29%|██▊       | 96/336 [28:22<1:09:41, 17.42s/it]
 29%|██▉       | 97/336 [28:38<1:07:29, 16.94s/it]
                                                  
{'loss': 0.0197, 'grad_norm': 0.1737026572227478, 'learning_rate': 1.6739057218446857e-06, 'kl': 0.0255, 'entropy': 0.0315, 'ce_loss': 0.0164, 'epoch': 1.15}

 29%|██▉       | 97/336 [28:38<1:07:29, 16.94s/it]
 29%|██▉       | 98/336 [28:53<1:05:14, 16.45s/it]
                                                  
{'loss': 0.0252, 'grad_norm': 0.21314632892608643, 'learning_rate': 1.666732615614169e-06, 'kl': 0.0219, 'entropy': -0.031, 'ce_loss': 0.0214, 'epoch': 1.17}

 29%|██▉       | 98/336 [28:53<1:05:14, 16.45s/it]
 29%|██▉       | 99/336 [29:15<1:11:44, 18.16s/it]
                                                  
{'loss': 0.0247, 'grad_norm': 0.2042514830827713, 'learning_rate': 1.6594972103453724e-06, 'kl': 0.032, 'entropy': -0.0283, 'ce_loss': 0.0123, 'epoch': 1.18}

 29%|██▉       | 99/336 [29:15<1:11:44, 18.16s/it]
 30%|██▉       | 100/336 [29:34<1:12:34, 18.45s/it]
                                                   
{'loss': 0.0197, 'grad_norm': 0.19269202649593353, 'learning_rate': 1.6522001821096019e-06, 'kl': 0.0325, 'entropy': -0.0947, 'ce_loss': 0.0156, 'epoch': 1.19}

 30%|██▉       | 100/336 [29:34<1:12:34, 18.45s/it]
 30%|███       | 101/336 [29:50<1:08:56, 17.60s/it]
                                                   
{'loss': 0.0323, 'grad_norm': 0.2899748384952545, 'learning_rate': 1.6448422127361705e-06, 'kl': 0.0195, 'entropy': -0.0148, 'ce_loss': 0.0249, 'epoch': 1.2}

 30%|███       | 101/336 [29:50<1:08:56, 17.60s/it]
 30%|███       | 102/336 [30:09<1:10:28, 18.07s/it]
                                                   
{'loss': 0.0198, 'grad_norm': 0.17614398896694183, 'learning_rate': 1.6374239897486897e-06, 'kl': 0.0184, 'entropy': -0.0347, 'ce_loss': 0.0186, 'epoch': 1.21}

 30%|███       | 102/336 [30:09<1:10:28, 18.07s/it]
 31%|███       | 103/336 [30:25<1:07:29, 17.38s/it]
                                                   
{'loss': 0.0272, 'grad_norm': 0.23129649460315704, 'learning_rate': 1.6299462063008269e-06, 'kl': 0.0576, 'entropy': -0.0552, 'ce_loss': 0.0136, 'epoch': 1.23}

 31%|███       | 103/336 [30:25<1:07:29, 17.38s/it]
 31%|███       | 104/336 [30:40<1:05:08, 16.85s/it]
                                                   
{'loss': 0.0243, 'grad_norm': 0.20547500252723694, 'learning_rate': 1.6224095611115383e-06, 'kl': 0.032, 'entropy': -0.0177, 'ce_loss': 0.0184, 'epoch': 1.24}

 31%|███       | 104/336 [30:40<1:05:08, 16.85s/it]
 31%|███▏      | 105/336 [31:00<1:07:26, 17.52s/it]
                                                   
{'loss': 0.0271, 'grad_norm': 0.226070374250412, 'learning_rate': 1.614814758399781e-06, 'kl': 0.0337, 'entropy': 0.0398, 'ce_loss': 0.0058, 'epoch': 1.25}

 31%|███▏      | 105/336 [31:00<1:07:26, 17.52s/it]
 32%|███▏      | 106/336 [31:15<1:04:45, 16.89s/it]
                                                   
{'loss': 0.025, 'grad_norm': 0.22863830626010895, 'learning_rate': 1.6071625078187112e-06, 'kl': 0.0129, 'entropy': -0.0718, 'ce_loss': 0.0122, 'epoch': 1.26}

 32%|███▏      | 106/336 [31:15<1:04:45, 16.89s/it]
 32%|███▏      | 107/336 [31:35<1:07:49, 17.77s/it]
                                                   
{'loss': 0.0269, 'grad_norm': 0.26432666182518005, 'learning_rate': 1.599453524389374e-06, 'kl': 0.0474, 'entropy': -0.0254, 'ce_loss': 0.0065, 'epoch': 1.27}

 32%|███▏      | 107/336 [31:35<1:07:49, 17.77s/it]
 32%|███▏      | 108/336 [31:50<1:05:10, 17.15s/it]
                                                   
{'loss': 0.0268, 'grad_norm': 0.2612511217594147, 'learning_rate': 1.5916885284338935e-06, 'kl': 0.0265, 'entropy': -0.0654, 'ce_loss': 0.0141, 'epoch': 1.29}

 32%|███▏      | 108/336 [31:50<1:05:10, 17.15s/it]
 32%|███▏      | 109/336 [32:09<1:06:26, 17.56s/it]
                                                   
{'loss': 0.018, 'grad_norm': 0.18477967381477356, 'learning_rate': 1.5838682455081657e-06, 'kl': 0.0082, 'entropy': -0.0437, 'ce_loss': 0.0315, 'epoch': 1.3}

 32%|███▏      | 109/336 [32:09<1:06:26, 17.56s/it]
 33%|███▎      | 110/336 [32:25<1:04:00, 16.99s/it]
                                                   
{'loss': 0.0256, 'grad_norm': 0.20217899978160858, 'learning_rate': 1.5759934063340624e-06, 'kl': 0.0422, 'entropy': -0.1001, 'ce_loss': 0.0087, 'epoch': 1.31}

 33%|███▎      | 110/336 [32:25<1:04:00, 16.99s/it]
 33%|███▎      | 111/336 [32:46<1:08:52, 18.36s/it]
                                                   
{'loss': 0.0197, 'grad_norm': 0.1906185746192932, 'learning_rate': 1.5680647467311555e-06, 'kl': 0.028, 'entropy': -0.0302, 'ce_loss': 0.0081, 'epoch': 1.32}

 33%|███▎      | 111/336 [32:46<1:08:52, 18.36s/it]
 33%|███▎      | 112/336 [33:09<1:13:07, 19.59s/it]
                                                   
{'loss': 0.0204, 'grad_norm': 0.1959102302789688, 'learning_rate': 1.56008300754796e-06, 'kl': 0.0425, 'entropy': 0.0223, 'ce_loss': 0.0217, 'epoch': 1.33}

 33%|███▎      | 112/336 [33:09<1:13:07, 19.59s/it]
 34%|███▎      | 113/336 [33:27<1:11:36, 19.27s/it]
                                                   
{'loss': 0.0197, 'grad_norm': 0.17761990427970886, 'learning_rate': 1.5520489345927094e-06, 'kl': 0.0327, 'entropy': -0.0562, 'ce_loss': 0.0186, 'epoch': 1.35}

 34%|███▎      | 113/336 [33:27<1:11:36, 19.27s/it]
 34%|███▍      | 114/336 [33:43<1:07:53, 18.35s/it]
                                                   
{'loss': 0.0239, 'grad_norm': 0.2238835245370865, 'learning_rate': 1.5439632785636705e-06, 'kl': 0.0192, 'entropy': 0.0053, 'ce_loss': 0.0134, 'epoch': 1.36}

 34%|███▍      | 114/336 [33:43<1:07:53, 18.35s/it]
 34%|███▍      | 115/336 [33:59<1:04:39, 17.55s/it]
                                                   
{'loss': 0.0272, 'grad_norm': 0.22165977954864502, 'learning_rate': 1.5358267949789964e-06, 'kl': 0.0114, 'entropy': -0.0564, 'ce_loss': 0.0137, 'epoch': 1.37}

 34%|███▍      | 115/336 [33:59<1:04:39, 17.55s/it]
 35%|███▍      | 116/336 [34:15<1:02:12, 16.97s/it]
                                                   
{'loss': 0.0264, 'grad_norm': 0.23063895106315613, 'learning_rate': 1.5276402441061327e-06, 'kl': 0.0405, 'entropy': -0.0654, 'ce_loss': 0.019, 'epoch': 1.38}

 35%|███▍      | 116/336 [34:15<1:02:12, 16.97s/it]
 35%|███▍      | 117/336 [34:30<1:00:38, 16.61s/it]
                                                   
{'loss': 0.0233, 'grad_norm': 0.20103129744529724, 'learning_rate': 1.5194043908907772e-06, 'kl': 0.0231, 'entropy': 0.0317, 'ce_loss': 0.025, 'epoch': 1.39}

 35%|███▍      | 117/336 [34:30<1:00:38, 16.61s/it]
 35%|███▌      | 118/336 [34:46<59:10, 16.29s/it]  
                                                 
{'loss': 0.0265, 'grad_norm': 0.22980648279190063, 'learning_rate': 1.5111200048854054e-06, 'kl': 0.0067, 'entropy': -0.0408, 'ce_loss': 0.0121, 'epoch': 1.4}

 35%|███▌      | 118/336 [34:46<59:10, 16.29s/it]
 35%|███▌      | 119/336 [35:02<58:07, 16.07s/it]
                                                 
{'loss': 0.0295, 'grad_norm': 0.24733252823352814, 'learning_rate': 1.5027878601773632e-06, 'kl': 0.0167, 'entropy': -0.054, 'ce_loss': 0.0217, 'epoch': 1.42}

 35%|███▌      | 119/336 [35:02<58:07, 16.07s/it]
 36%|███▌      | 120/336 [35:17<57:25, 15.95s/it]
                                                 
{'loss': 0.0181, 'grad_norm': 0.1668197065591812, 'learning_rate': 1.494408735316537e-06, 'kl': 0.0227, 'entropy': -0.009, 'ce_loss': 0.0184, 'epoch': 1.43}

 36%|███▌      | 120/336 [35:17<57:25, 15.95s/it]
 36%|███▌      | 121/336 [35:33<57:04, 15.93s/it]
                                                 
{'loss': 0.0273, 'grad_norm': 0.23297691345214844, 'learning_rate': 1.4859834132426058e-06, 'kl': 0.0359, 'entropy': -0.0361, 'ce_loss': 0.0179, 'epoch': 1.44}

 36%|███▌      | 121/336 [35:33<57:04, 15.93s/it]
 36%|███▋      | 122/336 [35:55<1:03:28, 17.80s/it]
                                                   
{'loss': 0.0201, 'grad_norm': 0.1959342062473297, 'learning_rate': 1.4775126812118863e-06, 'kl': 0.0244, 'entropy': 0.0123, 'ce_loss': 0.0129, 'epoch': 1.45}

 36%|███▋      | 122/336 [35:55<1:03:28, 17.80s/it]
 37%|███▋      | 123/336 [36:11<1:01:10, 17.23s/it]
                                                   
{'loss': 0.0227, 'grad_norm': 0.19156020879745483, 'learning_rate': 1.4689973307237686e-06, 'kl': 0.0164, 'entropy': -0.0674, 'ce_loss': 0.0214, 'epoch': 1.46}

 37%|███▋      | 123/336 [36:11<1:01:10, 17.23s/it]
 37%|███▋      | 124/336 [36:31<1:03:45, 18.05s/it]
                                                   
{'loss': 0.02, 'grad_norm': 0.1831812560558319, 'learning_rate': 1.4604381574467614e-06, 'kl': 0.0334, 'entropy': -0.0136, 'ce_loss': 0.0232, 'epoch': 1.48}

 37%|███▋      | 124/336 [36:31<1:03:45, 18.05s/it]
 37%|███▋      | 125/336 [36:47<1:01:23, 17.46s/it]
                                                   
{'loss': 0.0216, 'grad_norm': 0.1912676841020584, 'learning_rate': 1.451835961144145e-06, 'kl': 0.0342, 'entropy': -0.0332, 'ce_loss': 0.0058, 'epoch': 1.49}

 37%|███▋      | 125/336 [36:47<1:01:23, 17.46s/it]
 38%|███▊      | 126/336 [37:03<59:26, 16.99s/it]  
                                                 
{'loss': 0.0242, 'grad_norm': 0.19266380369663239, 'learning_rate': 1.4431915455992414e-06, 'kl': 0.0216, 'entropy': -0.0322, 'ce_loss': 0.0111, 'epoch': 1.5}

 38%|███▊      | 126/336 [37:03<59:26, 16.99s/it]
 38%|███▊      | 127/336 [37:26<1:04:51, 18.62s/it]
                                                   
{'loss': 0.0234, 'grad_norm': 0.19839657843112946, 'learning_rate': 1.4345057185403098e-06, 'kl': 0.0286, 'entropy': -0.0264, 'ce_loss': 0.011, 'epoch': 1.51}

 38%|███▊      | 127/336 [37:26<1:04:51, 18.62s/it]
 38%|███▊      | 128/336 [37:43<1:03:04, 18.19s/it]
                                                   
{'loss': 0.0309, 'grad_norm': 0.25261247158050537, 'learning_rate': 1.4257792915650725e-06, 'kl': 0.0354, 'entropy': 0.001, 'ce_loss': 0.0211, 'epoch': 1.52}

 38%|███▊      | 128/336 [37:43<1:03:04, 18.19s/it]
 38%|███▊      | 129/336 [38:02<1:04:08, 18.59s/it]
                                                   
{'loss': 0.0263, 'grad_norm': 0.22662681341171265, 'learning_rate': 1.4170130800648812e-06, 'kl': 0.0101, 'entropy': -0.014, 'ce_loss': 0.0086, 'epoch': 1.54}

 38%|███▊      | 129/336 [38:02<1:04:08, 18.59s/it]
 39%|███▊      | 130/336 [38:21<1:04:16, 18.72s/it]
                                                   
{'loss': 0.0201, 'grad_norm': 0.17634278535842896, 'learning_rate': 1.408207903148525e-06, 'kl': 0.0121, 'entropy': -0.0188, 'ce_loss': 0.0124, 'epoch': 1.55}

 39%|███▊      | 130/336 [38:21<1:04:16, 18.72s/it]
 39%|███▉      | 131/336 [38:37<1:00:55, 17.83s/it]
                                                   
{'loss': 0.0216, 'grad_norm': 0.18929998576641083, 'learning_rate': 1.3993645835656952e-06, 'kl': 0.015, 'entropy': -0.1147, 'ce_loss': 0.0246, 'epoch': 1.56}

 39%|███▉      | 131/336 [38:37<1:00:55, 17.83s/it]
 39%|███▉      | 132/336 [38:55<1:00:57, 17.93s/it]
                                                   
{'loss': 0.0238, 'grad_norm': 0.2257225066423416, 'learning_rate': 1.3904839476301088e-06, 'kl': 0.0347, 'entropy': -0.0623, 'ce_loss': 0.0086, 'epoch': 1.57}

 39%|███▉      | 132/336 [38:55<1:00:57, 17.93s/it]
 40%|███▉      | 133/336 [39:18<1:05:38, 19.40s/it]
                                                   
{'loss': 0.0164, 'grad_norm': 0.16117718815803528, 'learning_rate': 1.3815668251422953e-06, 'kl': 0.007, 'entropy': 0.019, 'ce_loss': 0.0245, 'epoch': 1.58}

 40%|███▉      | 133/336 [39:18<1:05:38, 19.40s/it]
 40%|███▉      | 134/336 [39:37<1:05:10, 19.36s/it]
                                                   
{'loss': 0.0174, 'grad_norm': 0.1562952846288681, 'learning_rate': 1.3726140493120637e-06, 'kl': 0.04, 'entropy': -0.0471, 'ce_loss': 0.0057, 'epoch': 1.6}

 40%|███▉      | 134/336 [39:37<1:05:10, 19.36s/it]
 40%|████      | 135/336 [39:57<1:04:59, 19.40s/it]
                                                   
{'loss': 0.0224, 'grad_norm': 0.26663967967033386, 'learning_rate': 1.363626456680647e-06, 'kl': 0.0076, 'entropy': -0.0405, 'ce_loss': 0.0135, 'epoch': 1.61}

 40%|████      | 135/336 [39:57<1:04:59, 19.40s/it]
 40%|████      | 136/336 [40:16<1:04:02, 19.21s/it]
                                                   
{'loss': 0.0231, 'grad_norm': 0.19622810184955597, 'learning_rate': 1.3546048870425354e-06, 'kl': 0.0339, 'entropy': -0.0459, 'ce_loss': 0.0174, 'epoch': 1.62}

 40%|████      | 136/336 [40:16<1:04:02, 19.21s/it]
 41%|████      | 137/336 [40:35<1:03:45, 19.22s/it]
                                                   
{'loss': 0.0223, 'grad_norm': 0.22339412569999695, 'learning_rate': 1.3455501833670087e-06, 'kl': 0.024, 'entropy': -0.0359, 'ce_loss': 0.0137, 'epoch': 1.63}

 41%|████      | 137/336 [40:35<1:03:45, 19.22s/it]
 41%|████      | 138/336 [40:51<59:57, 18.17s/it]  
                                                 
{'loss': 0.025, 'grad_norm': 0.23139812052249908, 'learning_rate': 1.336463191719367e-06, 'kl': 0.0225, 'entropy': 0.0078, 'ce_loss': 0.0123, 'epoch': 1.64}

 41%|████      | 138/336 [40:51<59:57, 18.17s/it]
 41%|████▏     | 139/336 [41:06<57:08, 17.40s/it]
                                                 
{'loss': 0.0303, 'grad_norm': 0.25350722670555115, 'learning_rate': 1.3273447611818766e-06, 'kl': 0.0195, 'entropy': -0.0481, 'ce_loss': 0.0102, 'epoch': 1.65}

 41%|████▏     | 139/336 [41:06<57:08, 17.40s/it]
 42%|████▏     | 140/336 [41:23<56:47, 17.38s/it]
                                                 
{'loss': 0.0249, 'grad_norm': 0.21892692148685455, 'learning_rate': 1.3181957437744332e-06, 'kl': 0.0237, 'entropy': -0.0242, 'ce_loss': 0.0229, 'epoch': 1.67}

 42%|████▏     | 140/336 [41:23<56:47, 17.38s/it]
 42%|████▏     | 141/336 [41:42<57:39, 17.74s/it]
                                                 
{'loss': 0.0177, 'grad_norm': 0.16795045137405396, 'learning_rate': 1.3090169943749473e-06, 'kl': 0.0464, 'entropy': 0.0137, 'ce_loss': 0.0068, 'epoch': 1.68}

 42%|████▏     | 141/336 [41:42<57:39, 17.74s/it]
 42%|████▏     | 142/336 [41:58<55:48, 17.26s/it]
                                                 
{'loss': 0.0307, 'grad_norm': 0.24835337698459625, 'learning_rate': 1.2998093706394675e-06, 'kl': 0.0098, 'entropy': -0.0178, 'ce_loss': 0.0141, 'epoch': 1.69}

 42%|████▏     | 142/336 [41:58<55:48, 17.26s/it]
 43%|████▎     | 143/336 [42:16<56:18, 17.51s/it]
                                                 
{'loss': 0.0179, 'grad_norm': 0.15057280659675598, 'learning_rate': 1.2905737329220392e-06, 'kl': 0.0354, 'entropy': -0.125, 'ce_loss': 0.0115, 'epoch': 1.7}

 43%|████▎     | 143/336 [42:16<56:18, 17.51s/it]
 43%|████▎     | 144/336 [42:35<57:23, 17.94s/it]
                                                 
{'loss': 0.0217, 'grad_norm': 0.20697814226150513, 'learning_rate': 1.2813109441943164e-06, 'kl': 0.0132, 'entropy': -0.0342, 'ce_loss': 0.0128, 'epoch': 1.71}

 43%|████▎     | 144/336 [42:35<57:23, 17.94s/it]
 43%|████▎     | 145/336 [42:51<54:51, 17.24s/it]
                                                 
{'loss': 0.0272, 'grad_norm': 0.23005861043930054, 'learning_rate': 1.2720218699649241e-06, 'kl': 0.0447, 'entropy': -0.0659, 'ce_loss': 0.0129, 'epoch': 1.73}

 43%|████▎     | 145/336 [42:51<54:51, 17.24s/it]
 43%|████▎     | 146/336 [43:10<56:23, 17.81s/it]
                                                 
{'loss': 0.0228, 'grad_norm': 0.20512734353542328, 'learning_rate': 1.262707378198587e-06, 'kl': 0.0148, 'entropy': -0.022, 'ce_loss': 0.0205, 'epoch': 1.74}

 43%|████▎     | 146/336 [43:10<56:23, 17.81s/it]
 44%|████▍     | 147/336 [43:25<53:50, 17.09s/it]
                                                 
{'loss': 0.0195, 'grad_norm': 0.18244057893753052, 'learning_rate': 1.2533683392350262e-06, 'kl': 0.0074, 'entropy': 0.0103, 'ce_loss': 0.0126, 'epoch': 1.75}

 44%|████▍     | 147/336 [43:25<53:50, 17.09s/it]
 44%|████▍     | 148/336 [43:45<55:46, 17.80s/it]
                                                 
{'loss': 0.0204, 'grad_norm': 0.18747949600219727, 'learning_rate': 1.2440056257076374e-06, 'kl': 0.0437, 'entropy': -0.0479, 'ce_loss': 0.0106, 'epoch': 1.76}

 44%|████▍     | 148/336 [43:45<55:46, 17.80s/it]
 44%|████▍     | 149/336 [44:05<57:46, 18.54s/it]
                                                 
{'loss': 0.0199, 'grad_norm': 0.19330933690071106, 'learning_rate': 1.23462011246195e-06, 'kl': 0.0126, 'entropy': -0.0552, 'ce_loss': 0.0221, 'epoch': 1.77}

 44%|████▍     | 149/336 [44:05<57:46, 18.54s/it]
 45%|████▍     | 150/336 [44:21<54:45, 17.66s/it]
                                                 
{'loss': 0.0238, 'grad_norm': 0.1976664662361145, 'learning_rate': 1.2252126764738844e-06, 'kl': 0.0078, 'entropy': -0.0659, 'ce_loss': 0.021, 'epoch': 1.79}

 45%|████▍     | 150/336 [44:21<54:45, 17.66s/it]
 45%|████▍     | 151/336 [44:43<58:28, 18.97s/it]
                                                 
{'loss': 0.02, 'grad_norm': 0.17954564094543457, 'learning_rate': 1.2157841967678063e-06, 'kl': 0.0168, 'entropy': 0.0046, 'ce_loss': 0.0035, 'epoch': 1.8}

 45%|████▍     | 151/336 [44:43<58:28, 18.97s/it]
 45%|████▌     | 152/336 [44:59<55:42, 18.16s/it]
                                                 
{'loss': 0.0263, 'grad_norm': 0.20568442344665527, 'learning_rate': 1.2063355543343923e-06, 'kl': 0.0344, 'entropy': -0.0776, 'ce_loss': 0.018, 'epoch': 1.81}

 45%|████▌     | 152/336 [44:59<55:42, 18.16s/it]
 46%|████▌     | 153/336 [45:15<53:08, 17.43s/it]
                                                 
{'loss': 0.0297, 'grad_norm': 0.2846704125404358, 'learning_rate': 1.1968676320483101e-06, 'kl': 0.0135, 'entropy': -0.0031, 'ce_loss': 0.011, 'epoch': 1.82}

 46%|████▌     | 153/336 [45:15<53:08, 17.43s/it]
 46%|████▌     | 154/336 [45:34<54:21, 17.92s/it]
                                                 
{'loss': 0.0218, 'grad_norm': 0.19644086062908173, 'learning_rate': 1.1873813145857248e-06, 'kl': 0.0334, 'entropy': -0.0391, 'ce_loss': 0.0388, 'epoch': 1.83}

 46%|████▌     | 154/336 [45:34<54:21, 17.92s/it]
 46%|████▌     | 155/336 [45:53<55:02, 18.24s/it]
                                                 
{'loss': 0.0236, 'grad_norm': 0.20494718849658966, 'learning_rate': 1.1778774883416322e-06, 'kl': 0.0227, 'entropy': -0.0212, 'ce_loss': 0.0207, 'epoch': 1.85}

 46%|████▌     | 155/336 [45:53<55:02, 18.24s/it]
 46%|████▋     | 156/336 [46:12<55:20, 18.45s/it]
                                                 
{'loss': 0.0174, 'grad_norm': 0.16688618063926697, 'learning_rate': 1.1683570413470383e-06, 'kl': 0.0198, 'entropy': -0.0238, 'ce_loss': 0.0162, 'epoch': 1.86}

 46%|████▋     | 156/336 [46:12<55:20, 18.45s/it]
 47%|████▋     | 157/336 [46:32<56:35, 18.97s/it]
                                                 
{'loss': 0.0229, 'grad_norm': 0.1961638331413269, 'learning_rate': 1.1588208631859807e-06, 'kl': 0.0284, 'entropy': -0.0869, 'ce_loss': 0.0147, 'epoch': 1.87}

 47%|████▋     | 157/336 [46:32<56:35, 18.97s/it]
 47%|████▋     | 158/336 [46:51<56:11, 18.94s/it]
                                                 
{'loss': 0.0219, 'grad_norm': 0.2146499902009964, 'learning_rate': 1.149269844912404e-06, 'kl': 0.0243, 'entropy': -0.0603, 'ce_loss': 0.0058, 'epoch': 1.88}

 47%|████▋     | 158/336 [46:51<56:11, 18.94s/it]
 47%|████▋     | 159/336 [47:10<56:02, 18.99s/it]
                                                 
{'loss': 0.0245, 'grad_norm': 0.2261628657579422, 'learning_rate': 1.1397048789669059e-06, 'kl': 0.0242, 'entropy': -0.0796, 'ce_loss': 0.0189, 'epoch': 1.89}

 47%|████▋     | 159/336 [47:10<56:02, 18.99s/it]
 48%|████▊     | 160/336 [47:30<56:31, 19.27s/it]
                                                 
{'loss': 0.0222, 'grad_norm': 0.18682613968849182, 'learning_rate': 1.1301268590933434e-06, 'kl': 0.0203, 'entropy': -0.0576, 'ce_loss': 0.0128, 'epoch': 1.9}

 48%|████▊     | 160/336 [47:30<56:31, 19.27s/it]
 48%|████▊     | 161/336 [47:45<53:01, 18.18s/it]
                                                 
{'loss': 0.0258, 'grad_norm': 0.2125205546617508, 'learning_rate': 1.1205366802553228e-06, 'kl': 0.0183, 'entropy': -0.0405, 'ce_loss': 0.0125, 'epoch': 1.92}

 48%|████▊     | 161/336 [47:45<53:01, 18.18s/it]
 48%|████▊     | 162/336 [48:04<52:44, 18.19s/it]
                                                 
{'loss': 0.0246, 'grad_norm': 0.21307235956192017, 'learning_rate': 1.110935238552578e-06, 'kl': 0.0459, 'entropy': -0.1074, 'ce_loss': 0.0112, 'epoch': 1.93}

 48%|████▊     | 162/336 [48:04<52:44, 18.19s/it]
 49%|████▊     | 163/336 [48:19<50:13, 17.42s/it]
                                                 
{'loss': 0.0283, 'grad_norm': 0.22004197537899017, 'learning_rate': 1.1013234311372353e-06, 'kl': 0.0327, 'entropy': -0.0884, 'ce_loss': 0.0336, 'epoch': 1.94}

 49%|████▊     | 163/336 [48:19<50:13, 17.42s/it]
 49%|████▉     | 164/336 [48:39<51:39, 18.02s/it]
                                                 
{'loss': 0.021, 'grad_norm': 0.20040468871593475, 'learning_rate': 1.0917021561299862e-06, 'kl': 0.0292, 'entropy': -0.0161, 'ce_loss': 0.0119, 'epoch': 1.95}

 49%|████▉     | 164/336 [48:39<51:39, 18.02s/it]
 49%|████▉     | 165/336 [49:00<54:31, 19.13s/it]
                                                 
{'loss': 0.0211, 'grad_norm': 0.17962150275707245, 'learning_rate': 1.0820723125361684e-06, 'kl': 0.0253, 'entropy': -0.063, 'ce_loss': 0.0112, 'epoch': 1.96}

 49%|████▉     | 165/336 [49:00<54:31, 19.13s/it]
 49%|████▉     | 166/336 [49:19<53:47, 18.99s/it]
                                                 
{'loss': 0.0195, 'grad_norm': 0.17014382779598236, 'learning_rate': 1.0724348001617625e-06, 'kl': 0.0913, 'entropy': -0.0869, 'ce_loss': 0.0093, 'epoch': 1.98}

 49%|████▉     | 166/336 [49:19<53:47, 18.99s/it]
 50%|████▉     | 167/336 [49:35<50:39, 17.99s/it]
                                                 
{'loss': 0.0222, 'grad_norm': 0.18333743512630463, 'learning_rate': 1.0627905195293135e-06, 'kl': 0.0256, 'entropy': -0.0571, 'ce_loss': 0.0128, 'epoch': 1.99}

 50%|████▉     | 167/336 [49:35<50:39, 17.99s/it]
 50%|█████     | 168/336 [49:51<48:40, 17.39s/it]
                                                 
{'loss': 0.0272, 'grad_norm': 0.24237695336341858, 'learning_rate': 1.0531403717937886e-06, 'kl': 0.0027, 'entropy': -0.0026, 'ce_loss': 0.0052, 'epoch': 2.0}

 50%|█████     | 168/336 [49:51<48:40, 17.39s/it]
 50%|█████     | 169/336 [50:07<47:08, 16.94s/it]
                                                 
{'loss': 0.0162, 'grad_norm': 0.15395523607730865, 'learning_rate': 1.0434852586583737e-06, 'kl': 0.0449, 'entropy': -0.0021, 'ce_loss': 0.0147, 'epoch': 2.01}

 50%|█████     | 169/336 [50:07<47:08, 16.94s/it]
 51%|█████     | 170/336 [50:24<47:01, 17.00s/it]
                                                 
{'loss': 0.0146, 'grad_norm': 0.13404767215251923, 'learning_rate': 1.0338260822902165e-06, 'kl': 0.0332, 'entropy': -0.0286, 'ce_loss': 0.0201, 'epoch': 2.02}

 51%|█████     | 170/336 [50:24<47:01, 17.00s/it]
 51%|█████     | 171/336 [50:41<46:48, 17.02s/it]
                                                 
{'loss': 0.0194, 'grad_norm': 0.2107783555984497, 'learning_rate': 1.0241637452361322e-06, 'kl': 0.0859, 'entropy': -0.1113, 'ce_loss': 0.0114, 'epoch': 2.04}

 51%|█████     | 171/336 [50:41<46:48, 17.02s/it]
 51%|█████     | 172/336 [50:57<45:28, 16.64s/it]
                                                 
{'loss': 0.0213, 'grad_norm': 0.17959530651569366, 'learning_rate': 1.0144991503382673e-06, 'kl': 0.0581, 'entropy': -0.0378, 'ce_loss': 0.0105, 'epoch': 2.05}

 51%|█████     | 172/336 [50:57<45:28, 16.64s/it]
 51%|█████▏    | 173/336 [51:12<44:24, 16.34s/it]
                                                 
{'loss': 0.0161, 'grad_norm': 0.1580849587917328, 'learning_rate': 1.0048332006497404e-06, 'kl': 0.0408, 'entropy': -0.0569, 'ce_loss': 0.0214, 'epoch': 2.06}

 51%|█████▏    | 173/336 [51:12<44:24, 16.34s/it]
 52%|█████▏    | 174/336 [51:29<44:25, 16.45s/it]
                                                 
{'loss': 0.0162, 'grad_norm': 0.196575328707695, 'learning_rate': 9.951667993502597e-07, 'kl': 0.0728, 'entropy': -0.1484, 'ce_loss': 0.0088, 'epoch': 2.07}

 52%|█████▏    | 174/336 [51:29<44:25, 16.45s/it]
 52%|█████▏    | 175/336 [51:45<43:31, 16.22s/it]
                                                 
{'loss': 0.018, 'grad_norm': 0.17969413101673126, 'learning_rate': 9.855008496617326e-07, 'kl': 0.0391, 'entropy': -0.0245, 'ce_loss': 0.0064, 'epoch': 2.08}

 52%|█████▏    | 175/336 [51:45<43:31, 16.22s/it]
 52%|█████▏    | 176/336 [52:03<44:57, 16.86s/it]
                                                 
{'loss': 0.0143, 'grad_norm': 0.1602613776922226, 'learning_rate': 9.75836254763868e-07, 'kl': 0.03, 'entropy': -0.0674, 'ce_loss': 0.009, 'epoch': 2.1}

 52%|█████▏    | 176/336 [52:03<44:57, 16.86s/it]
 53%|█████▎    | 177/336 [52:22<46:28, 17.54s/it]
                                                 
{'loss': 0.0209, 'grad_norm': 0.19288259744644165, 'learning_rate': 9.661739177097834e-07, 'kl': 0.026, 'entropy': -0.0718, 'ce_loss': 0.0052, 'epoch': 2.11}

 53%|█████▎    | 177/336 [52:22<46:28, 17.54s/it]
 53%|█████▎    | 178/336 [52:38<44:40, 16.96s/it]
                                                 
{'loss': 0.0202, 'grad_norm': 0.21120965480804443, 'learning_rate': 9.565147413416265e-07, 'kl': 0.0542, 'entropy': -0.1177, 'ce_loss': 0.0089, 'epoch': 2.12}

 53%|█████▎    | 178/336 [52:38<44:40, 16.96s/it]
 53%|█████▎    | 179/336 [52:53<43:27, 16.61s/it]
                                                 
{'loss': 0.0118, 'grad_norm': 0.15320152044296265, 'learning_rate': 9.468596282062113e-07, 'kl': 0.0167, 'entropy': -0.0544, 'ce_loss': 0.0139, 'epoch': 2.13}

 53%|█████▎    | 179/336 [52:53<43:27, 16.61s/it]
 54%|█████▎    | 180/336 [53:12<44:59, 17.31s/it]
                                                 
{'loss': 0.0175, 'grad_norm': 0.1730872541666031, 'learning_rate': 9.372094804706866e-07, 'kl': 0.0532, 'entropy': 0.0073, 'ce_loss': 0.0102, 'epoch': 2.14}

 54%|█████▎    | 180/336 [53:12<44:59, 17.31s/it]
 54%|█████▍    | 181/336 [53:31<45:53, 17.76s/it]
                                                 
{'loss': 0.0142, 'grad_norm': 0.15351615846157074, 'learning_rate': 9.275651998382377e-07, 'kl': 0.0293, 'entropy': -0.0095, 'ce_loss': 0.0073, 'epoch': 2.15}

 54%|█████▍    | 181/336 [53:31<45:53, 17.76s/it]
 54%|█████▍    | 182/336 [53:50<46:30, 18.12s/it]
                                                 
{'loss': 0.0221, 'grad_norm': 0.2354268729686737, 'learning_rate': 9.179276874638314e-07, 'kl': 0.0596, 'entropy': -0.1426, 'ce_loss': 0.029, 'epoch': 2.17}

 54%|█████▍    | 182/336 [53:50<46:30, 18.12s/it]
 54%|█████▍    | 183/336 [54:06<44:30, 17.45s/it]
                                                 
{'loss': 0.0173, 'grad_norm': 0.1883721500635147, 'learning_rate': 9.082978438700138e-07, 'kl': 0.0532, 'entropy': -0.0776, 'ce_loss': 0.0161, 'epoch': 2.18}

 54%|█████▍    | 183/336 [54:06<44:30, 17.45s/it]
 55%|█████▍    | 184/336 [54:22<43:24, 17.13s/it]
                                                 
{'loss': 0.0168, 'grad_norm': 0.19475631415843964, 'learning_rate': 8.986765688627651e-07, 'kl': 0.0549, 'entropy': -0.0366, 'ce_loss': 0.01, 'epoch': 2.19}

 55%|█████▍    | 184/336 [54:22<43:24, 17.13s/it]
 55%|█████▌    | 185/336 [54:38<42:00, 16.69s/it]
                                                 
{'loss': 0.0155, 'grad_norm': 0.19626420736312866, 'learning_rate': 8.890647614474222e-07, 'kl': 0.0237, 'entropy': -0.0209, 'ce_loss': 0.0138, 'epoch': 2.2}

 55%|█████▌    | 185/336 [54:38<42:00, 16.69s/it]
 55%|█████▌    | 186/336 [55:00<45:51, 18.34s/it]
                                                 
{'loss': 0.0172, 'grad_norm': 0.21107307076454163, 'learning_rate': 8.79463319744677e-07, 'kl': 0.0613, 'entropy': 0.0027, 'ce_loss': 0.012, 'epoch': 2.21}

 55%|█████▌    | 186/336 [55:00<45:51, 18.34s/it]
 56%|█████▌    | 187/336 [55:18<45:11, 18.20s/it]
                                                 
{'loss': 0.0184, 'grad_norm': 0.20781706273555756, 'learning_rate': 8.698731409066568e-07, 'kl': 0.0221, 'entropy': -0.0269, 'ce_loss': 0.0102, 'epoch': 2.23}

 56%|█████▌    | 187/336 [55:18<45:11, 18.20s/it]
 56%|█████▌    | 188/336 [55:34<43:28, 17.62s/it]
                                                 
{'loss': 0.0213, 'grad_norm': 0.20844508707523346, 'learning_rate': 8.602951210330941e-07, 'kl': 0.0811, 'entropy': -0.0398, 'ce_loss': 0.0154, 'epoch': 2.24}

 56%|█████▌    | 188/336 [55:34<43:28, 17.62s/it]
 56%|█████▋    | 189/336 [55:53<43:32, 17.77s/it]
                                                 
{'loss': 0.0177, 'grad_norm': 0.21321342885494232, 'learning_rate': 8.507301550875959e-07, 'kl': 0.0032, 'entropy': -0.0056, 'ce_loss': 0.0025, 'epoch': 2.25}

 56%|█████▋    | 189/336 [55:53<43:32, 17.77s/it]
 57%|█████▋    | 190/336 [56:13<45:29, 18.69s/it]
                                                 
{'loss': 0.0166, 'grad_norm': 0.20195260643959045, 'learning_rate': 8.411791368140195e-07, 'kl': 0.0266, 'entropy': -0.0417, 'ce_loss': 0.0066, 'epoch': 2.26}

 57%|█████▋    | 190/336 [56:13<45:29, 18.69s/it]
 57%|█████▋    | 191/336 [56:29<43:00, 17.80s/it]
                                                 
{'loss': 0.0183, 'grad_norm': 0.20009654760360718, 'learning_rate': 8.316429586529614e-07, 'kl': 0.0574, 'entropy': -0.0474, 'ce_loss': 0.0076, 'epoch': 2.27}

 57%|█████▋    | 191/336 [56:29<43:00, 17.80s/it]
 57%|█████▋    | 192/336 [56:48<43:37, 18.18s/it]
                                                 
{'loss': 0.0218, 'grad_norm': 0.26318904757499695, 'learning_rate': 8.221225116583676e-07, 'kl': 0.0325, 'entropy': -0.0967, 'ce_loss': 0.0263, 'epoch': 2.29}

 57%|█████▋    | 192/336 [56:48<43:37, 18.18s/it]
 57%|█████▋    | 193/336 [57:05<42:17, 17.74s/it]
                                                 
{'loss': 0.0186, 'grad_norm': 0.19039058685302734, 'learning_rate': 8.126186854142751e-07, 'kl': 0.0264, 'entropy': -0.0669, 'ce_loss': 0.0072, 'epoch': 2.3}

 57%|█████▋    | 193/336 [57:05<42:17, 17.74s/it]
 58%|█████▊    | 194/336 [57:20<40:26, 17.09s/it]
                                                 
{'loss': 0.021, 'grad_norm': 0.23199713230133057, 'learning_rate': 8.031323679516899e-07, 'kl': 0.0188, 'entropy': -0.0713, 'ce_loss': 0.0116, 'epoch': 2.31}

 58%|█████▊    | 194/336 [57:20<40:26, 17.09s/it]
 58%|█████▊    | 195/336 [57:40<42:04, 17.90s/it]
                                                 
{'loss': 0.0137, 'grad_norm': 0.18526126444339752, 'learning_rate': 7.936644456656081e-07, 'kl': 0.0635, 'entropy': -0.0996, 'ce_loss': 0.0047, 'epoch': 2.32}

 58%|█████▊    | 195/336 [57:40<42:04, 17.90s/it]
 58%|█████▊    | 196/336 [57:56<40:07, 17.20s/it]
                                                 
{'loss': 0.0189, 'grad_norm': 0.22145313024520874, 'learning_rate': 7.84215803232194e-07, 'kl': 0.0291, 'entropy': -0.0659, 'ce_loss': 0.0072, 'epoch': 2.33}

 58%|█████▊    | 196/336 [57:56<40:07, 17.20s/it]
 59%|█████▊    | 197/336 [58:12<38:52, 16.78s/it]
                                                 
{'loss': 0.0189, 'grad_norm': 0.20441696047782898, 'learning_rate': 7.747873235261156e-07, 'kl': 0.0334, 'entropy': -0.0525, 'ce_loss': 0.022, 'epoch': 2.35}

 59%|█████▊    | 197/336 [58:12<38:52, 16.78s/it]
 59%|█████▉    | 198/336 [58:30<39:56, 17.37s/it]
                                                 
{'loss': 0.0165, 'grad_norm': 0.17816822230815887, 'learning_rate': 7.653798875380499e-07, 'kl': 0.0322, 'entropy': -0.0659, 'ce_loss': 0.0168, 'epoch': 2.36}

 59%|█████▉    | 198/336 [58:30<39:56, 17.37s/it]
 59%|█████▉    | 199/336 [58:52<42:45, 18.73s/it]
                                                 
{'loss': 0.0183, 'grad_norm': 0.18242992460727692, 'learning_rate': 7.559943742923625e-07, 'kl': 0.0056, 'entropy': -0.0041, 'ce_loss': 0.0044, 'epoch': 2.37}

 59%|█████▉    | 199/336 [58:52<42:45, 18.73s/it]
 60%|█████▉    | 200/336 [59:14<44:10, 19.49s/it]
                                                 
{'loss': 0.0144, 'grad_norm': 0.16184532642364502, 'learning_rate': 7.466316607649736e-07, 'kl': 0.0025, 'entropy': -0.0045, 'ce_loss': 0.0053, 'epoch': 2.38}

 60%|█████▉    | 200/336 [59:14<44:10, 19.49s/it]
 60%|█████▉    | 201/336 [59:32<43:16, 19.24s/it]
                                                 
{'loss': 0.0128, 'grad_norm': 0.1520135998725891, 'learning_rate': 7.372926218014131e-07, 'kl': 0.0591, 'entropy': -0.0311, 'ce_loss': 0.0276, 'epoch': 2.39}

 60%|█████▉    | 201/336 [59:32<43:16, 19.24s/it]
 60%|██████    | 202/336 [59:48<40:33, 18.16s/it]
                                                 
{'loss': 0.0224, 'grad_norm': 0.2052118480205536, 'learning_rate': 7.279781300350757e-07, 'kl': 0.1836, 'entropy': -0.1348, 'ce_loss': 0.0101, 'epoch': 2.4}

 60%|██████    | 202/336 [59:48<40:33, 18.16s/it]
 60%|██████    | 203/336 [1:00:04<38:35, 17.41s/it]
                                                   
{'loss': 0.0187, 'grad_norm': 0.19855929911136627, 'learning_rate': 7.186890558056836e-07, 'kl': 0.0145, 'entropy': -0.0366, 'ce_loss': 0.0135, 'epoch': 2.42}

 60%|██████    | 203/336 [1:00:04<38:35, 17.41s/it]
 61%|██████    | 204/336 [1:00:19<37:02, 16.83s/it]
                                                   
{'loss': 0.0177, 'grad_norm': 0.1831832379102707, 'learning_rate': 7.09426267077961e-07, 'kl': 0.0391, 'entropy': -0.0027, 'ce_loss': 0.018, 'epoch': 2.43}

 61%|██████    | 204/336 [1:00:19<37:02, 16.83s/it]
 61%|██████    | 205/336 [1:00:35<36:01, 16.50s/it]
                                                   
{'loss': 0.0163, 'grad_norm': 0.16797998547554016, 'learning_rate': 7.001906293605329e-07, 'kl': 0.0679, 'entropy': -0.0106, 'ce_loss': 0.0109, 'epoch': 2.44}

 61%|██████    | 205/336 [1:00:35<36:01, 16.50s/it]
 61%|██████▏   | 206/336 [1:00:55<38:06, 17.59s/it]
                                                   
{'loss': 0.0153, 'grad_norm': 0.1700359433889389, 'learning_rate': 6.909830056250526e-07, 'kl': 0.0262, 'entropy': -0.0146, 'ce_loss': 0.004, 'epoch': 2.45}

 61%|██████▏   | 206/336 [1:00:55<38:06, 17.59s/it]
 62%|██████▏   | 207/336 [1:01:11<37:00, 17.21s/it]
                                                   
{'loss': 0.02, 'grad_norm': 0.18519534170627594, 'learning_rate': 6.81804256225567e-07, 'kl': 0.0576, 'entropy': -0.0767, 'ce_loss': 0.0062, 'epoch': 2.46}

 62%|██████▏   | 207/336 [1:01:11<37:00, 17.21s/it]
 62%|██████▏   | 208/336 [1:01:37<42:05, 19.73s/it]
                                                   
{'loss': 0.0195, 'grad_norm': 0.1842510849237442, 'learning_rate': 6.726552388181233e-07, 'kl': 0.022, 'entropy': -0.0552, 'ce_loss': 0.0121, 'epoch': 2.48}

 62%|██████▏   | 208/336 [1:01:37<42:05, 19.73s/it]
 62%|██████▏   | 209/336 [1:01:56<41:25, 19.57s/it]
                                                   
{'loss': 0.0168, 'grad_norm': 0.18451088666915894, 'learning_rate': 6.63536808280633e-07, 'kl': 0.0503, 'entropy': -0.0537, 'ce_loss': 0.012, 'epoch': 2.49}

 62%|██████▏   | 209/336 [1:01:56<41:25, 19.57s/it]
 62%|██████▎   | 210/336 [1:02:12<38:44, 18.45s/it]
                                                   
{'loss': 0.0178, 'grad_norm': 0.18919403851032257, 'learning_rate': 6.544498166329912e-07, 'kl': 0.0049, 'entropy': -0.0171, 'ce_loss': 0.0278, 'epoch': 2.5}

 62%|██████▎   | 210/336 [1:02:12<38:44, 18.45s/it]
 63%|██████▎   | 211/336 [1:02:27<36:38, 17.59s/it]
                                                   
{'loss': 0.019, 'grad_norm': 0.19489766657352448, 'learning_rate': 6.453951129574643e-07, 'kl': 0.0859, 'entropy': -0.1299, 'ce_loss': 0.0093, 'epoch': 2.51}

 63%|██████▎   | 211/336 [1:02:27<36:38, 17.59s/it]
 63%|██████▎   | 212/336 [1:02:45<36:06, 17.47s/it]
                                                   
{'loss': 0.0247, 'grad_norm': 0.22543658316135406, 'learning_rate': 6.363735433193529e-07, 'kl': 0.0447, 'entropy': -0.0586, 'ce_loss': 0.0098, 'epoch': 2.52}

 63%|██████▎   | 212/336 [1:02:45<36:06, 17.47s/it]
 63%|██████▎   | 213/336 [1:03:02<35:28, 17.31s/it]
                                                   
{'loss': 0.0191, 'grad_norm': 0.2101595252752304, 'learning_rate': 6.273859506879364e-07, 'kl': 0.0194, 'entropy': -0.0265, 'ce_loss': 0.0077, 'epoch': 2.54}

 63%|██████▎   | 213/336 [1:03:02<35:28, 17.31s/it]
 64%|██████▎   | 214/336 [1:03:21<36:20, 17.87s/it]
                                                   
{'loss': 0.0197, 'grad_norm': 0.22376661002635956, 'learning_rate': 6.18433174857705e-07, 'kl': 0.063, 'entropy': -0.0239, 'ce_loss': 0.0106, 'epoch': 2.55}

 64%|██████▎   | 214/336 [1:03:21<36:20, 17.87s/it]
 64%|██████▍   | 215/336 [1:03:40<36:42, 18.20s/it]
                                                   
{'loss': 0.0172, 'grad_norm': 0.20559558272361755, 'learning_rate': 6.095160523698912e-07, 'kl': 0.0126, 'entropy': -0.063, 'ce_loss': 0.0269, 'epoch': 2.56}

 64%|██████▍   | 215/336 [1:03:40<36:42, 18.20s/it]
 64%|██████▍   | 216/336 [1:04:02<38:43, 19.36s/it]
                                                   
{'loss': 0.0151, 'grad_norm': 0.17532405257225037, 'learning_rate': 6.006354164343046e-07, 'kl': 0.032, 'entropy': -0.0762, 'ce_loss': 0.0291, 'epoch': 2.57}

 64%|██████▍   | 216/336 [1:04:02<38:43, 19.36s/it]
 65%|██████▍   | 217/336 [1:04:21<38:07, 19.22s/it]
                                                   
{'loss': 0.0198, 'grad_norm': 0.2614968419075012, 'learning_rate': 5.917920968514751e-07, 'kl': 0.1108, 'entropy': -0.0991, 'ce_loss': 0.007, 'epoch': 2.58}

 65%|██████▍   | 217/336 [1:04:21<38:07, 19.22s/it]
 65%|██████▍   | 218/336 [1:04:40<37:41, 19.16s/it]
                                                   
{'loss': 0.0167, 'grad_norm': 0.17147068679332733, 'learning_rate': 5.829869199351187e-07, 'kl': 0.0383, 'entropy': -0.0864, 'ce_loss': 0.0241, 'epoch': 2.6}

 65%|██████▍   | 218/336 [1:04:40<37:41, 19.16s/it]
 65%|██████▌   | 219/336 [1:05:01<38:45, 19.87s/it]
                                                   
{'loss': 0.0145, 'grad_norm': 0.15350160002708435, 'learning_rate': 5.742207084349273e-07, 'kl': 0.0457, 'entropy': -0.0289, 'ce_loss': 0.013, 'epoch': 2.61}

 65%|██████▌   | 219/336 [1:05:01<38:45, 19.87s/it]
 65%|██████▌   | 220/336 [1:05:17<36:06, 18.67s/it]
                                                   
{'loss': 0.0167, 'grad_norm': 0.18451078236103058, 'learning_rate': 5.654942814596901e-07, 'kl': 0.0703, 'entropy': -0.052, 'ce_loss': 0.0216, 'epoch': 2.62}

 65%|██████▌   | 220/336 [1:05:17<36:06, 18.67s/it]
 66%|██████▌   | 221/336 [1:05:35<35:36, 18.58s/it]
                                                   
{'loss': 0.0171, 'grad_norm': 0.18586644530296326, 'learning_rate': 5.568084544007588e-07, 'kl': 0.0544, 'entropy': -0.0757, 'ce_loss': 0.0097, 'epoch': 2.63}

 66%|██████▌   | 221/336 [1:05:35<35:36, 18.58s/it]
 66%|██████▌   | 222/336 [1:05:52<34:10, 17.99s/it]
                                                   
{'loss': 0.0156, 'grad_norm': 0.17732904851436615, 'learning_rate': 5.48164038855855e-07, 'kl': 0.0104, 'entropy': 0.027, 'ce_loss': 0.009, 'epoch': 2.64}

 66%|██████▌   | 222/336 [1:05:52<34:10, 17.99s/it]
 66%|██████▋   | 223/336 [1:06:09<33:22, 17.72s/it]
                                                   
{'loss': 0.0151, 'grad_norm': 0.1597178876399994, 'learning_rate': 5.395618425532389e-07, 'kl': 0.0007, 'entropy': -0.0108, 'ce_loss': 0.0102, 'epoch': 2.65}

 66%|██████▋   | 223/336 [1:06:09<33:22, 17.72s/it]
 67%|██████▋   | 224/336 [1:06:28<33:42, 18.05s/it]
                                                   
{'loss': 0.0139, 'grad_norm': 0.16127213835716248, 'learning_rate': 5.310026692762314e-07, 'kl': 0.063, 'entropy': -0.0089, 'ce_loss': 0.0045, 'epoch': 2.67}

 67%|██████▋   | 224/336 [1:06:28<33:42, 18.05s/it]
 67%|██████▋   | 225/336 [1:06:44<32:07, 17.36s/it]
                                                   
{'loss': 0.0205, 'grad_norm': 0.20752082765102386, 'learning_rate': 5.224873187881136e-07, 'kl': 0.0508, 'entropy': -0.0178, 'ce_loss': 0.0216, 'epoch': 2.68}

 67%|██████▋   | 225/336 [1:06:44<32:07, 17.36s/it]
 67%|██████▋   | 226/336 [1:07:00<31:15, 17.05s/it]
                                                   
{'loss': 0.016, 'grad_norm': 0.17134831845760345, 'learning_rate': 5.140165867573939e-07, 'kl': 0.0503, 'entropy': -0.0815, 'ce_loss': 0.0059, 'epoch': 2.69}

 67%|██████▋   | 226/336 [1:07:00<31:15, 17.05s/it]
 68%|██████▊   | 227/336 [1:07:15<30:06, 16.57s/it]
                                                   
{'loss': 0.0171, 'grad_norm': 0.1890375167131424, 'learning_rate': 5.055912646834635e-07, 'kl': 0.0214, 'entropy': -0.0396, 'ce_loss': 0.0122, 'epoch': 2.7}

 68%|██████▊   | 227/336 [1:07:15<30:06, 16.57s/it]
 68%|██████▊   | 228/336 [1:07:37<32:35, 18.11s/it]
                                                   
{'loss': 0.0138, 'grad_norm': 0.15590184926986694, 'learning_rate': 4.972121398226371e-07, 'kl': 0.0334, 'entropy': 0.0209, 'ce_loss': 0.0108, 'epoch': 2.71}

 68%|██████▊   | 228/336 [1:07:37<32:35, 18.11s/it]
 68%|██████▊   | 229/336 [1:07:53<31:03, 17.41s/it]
                                                   
{'loss': 0.0198, 'grad_norm': 0.1960582584142685, 'learning_rate': 4.888799951145947e-07, 'kl': 0.0315, 'entropy': -0.0879, 'ce_loss': 0.0134, 'epoch': 2.73}

 68%|██████▊   | 229/336 [1:07:53<31:03, 17.41s/it]
 68%|██████▊   | 230/336 [1:08:08<29:42, 16.82s/it]
                                                   
{'loss': 0.0173, 'grad_norm': 0.19161897897720337, 'learning_rate': 4.805956091092227e-07, 'kl': 0.0208, 'entropy': -0.0532, 'ce_loss': 0.0134, 'epoch': 2.74}

 68%|██████▊   | 230/336 [1:08:08<29:42, 16.82s/it]
 69%|██████▉   | 231/336 [1:08:24<28:54, 16.52s/it]
                                                   
{'loss': 0.023, 'grad_norm': 0.2421635091304779, 'learning_rate': 4.7235975589386713e-07, 'kl': 0.0532, 'entropy': -0.0908, 'ce_loss': 0.0157, 'epoch': 2.75}

 69%|██████▉   | 231/336 [1:08:24<28:54, 16.52s/it]
 69%|██████▉   | 232/336 [1:08:40<28:11, 16.26s/it]
                                                   
{'loss': 0.0175, 'grad_norm': 0.2146768867969513, 'learning_rate': 4.641732050210031e-07, 'kl': 0.0703, 'entropy': 0.0269, 'ce_loss': 0.0164, 'epoch': 2.76}

 69%|██████▉   | 232/336 [1:08:40<28:11, 16.26s/it]
 69%|██████▉   | 233/336 [1:08:56<27:43, 16.15s/it]
                                                   
{'loss': 0.0159, 'grad_norm': 0.19451335072517395, 'learning_rate': 4.5603672143632945e-07, 'kl': 0.0148, 'entropy': -0.0435, 'ce_loss': 0.019, 'epoch': 2.77}

 69%|██████▉   | 233/336 [1:08:56<27:43, 16.15s/it]
 70%|██████▉   | 234/336 [1:09:17<29:52, 17.58s/it]
                                                   
{'loss': 0.0164, 'grad_norm': 0.18908201158046722, 'learning_rate': 4.479510654072909e-07, 'kl': 0.0459, 'entropy': -0.0204, 'ce_loss': 0.0044, 'epoch': 2.79}

 70%|██████▉   | 234/336 [1:09:17<29:52, 17.58s/it]
 70%|██████▉   | 235/336 [1:09:36<30:19, 18.01s/it]
                                                   
{'loss': 0.0178, 'grad_norm': 0.18985359370708466, 'learning_rate': 4.399169924520403e-07, 'kl': 0.0356, 'entropy': -0.0447, 'ce_loss': 0.0174, 'epoch': 2.8}

 70%|██████▉   | 235/336 [1:09:36<30:19, 18.01s/it]
 70%|███████   | 236/336 [1:09:51<28:50, 17.30s/it]
                                                   
{'loss': 0.0162, 'grad_norm': 0.1735706329345703, 'learning_rate': 4.3193525326884426e-07, 'kl': 0.0131, 'entropy': -0.0542, 'ce_loss': 0.0136, 'epoch': 2.81}

 70%|███████   | 236/336 [1:09:51<28:50, 17.30s/it]
 71%|███████   | 237/336 [1:10:07<27:43, 16.80s/it]
                                                   
{'loss': 0.0227, 'grad_norm': 0.23269647359848022, 'learning_rate': 4.240065936659374e-07, 'kl': 0.0447, 'entropy': -0.0664, 'ce_loss': 0.0172, 'epoch': 2.82}

 71%|███████   | 237/336 [1:10:07<27:43, 16.80s/it]
 71%|███████   | 238/336 [1:10:23<27:02, 16.56s/it]
                                                   
{'loss': 0.0174, 'grad_norm': 0.1907908022403717, 'learning_rate': 4.1613175449183446e-07, 'kl': 0.0522, 'entropy': -0.027, 'ce_loss': 0.0101, 'epoch': 2.83}

 71%|███████   | 238/336 [1:10:23<27:02, 16.56s/it]
 71%|███████   | 239/336 [1:10:46<29:39, 18.35s/it]
                                                   
{'loss': 0.0164, 'grad_norm': 0.18256975710391998, 'learning_rate': 4.0831147156610676e-07, 'kl': 0.0352, 'entropy': -0.0698, 'ce_loss': 0.0075, 'epoch': 2.85}

 71%|███████   | 239/336 [1:10:46<29:39, 18.35s/it]
 71%|███████▏  | 240/336 [1:11:07<30:48, 19.25s/it]
                                                   
{'loss': 0.0142, 'grad_norm': 0.17074568569660187, 'learning_rate': 4.0054647561062615e-07, 'kl': 0.0688, 'entropy': 0.0356, 'ce_loss': 0.0199, 'epoch': 2.86}

 71%|███████▏  | 240/336 [1:11:07<30:48, 19.25s/it]
 72%|███████▏  | 241/336 [1:11:23<28:52, 18.23s/it]
                                                   
{'loss': 0.0194, 'grad_norm': 0.19037066400051117, 'learning_rate': 3.928374921812888e-07, 'kl': 0.0447, 'entropy': -0.0684, 'ce_loss': 0.0054, 'epoch': 2.87}

 72%|███████▏  | 241/336 [1:11:23<28:52, 18.23s/it]
 72%|███████▏  | 242/336 [1:11:42<28:54, 18.46s/it]
                                                   
{'loss': 0.0185, 'grad_norm': 0.1752237230539322, 'learning_rate': 3.851852416002187e-07, 'kl': 0.0229, 'entropy': -0.0327, 'ce_loss': 0.0198, 'epoch': 2.88}

 72%|███████▏  | 242/336 [1:11:42<28:54, 18.46s/it]
 72%|███████▏  | 243/336 [1:12:00<28:25, 18.34s/it]
                                                   
{'loss': 0.0162, 'grad_norm': 0.20181013643741608, 'learning_rate': 3.7759043888846173e-07, 'kl': 0.0325, 'entropy': -0.0439, 'ce_loss': 0.0097, 'epoch': 2.89}

 72%|███████▏  | 243/336 [1:12:00<28:25, 18.34s/it]
 73%|███████▎  | 244/336 [1:12:19<28:35, 18.64s/it]
                                                   
{'loss': 0.0164, 'grad_norm': 0.20712324976921082, 'learning_rate': 3.7005379369917324e-07, 'kl': 0.0243, 'entropy': -0.0796, 'ce_loss': 0.012, 'epoch': 2.9}

 73%|███████▎  | 244/336 [1:12:19<28:35, 18.64s/it]
 73%|███████▎  | 245/336 [1:12:35<27:14, 17.96s/it]
                                                   
{'loss': 0.0186, 'grad_norm': 0.19520629942417145, 'learning_rate': 3.625760102513102e-07, 'kl': 0.0562, 'entropy': -0.0293, 'ce_loss': 0.0083, 'epoch': 2.92}

 73%|███████▎  | 245/336 [1:12:35<27:14, 17.96s/it]
 73%|███████▎  | 246/336 [1:12:54<27:05, 18.06s/it]
                                                   
{'loss': 0.0166, 'grad_norm': 0.19249661266803741, 'learning_rate': 3.551577872638296e-07, 'kl': 0.0854, 'entropy': -0.1016, 'ce_loss': 0.0133, 'epoch': 2.93}

 73%|███████▎  | 246/336 [1:12:54<27:05, 18.06s/it]
 74%|███████▎  | 247/336 [1:13:10<25:47, 17.38s/it]
                                                   
{'loss': 0.0222, 'grad_norm': 0.2110520452260971, 'learning_rate': 3.477998178903981e-07, 'kl': 0.0254, 'entropy': -0.0016, 'ce_loss': 0.0083, 'epoch': 2.94}

 74%|███████▎  | 247/336 [1:13:10<25:47, 17.38s/it]
 74%|███████▍  | 248/336 [1:13:25<24:47, 16.90s/it]
                                                   
{'loss': 0.02, 'grad_norm': 0.21636487543582916, 'learning_rate': 3.4050278965462763e-07, 'kl': 0.0188, 'entropy': -0.0415, 'ce_loss': 0.0099, 'epoch': 2.95}

 74%|███████▍  | 248/336 [1:13:25<24:47, 16.90s/it]
 74%|███████▍  | 249/336 [1:13:47<26:31, 18.29s/it]
                                                   
{'loss': 0.0149, 'grad_norm': 0.1615242063999176, 'learning_rate': 3.3326738438583114e-07, 'kl': 0.03, 'entropy': -0.0845, 'ce_loss': 0.0152, 'epoch': 2.96}

 74%|███████▍  | 249/336 [1:13:47<26:31, 18.29s/it]
 74%|███████▍  | 250/336 [1:14:03<25:11, 17.58s/it]
                                                   
{'loss': 0.0184, 'grad_norm': 0.2158062607049942, 'learning_rate': 3.260942781553142e-07, 'kl': 0.04, 'entropy': -0.0505, 'ce_loss': 0.0089, 'epoch': 2.98}

 74%|███████▍  | 250/336 [1:14:03<25:11, 17.58s/it]
 75%|███████▍  | 251/336 [1:14:19<24:15, 17.13s/it]
                                                   
{'loss': 0.0187, 'grad_norm': 0.19854053854942322, 'learning_rate': 3.189841412132027e-07, 'kl': 0.0152, 'entropy': -0.0767, 'ce_loss': 0.0101, 'epoch': 2.99}

 75%|███████▍  | 251/336 [1:14:19<24:15, 17.13s/it]
 75%|███████▌  | 252/336 [1:14:38<24:43, 17.66s/it]
                                                   
{'loss': 0.0174, 'grad_norm': 0.211074098944664, 'learning_rate': 3.1193763792581594e-07, 'kl': 0.0262, 'entropy': -0.0339, 'ce_loss': 0.0075, 'epoch': 3.0}

 75%|███████▌  | 252/336 [1:14:38<24:43, 17.66s/it]
 75%|███████▌  | 253/336 [1:14:56<24:50, 17.96s/it]
                                                   
{'loss': 0.0118, 'grad_norm': 0.14704979956150055, 'learning_rate': 3.0495542671358744e-07, 'kl': 0.0293, 'entropy': -0.0493, 'ce_loss': 0.0126, 'epoch': 3.01}

 75%|███████▌  | 253/336 [1:14:56<24:50, 17.96s/it]
 76%|███████▌  | 254/336 [1:15:12<23:37, 17.29s/it]
                                                   
{'loss': 0.0189, 'grad_norm': 0.17956186830997467, 'learning_rate': 2.980381599895433e-07, 'kl': 0.0413, 'entropy': -0.008, 'ce_loss': 0.006, 'epoch': 3.02}

 76%|███████▌  | 254/336 [1:15:12<23:37, 17.29s/it]
 76%|███████▌  | 255/336 [1:15:29<23:01, 17.05s/it]
                                                   
{'loss': 0.0158, 'grad_norm': 0.1652800440788269, 'learning_rate': 2.91186484098342e-07, 'kl': 0.0447, 'entropy': -0.084, 'ce_loss': 0.0198, 'epoch': 3.04}

 76%|███████▌  | 255/336 [1:15:29<23:01, 17.05s/it]
 76%|███████▌  | 256/336 [1:15:45<22:15, 16.70s/it]
                                                   
{'loss': 0.0172, 'grad_norm': 0.1860676407814026, 'learning_rate': 2.84401039255879e-07, 'kl': 0.1084, 'entropy': -0.1768, 'ce_loss': 0.0095, 'epoch': 3.05}

 76%|███████▌  | 256/336 [1:15:45<22:15, 16.70s/it]
 76%|███████▋  | 257/336 [1:16:04<23:02, 17.50s/it]
                                                   
{'loss': 0.0137, 'grad_norm': 0.17659200727939606, 'learning_rate': 2.776824594894661e-07, 'kl': 0.0298, 'entropy': -0.0393, 'ce_loss': 0.0067, 'epoch': 3.06}

 76%|███████▋  | 257/336 [1:16:04<23:02, 17.50s/it]
 77%|███████▋  | 258/336 [1:16:22<23:04, 17.75s/it]
                                                   
{'loss': 0.0139, 'grad_norm': 0.1678144633769989, 'learning_rate': 2.7103137257858863e-07, 'kl': 0.0044, 'entropy': -0.003, 'ce_loss': 0.0071, 'epoch': 3.07}

 77%|███████▋  | 258/336 [1:16:22<23:04, 17.75s/it]
 77%|███████▋  | 259/336 [1:16:41<23:17, 18.14s/it]
                                                   
{'loss': 0.0137, 'grad_norm': 0.13771401345729828, 'learning_rate': 2.644483999962449e-07, 'kl': -0.0025, 'entropy': -0.0101, 'ce_loss': 0.0261, 'epoch': 3.08}

 77%|███████▋  | 259/336 [1:16:41<23:17, 18.14s/it]
 77%|███████▋  | 260/336 [1:16:57<22:12, 17.53s/it]
                                                   
{'loss': 0.0142, 'grad_norm': 0.15306684374809265, 'learning_rate': 2.579341568508779e-07, 'kl': 0.0214, 'entropy': -0.0198, 'ce_loss': 0.0163, 'epoch': 3.1}

 77%|███████▋  | 260/336 [1:16:57<22:12, 17.53s/it]
 78%|███████▊  | 261/336 [1:17:13<21:16, 17.02s/it]
                                                   
{'loss': 0.0177, 'grad_norm': 0.19758057594299316, 'learning_rate': 2.514892518288988e-07, 'kl': 0.041, 'entropy': -0.0322, 'ce_loss': 0.0069, 'epoch': 3.11}

 78%|███████▊  | 261/336 [1:17:13<21:16, 17.02s/it]
 78%|███████▊  | 262/336 [1:17:29<20:28, 16.60s/it]
                                                   
{'loss': 0.0178, 'grad_norm': 0.17641344666481018, 'learning_rate': 2.4511428713781236e-07, 'kl': -0.0012, 'entropy': -0.0288, 'ce_loss': 0.0192, 'epoch': 3.12}

 78%|███████▊  | 262/336 [1:17:29<20:28, 16.60s/it]
 78%|███████▊  | 263/336 [1:17:49<21:18, 17.52s/it]
                                                   
{'loss': 0.0137, 'grad_norm': 0.16513672471046448, 'learning_rate': 2.3880985844994673e-07, 'kl': 0.0603, 'entropy': -0.0781, 'ce_loss': 0.0111, 'epoch': 3.13}

 78%|███████▊  | 263/336 [1:17:49<21:18, 17.52s/it]
 79%|███████▊  | 264/336 [1:18:07<21:32, 17.95s/it]
                                                   
{'loss': 0.013, 'grad_norm': 0.15065506100654602, 'learning_rate': 2.3257655484679372e-07, 'kl': 0.022, 'entropy': -0.04, 'ce_loss': 0.0199, 'epoch': 3.14}

 79%|███████▊  | 264/336 [1:18:07<21:32, 17.95s/it]
 79%|███████▉  | 265/336 [1:18:23<20:30, 17.33s/it]
                                                   
{'loss': 0.0147, 'grad_norm': 0.15785431861877441, 'learning_rate': 2.264149587639671e-07, 'kl': 0.0439, 'entropy': -0.0845, 'ce_loss': 0.0042, 'epoch': 3.15}

 79%|███████▉  | 265/336 [1:18:23<20:30, 17.33s/it]
 79%|███████▉  | 266/336 [1:18:39<19:47, 16.97s/it]
                                                   
{'loss': 0.0135, 'grad_norm': 0.15365606546401978, 'learning_rate': 2.2032564593677772e-07, 'kl': 0.0525, 'entropy': -0.043, 'ce_loss': 0.0035, 'epoch': 3.17}

 79%|███████▉  | 266/336 [1:18:39<19:47, 16.97s/it]
 79%|███████▉  | 267/336 [1:18:56<19:22, 16.85s/it]
                                                   
{'loss': 0.0169, 'grad_norm': 0.16480816900730133, 'learning_rate': 2.1430918534643994e-07, 'kl': 0.0474, 'entropy': -0.0679, 'ce_loss': 0.0065, 'epoch': 3.18}

 79%|███████▉  | 267/336 [1:18:56<19:22, 16.85s/it]
 80%|███████▉  | 268/336 [1:19:15<19:46, 17.45s/it]
                                                   
{'loss': 0.0129, 'grad_norm': 0.15277600288391113, 'learning_rate': 2.0836613916690427e-07, 'kl': 0.0884, 'entropy': -0.1602, 'ce_loss': 0.006, 'epoch': 3.19}

 80%|███████▉  | 268/336 [1:19:15<19:46, 17.45s/it]
 80%|████████  | 269/336 [1:19:34<19:54, 17.83s/it]
                                                   
{'loss': 0.0149, 'grad_norm': 0.19228748977184296, 'learning_rate': 2.0249706271232946e-07, 'kl': 0.0231, 'entropy': 0.0119, 'ce_loss': 0.0048, 'epoch': 3.2}

 80%|████████  | 269/336 [1:19:34<19:54, 17.83s/it]
 80%|████████  | 270/336 [1:19:52<19:49, 18.02s/it]
                                                   
{'loss': 0.0138, 'grad_norm': 0.1653250902891159, 'learning_rate': 1.9670250438519386e-07, 'kl': 0.0461, 'entropy': -0.0708, 'ce_loss': 0.023, 'epoch': 3.21}

 80%|████████  | 270/336 [1:19:52<19:49, 18.02s/it]
 81%|████████  | 271/336 [1:20:11<19:50, 18.32s/it]
                                                   
{'loss': 0.0155, 'grad_norm': 0.1669386476278305, 'learning_rate': 1.9098300562505264e-07, 'kl': 0.0688, 'entropy': -0.0654, 'ce_loss': 0.0135, 'epoch': 3.23}

 81%|████████  | 271/336 [1:20:11<19:50, 18.32s/it]
 81%|████████  | 272/336 [1:20:31<20:02, 18.79s/it]
                                                   
{'loss': 0.0136, 'grad_norm': 0.1864534616470337, 'learning_rate': 1.8533910085794713e-07, 'kl': 0.0576, 'entropy': -0.0679, 'ce_loss': 0.0091, 'epoch': 3.24}

 81%|████████  | 272/336 [1:20:31<20:02, 18.79s/it]
 81%|████████▏ | 273/336 [1:20:53<20:52, 19.89s/it]
                                                   
{'loss': 0.0127, 'grad_norm': 0.14990662038326263, 'learning_rate': 1.7977131744646724e-07, 'kl': 0.0674, 'entropy': -0.0835, 'ce_loss': 0.0045, 'epoch': 3.25}

 81%|████████▏ | 273/336 [1:20:53<20:52, 19.89s/it]
 82%|████████▏ | 274/336 [1:21:10<19:24, 18.79s/it]
                                                   
{'loss': 0.0164, 'grad_norm': 0.1857193112373352, 'learning_rate': 1.742801756404759e-07, 'kl': 0.0786, 'entropy': -0.0771, 'ce_loss': 0.0023, 'epoch': 3.26}

 82%|████████▏ | 274/336 [1:21:10<19:24, 18.79s/it]
 82%|████████▏ | 275/336 [1:21:25<18:08, 17.85s/it]
                                                   
{'loss': 0.0159, 'grad_norm': 0.17385807633399963, 'learning_rate': 1.688661885284972e-07, 'kl': 0.0488, 'entropy': -0.0184, 'ce_loss': 0.0044, 'epoch': 3.27}

 82%|████████▏ | 275/336 [1:21:25<18:08, 17.85s/it]
 82%|████████▏ | 276/336 [1:21:41<17:11, 17.19s/it]
                                                   
{'loss': 0.0152, 'grad_norm': 0.17598123848438263, 'learning_rate': 1.6352986198977325e-07, 'kl': 0.0383, 'entropy': -0.0757, 'ce_loss': 0.024, 'epoch': 3.29}

 82%|████████▏ | 276/336 [1:21:41<17:11, 17.19s/it]
 82%|████████▏ | 277/336 [1:21:57<16:29, 16.78s/it]
                                                   
{'loss': 0.0136, 'grad_norm': 0.1699410378932953, 'learning_rate': 1.5827169464699575e-07, 'kl': 0.0116, 'entropy': -0.0452, 'ce_loss': 0.0106, 'epoch': 3.3}

 82%|████████▏ | 277/336 [1:21:57<16:29, 16.78s/it]
 83%|████████▎ | 278/336 [1:22:16<17:01, 17.61s/it]
                                                   
{'loss': 0.0121, 'grad_norm': 0.15897028148174286, 'learning_rate': 1.5309217781971416e-07, 'kl': 0.0596, 'entropy': -0.0413, 'ce_loss': 0.0116, 'epoch': 3.31}

 83%|████████▎ | 278/336 [1:22:16<17:01, 17.61s/it]
 83%|████████▎ | 279/336 [1:22:41<18:48, 19.79s/it]
                                                   
{'loss': 0.0116, 'grad_norm': 0.13833433389663696, 'learning_rate': 1.479917954784282e-07, 'kl': 0.0522, 'entropy': -0.1055, 'ce_loss': 0.012, 'epoch': 3.32}

 83%|████████▎ | 279/336 [1:22:41<18:48, 19.79s/it]
 83%|████████▎ | 280/336 [1:22:57<17:19, 18.57s/it]
                                                   
{'loss': 0.0151, 'grad_norm': 0.19401074945926666, 'learning_rate': 1.429710241993656e-07, 'kl': 0.062, 'entropy': -0.0366, 'ce_loss': 0.0134, 'epoch': 3.33}

 83%|████████▎ | 280/336 [1:22:57<17:19, 18.57s/it]
 84%|████████▎ | 281/336 [1:23:15<16:49, 18.36s/it]
                                                   
{'loss': 0.0148, 'grad_norm': 0.1647351086139679, 'learning_rate': 1.380303331199507e-07, 'kl': 0.0432, 'entropy': -0.0352, 'ce_loss': 0.0142, 'epoch': 3.35}

 84%|████████▎ | 281/336 [1:23:15<16:49, 18.36s/it]
 84%|████████▍ | 282/336 [1:23:34<16:46, 18.64s/it]
                                                   
{'loss': 0.0123, 'grad_norm': 0.14924506843090057, 'learning_rate': 1.3317018389496926e-07, 'kl': 0.0398, 'entropy': -0.0535, 'ce_loss': 0.0106, 'epoch': 3.36}

 84%|████████▍ | 282/336 [1:23:34<16:46, 18.64s/it]
 84%|████████▍ | 283/336 [1:23:50<15:42, 17.78s/it]
                                                   
{'loss': 0.014, 'grad_norm': 0.17482659220695496, 'learning_rate': 1.283910306534308e-07, 'kl': 0.0026, 'entropy': -0.0491, 'ce_loss': 0.0243, 'epoch': 3.37}

 84%|████████▍ | 283/336 [1:23:50<15:42, 17.78s/it]
 85%|████████▍ | 284/336 [1:24:06<14:54, 17.21s/it]
                                                   
{'loss': 0.0147, 'grad_norm': 0.15375792980194092, 'learning_rate': 1.2369331995613663e-07, 'kl': 0.0017, 'entropy': 0.0039, 'ce_loss': 0.0138, 'epoch': 3.38}

 85%|████████▍ | 284/336 [1:24:06<14:54, 17.21s/it]
 85%|████████▍ | 285/336 [1:24:21<14:14, 16.76s/it]
                                                   
{'loss': 0.0187, 'grad_norm': 0.20818421244621277, 'learning_rate': 1.1907749075395146e-07, 'kl': 0.0664, 'entropy': -0.0408, 'ce_loss': 0.0133, 'epoch': 3.39}

 85%|████████▍ | 285/336 [1:24:21<14:14, 16.76s/it]
 85%|████████▌ | 286/336 [1:24:43<15:12, 18.25s/it]
                                                   
{'loss': 0.0126, 'grad_norm': 0.14181439578533173, 'learning_rate': 1.145439743467902e-07, 'kl': 0.0742, 'entropy': -0.0859, 'ce_loss': 0.0047, 'epoch': 3.4}

 85%|████████▌ | 286/336 [1:24:43<15:12, 18.25s/it]
 85%|████████▌ | 287/336 [1:25:02<15:09, 18.56s/it]
                                                   
{'loss': 0.0153, 'grad_norm': 0.18021345138549805, 'learning_rate': 1.1009319434331621e-07, 'kl': 0.0255, 'entropy': -0.0505, 'ce_loss': 0.0076, 'epoch': 3.42}

 85%|████████▌ | 287/336 [1:25:02<15:09, 18.56s/it]
 86%|████████▌ | 288/336 [1:25:21<14:56, 18.68s/it]
                                                   
{'loss': 0.0139, 'grad_norm': 0.17935115098953247, 'learning_rate': 1.0572556662136035e-07, 'kl': 0.0238, 'entropy': -0.1021, 'ce_loss': 0.0081, 'epoch': 3.43}

 86%|████████▌ | 288/336 [1:25:21<14:56, 18.68s/it]
 86%|████████▌ | 289/336 [1:25:37<13:53, 17.74s/it]
                                                   
{'loss': 0.0191, 'grad_norm': 0.20373126864433289, 'learning_rate': 1.014414992890611e-07, 'kl': 0.085, 'entropy': -0.084, 'ce_loss': 0.0068, 'epoch': 3.44}

 86%|████████▌ | 289/336 [1:25:37<13:53, 17.74s/it]
 86%|████████▋ | 290/336 [1:25:56<13:53, 18.12s/it]
                                                   
{'loss': 0.0157, 'grad_norm': 0.1873011291027069, 'learning_rate': 9.724139264673114e-08, 'kl': 0.033, 'entropy': -0.0723, 'ce_loss': 0.001, 'epoch': 3.45}

 86%|████████▋ | 290/336 [1:25:56<13:53, 18.12s/it]
 87%|████████▋ | 291/336 [1:26:16<14:03, 18.75s/it]
                                                   
{'loss': 0.0112, 'grad_norm': 0.1414089798927307, 'learning_rate': 9.312563914945459e-08, 'kl': 0.0728, 'entropy': -0.0486, 'ce_loss': 0.0065, 'epoch': 3.46}

 87%|████████▋ | 291/336 [1:26:16<14:03, 18.75s/it]
 87%|████████▋ | 292/336 [1:26:34<13:36, 18.56s/it]
                                                   
{'loss': 0.0178, 'grad_norm': 0.20066796243190765, 'learning_rate': 8.909462337041507e-08, 'kl': 0.0359, 'entropy': -0.015, 'ce_loss': 0.0083, 'epoch': 3.48}

 87%|████████▋ | 292/336 [1:26:34<13:36, 18.56s/it]
 87%|████████▋ | 293/336 [1:26:53<13:20, 18.62s/it]
                                                   
{'loss': 0.014, 'grad_norm': 0.15551452338695526, 'learning_rate': 8.514872196496181e-08, 'kl': 0.042, 'entropy': -0.0068, 'ce_loss': 0.0015, 'epoch': 3.49}

 87%|████████▋ | 293/336 [1:26:53<13:20, 18.62s/it]
 88%|████████▊ | 294/336 [1:27:09<12:26, 17.77s/it]
                                                   
{'loss': 0.0155, 'grad_norm': 0.18158553540706635, 'learning_rate': 8.128830363541572e-08, 'kl': 0.0598, 'entropy': -0.0439, 'ce_loss': 0.0056, 'epoch': 3.5}

 88%|████████▊ | 294/336 [1:27:09<12:26, 17.77s/it]
 88%|████████▊ | 295/336 [1:27:27<12:15, 17.94s/it]
                                                   
{'loss': 0.0161, 'grad_norm': 0.18261036276817322, 'learning_rate': 7.751372909661768e-08, 'kl': 0.0041, 'entropy': -0.0078, 'ce_loss': 0.0069, 'epoch': 3.51}

 88%|████████▊ | 295/336 [1:27:27<12:15, 17.94s/it]
 88%|████████▊ | 296/336 [1:27:44<11:40, 17.50s/it]
                                                   
{'loss': 0.0117, 'grad_norm': 0.31206467747688293, 'learning_rate': 7.382535104222364e-08, 'kl': -0.0017, 'entropy': -0.0193, 'ce_loss': 0.0172, 'epoch': 3.52}

 88%|████████▊ | 296/336 [1:27:44<11:40, 17.50s/it]
 88%|████████▊ | 297/336 [1:28:02<11:30, 17.70s/it]
                                                   
{'loss': 0.0139, 'grad_norm': 0.1553187072277069, 'learning_rate': 7.022351411174865e-08, 'kl': -0.006, 'entropy': -0.0374, 'ce_loss': 0.015, 'epoch': 3.54}

 88%|████████▊ | 297/336 [1:28:02<11:30, 17.70s/it]
 89%|████████▊ | 298/336 [1:28:22<11:36, 18.32s/it]
                                                   
{'loss': 0.0126, 'grad_norm': 0.15778586268424988, 'learning_rate': 6.670855485836524e-08, 'kl': 0.1206, 'entropy': -0.1113, 'ce_loss': 0.008, 'epoch': 3.55}

 89%|████████▊ | 298/336 [1:28:22<11:36, 18.32s/it]
 89%|████████▉ | 299/336 [1:28:42<11:35, 18.81s/it]
                                                   
{'loss': 0.0141, 'grad_norm': 0.19020713865756989, 'learning_rate': 6.328080171745509e-08, 'kl': 0.0229, 'entropy': -0.032, 'ce_loss': 0.0108, 'epoch': 3.56}

 89%|████████▉ | 299/336 [1:28:42<11:35, 18.81s/it]
 89%|████████▉ | 300/336 [1:28:58<10:50, 18.08s/it]
                                                   
{'loss': 0.0159, 'grad_norm': 0.1756381243467331, 'learning_rate': 5.994057497592031e-08, 'kl': 0.085, 'entropy': -0.0796, 'ce_loss': 0.0052, 'epoch': 3.57}

 89%|████████▉ | 300/336 [1:28:58<10:50, 18.08s/it]
 90%|████████▉ | 301/336 [1:29:14<10:07, 17.35s/it]
                                                   
{'loss': 0.0158, 'grad_norm': 0.18305473029613495, 'learning_rate': 5.6688186742256835e-08, 'kl': 0.0525, 'entropy': -0.054, 'ce_loss': 0.0216, 'epoch': 3.58}

 90%|████████▉ | 301/336 [1:29:14<10:07, 17.35s/it]
 90%|████████▉ | 302/336 [1:29:30<09:42, 17.14s/it]
                                                   
{'loss': 0.0158, 'grad_norm': 0.1584191471338272, 'learning_rate': 5.352394091739021e-08, 'kl': 0.0444, 'entropy': -0.0796, 'ce_loss': 0.0089, 'epoch': 3.6}

 90%|████████▉ | 302/336 [1:29:30<09:42, 17.14s/it]
 90%|█████████ | 303/336 [1:29:49<09:42, 17.66s/it]
                                                   
{'loss': 0.0154, 'grad_norm': 0.17178170382976532, 'learning_rate': 5.0448133166279935e-08, 'kl': 0.0337, 'entropy': -0.0942, 'ce_loss': 0.0075, 'epoch': 3.61}

 90%|█████████ | 303/336 [1:29:49<09:42, 17.66s/it]
 90%|█████████ | 304/336 [1:30:05<09:07, 17.10s/it]
                                                   
{'loss': 0.016, 'grad_norm': 0.18547679483890533, 'learning_rate': 4.746105089029229e-08, 'kl': 0.0601, 'entropy': -0.0037, 'ce_loss': 0.0058, 'epoch': 3.62}

 90%|█████████ | 304/336 [1:30:05<09:07, 17.10s/it]
 91%|█████████ | 305/336 [1:30:24<09:04, 17.58s/it]
                                                   
{'loss': 0.0139, 'grad_norm': 0.16790246963500977, 'learning_rate': 4.456297320034641e-08, 'kl': 0.0542, 'entropy': -0.0459, 'ce_loss': 0.0075, 'epoch': 3.63}

 91%|█████████ | 305/336 [1:30:24<09:04, 17.58s/it]
 91%|█████████ | 306/336 [1:30:39<08:29, 17.00s/it]
                                                   
{'loss': 0.0143, 'grad_norm': 0.29298320412635803, 'learning_rate': 4.1754170890833774e-08, 'kl': 0.0344, 'entropy': -0.0493, 'ce_loss': 0.0171, 'epoch': 3.64}

 91%|█████████ | 306/336 [1:30:39<08:29, 17.00s/it]
 91%|█████████▏| 307/336 [1:30:58<08:30, 17.61s/it]
                                                   
{'loss': 0.0104, 'grad_norm': 0.13091427087783813, 'learning_rate': 3.9034906414315725e-08, 'kl': 0.0166, 'entropy': -0.0068, 'ce_loss': 0.0246, 'epoch': 3.65}

 91%|█████████▏| 307/336 [1:30:58<08:30, 17.61s/it]
 92%|█████████▏| 308/336 [1:31:14<07:56, 17.01s/it]
                                                   
{'loss': 0.0151, 'grad_norm': 0.18825221061706543, 'learning_rate': 3.6405433856999676e-08, 'kl': 0.0635, 'entropy': -0.0425, 'ce_loss': 0.0116, 'epoch': 3.67}

 92%|█████████▏| 308/336 [1:31:14<07:56, 17.01s/it]
 92%|█████████▏| 309/336 [1:31:32<07:50, 17.43s/it]
                                                   
{'loss': 0.0167, 'grad_norm': 0.20501720905303955, 'learning_rate': 3.386599891499764e-08, 'kl': 0.062, 'entropy': -0.0437, 'ce_loss': 0.0109, 'epoch': 3.68}

 92%|█████████▏| 309/336 [1:31:32<07:50, 17.43s/it]
 92%|█████████▏| 310/336 [1:31:52<07:48, 18.01s/it]
                                                   
{'loss': 0.0119, 'grad_norm': 0.14792554080486298, 'learning_rate': 3.141683887136892e-08, 'kl': 0.0537, 'entropy': -0.0369, 'ce_loss': 0.0092, 'epoch': 3.69}

 92%|█████████▏| 310/336 [1:31:52<07:48, 18.01s/it]
 93%|█████████▎| 311/336 [1:32:10<07:36, 18.25s/it]
                                                   
{'loss': 0.015, 'grad_norm': 0.18962831795215607, 'learning_rate': 2.9058182573947986e-08, 'kl': 0.1035, 'entropy': -0.0718, 'ce_loss': 0.0211, 'epoch': 3.7}

 93%|█████████▎| 311/336 [1:32:10<07:36, 18.25s/it]
 93%|█████████▎| 312/336 [1:32:29<07:23, 18.46s/it]
                                                   
{'loss': 0.0149, 'grad_norm': 0.1783580332994461, 'learning_rate': 2.6790250413961546e-08, 'kl': 0.0337, 'entropy': -0.0486, 'ce_loss': 0.0158, 'epoch': 3.71}

 93%|█████████▎| 312/336 [1:32:29<07:23, 18.46s/it]
 93%|█████████▎| 313/336 [1:32:45<06:46, 17.69s/it]
                                                   
{'loss': 0.0193, 'grad_norm': 0.21417374908924103, 'learning_rate': 2.4613254305434815e-08, 'kl': 0.0208, 'entropy': -0.0405, 'ce_loss': 0.0144, 'epoch': 3.73}

 93%|█████████▎| 313/336 [1:32:45<06:46, 17.69s/it]
 93%|█████████▎| 314/336 [1:33:04<06:36, 18.04s/it]
                                                   
{'loss': 0.0133, 'grad_norm': 0.1617756336927414, 'learning_rate': 2.2527397665391024e-08, 'kl': 0.0669, 'entropy': -0.0703, 'ce_loss': 0.0188, 'epoch': 3.74}

 93%|█████████▎| 314/336 [1:33:04<06:36, 18.04s/it]
 94%|█████████▍| 315/336 [1:33:20<06:05, 17.39s/it]
                                                   
{'loss': 0.0142, 'grad_norm': 0.17499667406082153, 'learning_rate': 2.053287539484405e-08, 'kl': 0.0322, 'entropy': -0.0269, 'ce_loss': 0.0077, 'epoch': 3.75}

 94%|█████████▍| 315/336 [1:33:20<06:05, 17.39s/it]
 94%|█████████▍| 316/336 [1:33:36<05:37, 16.88s/it]
                                                   
{'loss': 0.0168, 'grad_norm': 0.19160637259483337, 'learning_rate': 1.8629873860586564e-08, 'kl': 0.0175, 'entropy': -0.0879, 'ce_loss': 0.0451, 'epoch': 3.76}

 94%|█████████▍| 316/336 [1:33:36<05:37, 16.88s/it]
 94%|█████████▍| 317/336 [1:33:55<05:33, 17.56s/it]
                                                   
{'loss': 0.0125, 'grad_norm': 0.16696055233478546, 'learning_rate': 1.6818570877776718e-08, 'kl': 0.03, 'entropy': -0.0098, 'ce_loss': 0.0167, 'epoch': 3.77}

 94%|█████████▍| 317/336 [1:33:55<05:33, 17.56s/it]
 95%|█████████▍| 318/336 [1:34:14<05:24, 18.05s/it]
                                                   
{'loss': 0.0158, 'grad_norm': 0.17223559319972992, 'learning_rate': 1.5099135693322773e-08, 'kl': 0.0771, 'entropy': -0.0261, 'ce_loss': 0.0142, 'epoch': 3.79}

 95%|█████████▍| 318/336 [1:34:14<05:24, 18.05s/it]
 95%|█████████▍| 319/336 [1:34:30<04:54, 17.35s/it]
                                                   
{'loss': 0.019, 'grad_norm': 0.21838437020778656, 'learning_rate': 1.3471728970068985e-08, 'kl': 0.0713, 'entropy': -0.085, 'ce_loss': 0.0093, 'epoch': 3.8}

 95%|█████████▍| 319/336 [1:34:30<04:54, 17.35s/it]
 95%|█████████▌| 320/336 [1:34:46<04:32, 17.02s/it]
                                                   
{'loss': 0.0146, 'grad_norm': 0.16921240091323853, 'learning_rate': 1.1936502771783486e-08, 'kl': 0.0645, 'entropy': -0.05, 'ce_loss': 0.0086, 'epoch': 3.81}

 95%|█████████▌| 320/336 [1:34:46<04:32, 17.02s/it]
 96%|█████████▌| 321/336 [1:35:02<04:08, 16.58s/it]
                                                   
{'loss': 0.0141, 'grad_norm': 0.1600177139043808, 'learning_rate': 1.0493600548948877e-08, 'kl': 0.0294, 'entropy': -0.0142, 'ce_loss': 0.006, 'epoch': 3.82}

 96%|█████████▌| 321/336 [1:35:02<04:08, 16.58s/it]
 96%|█████████▌| 322/336 [1:35:17<03:47, 16.24s/it]
                                                   
{'loss': 0.0183, 'grad_norm': 0.20258741080760956, 'learning_rate': 9.143157125359513e-09, 'kl': 0.1006, 'entropy': -0.0801, 'ce_loss': 0.0065, 'epoch': 3.83}

 96%|█████████▌| 322/336 [1:35:17<03:47, 16.24s/it]
 96%|█████████▌| 323/336 [1:35:35<03:38, 16.80s/it]
                                                   
{'loss': 0.0154, 'grad_norm': 0.18779589235782623, 'learning_rate': 7.885298685522235e-09, 'kl': 0.0339, 'entropy': -0.0284, 'ce_loss': 0.0106, 'epoch': 3.85}

 96%|█████████▌| 323/336 [1:35:35<03:38, 16.80s/it]
 96%|█████████▋| 324/336 [1:35:52<03:20, 16.72s/it]
                                                   
{'loss': 0.0146, 'grad_norm': 0.16684141755104065, 'learning_rate': 6.720142762867032e-09, 'kl': 0.165, 'entropy': -0.1553, 'ce_loss': 0.0074, 'epoch': 3.86}

 96%|█████████▋| 324/336 [1:35:52<03:20, 16.72s/it]
 97%|█████████▋| 325/336 [1:36:14<03:21, 18.30s/it]
                                                   
{'loss': 0.0105, 'grad_norm': 0.135105162858963, 'learning_rate': 5.647798228764156e-09, 'kl': 0.011, 'entropy': -0.1157, 'ce_loss': 0.0246, 'epoch': 3.87}

 97%|█████████▋| 325/336 [1:36:14<03:21, 18.30s/it]
 97%|█████████▋| 326/336 [1:36:30<02:55, 17.57s/it]
                                                   
{'loss': 0.0142, 'grad_norm': 0.16241519153118134, 'learning_rate': 4.668365282351372e-09, 'kl': 0.028, 'entropy': -0.0454, 'ce_loss': 0.015, 'epoch': 3.88}

 97%|█████████▋| 326/336 [1:36:30<02:55, 17.57s/it]
 97%|█████████▋| 327/336 [1:36:45<02:33, 17.04s/it]
                                                   
{'loss': 0.0152, 'grad_norm': 0.17639409005641937, 'learning_rate': 3.7819354411713355e-09, 'kl': 0.0859, 'entropy': -0.0771, 'ce_loss': 0.0077, 'epoch': 3.89}

 97%|█████████▋| 327/336 [1:36:45<02:33, 17.04s/it]
 98%|█████████▊| 328/336 [1:37:04<02:20, 17.61s/it]
                                                   
{'loss': 0.0143, 'grad_norm': 0.1601157784461975, 'learning_rate': 2.9885915326203216e-09, 'kl': -0.0061, 'entropy': -0.0366, 'ce_loss': 0.0077, 'epoch': 3.9}

 98%|█████████▊| 328/336 [1:37:04<02:20, 17.61s/it]
 98%|█████████▊| 329/336 [1:37:20<01:59, 17.07s/it]
                                                   
{'loss': 0.0152, 'grad_norm': 0.17293810844421387, 'learning_rate': 2.2884076862089707e-09, 'kl': 0.0422, 'entropy': -0.0679, 'ce_loss': 0.016, 'epoch': 3.92}

 98%|█████████▊| 329/336 [1:37:20<01:59, 17.07s/it]
 98%|█████████▊| 330/336 [1:37:40<01:46, 17.78s/it]
                                                   
{'loss': 0.0151, 'grad_norm': 0.17773842811584473, 'learning_rate': 1.6814493266357199e-09, 'kl': 0.02, 'entropy': -0.043, 'ce_loss': 0.022, 'epoch': 3.93}

 98%|█████████▊| 330/336 [1:37:40<01:46, 17.78s/it]
 99%|█████████▊| 331/336 [1:37:55<01:25, 17.17s/it]
                                                   
{'loss': 0.0152, 'grad_norm': 0.17476435005664825, 'learning_rate': 1.1677731676733581e-09, 'kl': 0.0361, 'entropy': 0.0036, 'ce_loss': 0.0109, 'epoch': 3.94}

 99%|█████████▊| 331/336 [1:37:55<01:25, 17.17s/it]
 99%|█████████▉| 332/336 [1:38:11<01:06, 16.74s/it]
                                                   
{'loss': 0.0142, 'grad_norm': 0.17168329656124115, 'learning_rate': 7.474272068698217e-10, 'kl': 0.0532, 'entropy': -0.0459, 'ce_loss': 0.002, 'epoch': 3.95}

 99%|█████████▉| 332/336 [1:38:11<01:06, 16.74s/it]
 99%|█████████▉| 333/336 [1:38:30<00:52, 17.55s/it]
                                                   
{'loss': 0.011, 'grad_norm': 0.13916146755218506, 'learning_rate': 4.204507210633368e-10, 'kl': 0.0703, 'entropy': -0.1523, 'ce_loss': 0.0008, 'epoch': 3.96}

 99%|█████████▉| 333/336 [1:38:30<00:52, 17.55s/it]
 99%|█████████▉| 334/336 [1:38:49<00:35, 17.93s/it]
                                                   
{'loss': 0.0139, 'grad_norm': 0.18326641619205475, 'learning_rate': 1.8687426271246642e-10, 'kl': 0.0703, 'entropy': -0.0417, 'ce_loss': 0.0108, 'epoch': 3.98}

 99%|█████████▉| 334/336 [1:38:49<00:35, 17.93s/it]
100%|█████████▉| 335/336 [1:39:08<00:18, 18.28s/it]
                                                   
{'loss': 0.0145, 'grad_norm': 0.170999675989151, 'learning_rate': 4.6719657041283115e-11, 'kl': 0.0378, 'entropy': -0.0938, 'ce_loss': 0.0105, 'epoch': 3.99}

100%|█████████▉| 335/336 [1:39:08<00:18, 18.28s/it]
100%|██████████| 336/336 [1:39:24<00:00, 17.54s/it]
                                                   
{'loss': 0.014, 'grad_norm': 0.1720617562532425, 'learning_rate': 0.0, 'kl': 0.0508, 'entropy': -0.0742, 'ce_loss': 0.0039, 'epoch': 4.0}

100%|██████████| 336/336 [1:39:24<00:00, 17.54s/it][INFO|trainer.py:2665] 2025-04-13 16:37:52,456 >> 

Training completed. Do not forget to share your model on huggingface.co/models =)



                                                   
{'train_runtime': 5964.6993, 'train_samples_per_second': 1.803, 'train_steps_per_second': 0.056, 'train_loss': 0.021451761316885018, 'epoch': 4.0}

100%|██████████| 336/336 [1:39:24<00:00, 17.54s/it]
100%|██████████| 336/336 [1:39:24<00:00, 17.75s/it]
[INFO|trainer.py:3966] 2025-04-13 16:38:46,628 >> Saving model checkpoint to /home/stern/GRPO/offline_rl_v2/output
[INFO|configuration_utils.py:423] 2025-04-13 16:38:46,631 >> Configuration saved in /home/stern/GRPO/offline_rl_v2/output/config.json
[INFO|configuration_utils.py:908] 2025-04-13 16:38:46,632 >> Configuration saved in /home/stern/GRPO/offline_rl_v2/output/generation_config.json
[2025-04-13 16:38:59,677] [INFO] [launch.py:351:main] Process 1025164 exits successfully.
[2025-04-13 16:39:03,682] [INFO] [launch.py:351:main] Process 1025167 exits successfully.
[2025-04-13 16:39:07,686] [INFO] [launch.py:351:main] Process 1025168 exits successfully.
[2025-04-13 16:39:11,691] [INFO] [launch.py:351:main] Process 1025163 exits successfully.
[2025-04-13 16:39:16,696] [INFO] [launch.py:351:main] Process 1025166 exits successfully.
[2025-04-13 16:39:20,700] [INFO] [launch.py:351:main] Process 1025162 exits successfully.
[2025-04-13 16:39:24,705] [INFO] [launch.py:351:main] Process 1025165 exits successfully.
[INFO|modeling_utils.py:3594] 2025-04-13 16:40:11,943 >> The model is bigger than the maximum size per checkpoint (5GB) and is going to be split in 14 checkpoint shards. You can find where each parameters has been saved in the index located at /home/stern/GRPO/offline_rl_v2/output/model.safetensors.index.json.
[INFO|tokenization_utils_base.py:2510] 2025-04-13 16:40:11,945 >> tokenizer config file saved in /home/stern/GRPO/offline_rl_v2/output/tokenizer_config.json
[INFO|tokenization_utils_base.py:2519] 2025-04-13 16:40:11,946 >> Special tokens file saved in /home/stern/GRPO/offline_rl_v2/output/special_tokens_map.json
***** train metrics *****
  epoch                    =        4.0
  total_flos               =    86240GF
  train_loss               =     0.0215
  train_runtime            = 1:39:24.69
  train_samples            =       2688
  train_samples_per_second =      1.803
  train_steps_per_second   =      0.056
[2025-04-13 16:40:24,765] [INFO] [launch.py:351:main] Process 1025161 exits successfully.