apiche

Apiche's group workspace

Group: debug_gspo13

1-3

of 3

Tags

Notes

Author

apiche

State

Crashed

Start time

September 28th, 2025 2:42:50 AM

Runtime

1h 40m 49s

Tracked hours

5m 1s

Run path

apiche/pipeline-rl/debug_gspo13_finetune

Linux-5.15.0-1067-nvidia-x86_64-with-glibc2.39

Python version

CPython 3.11.11

Git repository

git clone git@github.com:ServiceNow/pipelinerl.git

Git state

git checkout -b "debug_gspo13/finetune" 977ccd227a5ec4bbb386300e93092b8194c2af47

Command

pipelinerl/entrypoints/run_finetune.py --config-dir results/debug_gspo13/conf --config-name exp_config output_dir=results/debug_gspo13 hydra.run.dir=results/debug_gspo13/finetune +me.weight_update_group_init_method=tcp://localhost:9000 +me.weight_update_group_world_size=5 +me.llm_urls=http://localhost:8080+http://localhost:8081+http://localhost:8082+http://localhost:8083

System Hardware

CPU count	112
Logical CPU count	224
GPU count	8
GPU type	NVIDIA H100 80GB HBM3

W&B CLI Version

0.19.11

Group

debug_gspo13

Config parameters are your model's inputs. Learn more

▶
Config parameters:{} 185 keys
- _cpu:
  "False"
- _mixed_precision:
  "no"
- accelerate_config:
  null
- actor.discount_factor:
  1
- actor.llm_max_rollouts:
  64
- actor.log_each_n_secs:
  0
- actor.problem_queue_size:
  64
- actor.result_queue_size:
  64
- actor.rollout_policy:
  "pipelinerl.domains.math.generate_math_rollout"
- actor.rollout_workers:
  1
- actor.shared_memory_entry_size:
  10,000,000
- actor.system_prompt:
  "Please reason step by step, and put your final answer within \boxed{}."
- actor.task_prompt:
  ""
- actor.throughput_window_size:
  50
- attempts:
  1
- backend:
  "nccl"
- dataset_loader:
  "pipelinerl.domains.math.load_datasets"
- debug:
  "False"
- debug.mode:
  ""
- debug.place_inference_workers:
  true
- debug.streams_from:
  null
- debug.use_existing_llms:
  false
- deepspeed_config:
  "deepspeed_stage3_bf16"
- deepspeed_plugins:
  "DeepSpeedPlugin(hf_ds_config=<accelerate.utils.deepspeed.HfDeepSpeedConfig object at 0x7ffc246e1fd0>, gradient_accumulation_steps=1, gradient_clipping='auto', zero_stage=3, is_train_batch_min=True, offload_optimizer_device='none', offload_param_device='none', offload_optimizer_nvme_path='none', offload_param_nvme_path='none', zero3_init_flag=True, zero3_save_16bit_model=True, transformer_moe_cls_names=None, enable_msamp=False, msamp_opt_level='O1')"
- device:
  "cuda:0"
- distributed_type:
  "DistributedType.DEEPSPEED"
- dynamo_plugin:
  "TorchDynamoPlugin(backend=<DynamoBackend.NO: 'NO'>, mode='default', fullgraph=False, dynamic=None, options=None, disable=False, use_regional_compilation=False)"
- environment._target_:
  "pipelinerl.domains.math.MathEnvironment"
- eval_every_n_versions:
  0
- finetune.also_save_steps:[] 0 items
- finetune.attempts:
  1
- finetune.attn_implementation:
  "flash_attention_2"
- finetune.auto_device_map:
  false
- finetune.config_name:
  "Qwen/Qwen2.5-0.5B"
- finetune.cuda_empty_cache:
  true
- finetune.data:
  null
- finetune.eval_callback._target_:
  "tapeagents.finetune.eval.dummy_eval_callback"
- finetune.eval_callback.config_name:
  ""
- finetune.force_restart:
  true
- finetune.gradient_accumulation_passes:
  16
- finetune.gradient_checkpointing:
  true
- finetune.gradient_clipping_threshold:
  0.3
- finetune.input:
  "training_data"
- finetune.interrupt_train_steps:
  -1
- finetune.keep_intermediate_checkpoints:
  true
- finetune.learning_rate:
  0.000001
- world.environment_start_port:
  7,777
- world.finetune_fraction:
  1
- world.preprocessor_fraction:
  0
- world.replicas:
  1

Summary metrics are your model's outputs. Learn more

▶
Summary metrics:{} 66 keys
- rl/advantage:
  3.743261814117432
- rl/clamp_log_ratio_new_old_indicator:
  0
- rl/clamp_log_ratio_ref_new_indicator:
  0
- rl/entropy:
  4.092408180236816
- rl/entropy_bonus_coef:
  0
- rl/ess:
  0.6677849530773315
- rl/input_size:
  105.75
- rl/kl:
  5.919403553009033
- rl/kl_coef:
  0
- rl/kl_new_old:
  0.4847049117088318
- rl/loss:
  -182.7752227783203
- rl/max_advantage:
  3.953125
- rl/max_kl:
  69.46488189697266
- rl/max_loss:
  -45.54624938964844
- rl/max_reward:
  0
- rl/max_token_weight:
  0.0625
- rl/min_advantage:
  2.890625
- rl/min_kl:
  0.0005342960357666016
- rl/min_loss:
  -45.8702392578125
- rl/min_reward:
  0
- rl/min_token_weight:
  0.0625
- rl/new_logprobs:
  -4.686008930206299
- rl/num_output_tokens_sum:
  40
- rl/old_logprobs:
  -4.1688618659973145
- rl/ratio_new_old:
  0.9675576686859132
- rl/ratio_new_old_squared_sum:
  55.71575927734375
- rl/ratio_new_old_sum:
  38.577789306640625
- rl/ratio_ref_new:
  7.436551094055176
- rl/ratio_ref_old:
  1
- rl/ref_logprobs:
  -4.1688618659973145
- rl/reward:
  0
- rl/token_weight:
  0.0625
- rl/value_loss:
  17.377090454101562
- rl/value_max:
  -2.890625
- rl/value_mean:
  -3.743261814117432
- rl/value_min:
  -3.953125
- rl/value_mse:
  14.041807174682615
- stats/completed_steps:
  139
- stats/epoch:
  0
- stats/grad_norm:
  501.21562853260406
- stats/lag:
  288
- stats/lr:
  0.000001
- stats/max_actor_version:
  1,920
- stats/min_actor_version:
  1,920
- stats/passes:
  155
- stats/queue/batches:
  1
- throughput/tokens_per_micro_batch:
  423
- throughput/tokens_per_sec:
  2,872.9807568211795
- throughput/tokens_per_step:
  1,692
- throughput/tokens_perGPU_per_sec:
  718.2451892052949

This run produced these artifacts as outputs. Total: 1. Learn more

wandb-history

run-debug_gspo13_finetune-history:v0