Add files using upload-large-folder tool

outputs/2025-12-17/14-54-44/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/14-56-09/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/14-56-09
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/16-01-29/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/16-01-29
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/16-12-46/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/16-12-46
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/16-12-46/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/16-36-56/.hydra/config.yaml

@@ -0,0 +1,897 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 3600
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: 1
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 25
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 128
+    group_size: 1
+    env_configs:
+      tags:
+      - rubikscube
+      n_groups:
+      - 128
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: gpt-4o-mini
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/16-51-20/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/17-26-19/.hydra/hydra.yaml

@@ -0,0 +1,179 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - system.CUDA_VISIBLE_DEVICES='0,1,2,3,4,5,6,7'
+    - trainer.project_name=ragen_latest_qwen_25_3b_it
+    - trainer.n_gpus_per_node=8
+    - model_path=Qwen/Qwen2.5-3B-Instruct
+    - trainer.experiment_name=frozenlake-static-starpos-stateestimate
+    - algorithm.adv_estimator=gae
+    - algorithm.kl_ctrl.kl_coef=0.0
+    - actor_rollout_ref.actor.kl_loss_coef=0.0
+    - actor_rollout_ref.actor.clip_ratio_high=0.28
+    - actor_rollout_ref.rollout.rollout_filter_ratio=0.25
+  job:
+    name: train
+    chdir: null
+    override_dirname: actor_rollout_ref.actor.clip_ratio_high=0.28,actor_rollout_ref.actor.kl_loss_coef=0.0,actor_rollout_ref.rollout.rollout_filter_ratio=0.25,algorithm.adv_estimator=gae,algorithm.kl_ctrl.kl_coef=0.0,model_path=Qwen/Qwen2.5-3B-Instruct,system.CUDA_VISIBLE_DEVICES='0,1,2,3,4,5,6,7',trainer.experiment_name=frozenlake-static-starpos-stateestimate,trainer.n_gpus_per_node=8,trainer.project_name=ragen_latest_qwen_25_3b_it
+    id: ???
+    num: ???
+    config_name: _3_frozen_lake
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: /mnt/general/wanghy/RAGEN/verl/verl/trainer/config
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/17-26-19
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/17-26-19/.hydra/overrides.yaml

@@ -0,0 +1,10 @@
+- system.CUDA_VISIBLE_DEVICES='0,1,2,3,4,5,6,7'
+- trainer.project_name=ragen_latest_qwen_25_3b_it
+- trainer.n_gpus_per_node=8
+- model_path=Qwen/Qwen2.5-3B-Instruct
+- trainer.experiment_name=frozenlake-static-starpos-stateestimate
+- algorithm.adv_estimator=gae
+- algorithm.kl_ctrl.kl_coef=0.0
+- actor_rollout_ref.actor.kl_loss_coef=0.0
+- actor_rollout_ref.actor.clip_ratio_high=0.28
+- actor_rollout_ref.rollout.rollout_filter_ratio=0.25

outputs/2025-12-17/17-57-14/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/18-05-39/.hydra/config.yaml

@@ -0,0 +1,861 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 3600
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: ragen_latest_qwen_25_3b_it
+  experiment_name: sokoban-box2-starpos-SPA
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 8
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: 'You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets You are provided with a symbol grid and the
+      zero-indexed coordinates of the player, each box, and each target. Coordinates
+      range from the top-left corner (0, 0) to the bottom-right corner (5, 5). When
+      you are exactly next to a box, you can push it by moving in the same direction.
+      You cannot push a box through a wall, and you cannot pull a box. The answer
+      should be a sequence of actions, like <answer>Right || Right || Up</answer>.
+      A sample full output is as follows: <think> <observation>######\n#_####\n#_P###\n#_X#_#\n#__O_#\n######\nPlayer
+      (P) is at (2,2); box (X) is at (3,2); target (O) is at (4,3).</observation>
+      1 Down – I push box to (4,2). 2 Left – I step to (3,1). 3 Down – I stand left
+      of box, ready to push it Right onto target. <prediction>######\n#_####\n#__###\n#__#_#\n#PXO_#\n######</prediction>
+      </think> <answer> Down || Left || Down </answer>'
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: 3
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: 0,1,2,3,4,5,6,7
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: /mnt/general/wanghy/SPA/sftckpt/checkpoints_2_sokoban-3B-text_with_coordinates-qwen/global_step_1375
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 25
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 256
+    group_size: 1
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 256
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null

outputs/2025-12-17/18-05-39/.hydra/hydra.yaml

@@ -0,0 +1,204 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath: []
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.mode=RUN
+    task:
+    - system.CUDA_VISIBLE_DEVICES='0,1,2,3,4,5,6,7'
+    - trainer.project_name=ragen_latest_qwen_25_3b_it
+    - trainer.n_gpus_per_node=8
+    - model_path=/mnt/general/wanghy/SPA/sftckpt/checkpoints_2_sokoban-3B-text_with_coordinates-qwen/global_step_1375
+    - custom_envs.CoordSokoban.env_config.num_boxes=2
+    - 'custom_envs.CoordSokoban.env_instruction=''You are solving the Sokoban puzzle.
+      You are the player and you need to push all boxes to targets You are provided
+      with a symbol grid and the zero-indexed coordinates of the player, each box,
+      and each target. Coordinates range from the top-left corner (0, 0) to the bottom-right
+      corner (5, 5). When you are exactly next to a box, you can push it by moving
+      in the same direction. You cannot push a box through a wall, and you cannot
+      pull a box. The answer should be a sequence of actions, like <answer>Right ||
+      Right || Up</answer>. A sample full output is as follows: <think> <observation>######\n#_####\n#_P###\n#_X#_#\n#__O_#\n######\nPlayer
+      (P) is at (2,2); box (X) is at (3,2); target (O) is at (4,3).</observation>
+      1 Down – I push box to (4,2). 2 Left – I step to (3,1). 3 Down – I stand left
+      of box, ready to push it Right onto target. <prediction>######\n#_####\n#__###\n#__#_#\n#PXO_#\n######</prediction>
+      </think> <answer> Down || Left || Down </answer>'''
+    - trainer.experiment_name=sokoban-box2-starpos-SPA
+    - algorithm.adv_estimator=gae
+    - algorithm.kl_ctrl.kl_coef=0.0
+    - actor_rollout_ref.actor.kl_loss_coef=0.0
+    - actor_rollout_ref.actor.clip_ratio_high=0.28
+    - actor_rollout_ref.rollout.rollout_filter_ratio=0.25
+  job:
+    name: train
+    chdir: null
+    override_dirname: 'actor_rollout_ref.actor.clip_ratio_high=0.28,actor_rollout_ref.actor.kl_loss_coef=0.0,actor_rollout_ref.rollout.rollout_filter_ratio=0.25,algorithm.adv_estimator=gae,algorithm.kl_ctrl.kl_coef=0.0,custom_envs.CoordSokoban.env_config.num_boxes=2,custom_envs.CoordSokoban.env_instruction=''You
+      are solving the Sokoban puzzle. You are the player and you need to push all
+      boxes to targets You are provided with a symbol grid and the zero-indexed coordinates
+      of the player, each box, and each target. Coordinates range from the top-left
+      corner (0, 0) to the bottom-right corner (5, 5). When you are exactly next to
+      a box, you can push it by moving in the same direction. You cannot push a box
+      through a wall, and you cannot pull a box. The answer should be a sequence of
+      actions, like <answer>Right || Right || Up</answer>. A sample full output is
+      as follows: <think> <observation>######\n#_####\n#_P###\n#_X#_#\n#__O_#\n######\nPlayer
+      (P) is at (2,2); box (X) is at (3,2); target (O) is at (4,3).</observation>
+      1 Down – I push box to (4,2). 2 Left – I step to (3,1). 3 Down – I stand left
+      of box, ready to push it Right onto target. <prediction>######\n#_####\n#__###\n#__#_#\n#PXO_#\n######</prediction>
+      </think> <answer> Down || Left || Down </answer>'',model_path=/mnt/general/wanghy/SPA/sftckpt/checkpoints_2_sokoban-3B-text_with_coordinates-qwen/global_step_1375,system.CUDA_VISIBLE_DEVICES=''0,1,2,3,4,5,6,7'',trainer.experiment_name=sokoban-box2-starpos-SPA,trainer.n_gpus_per_node=8,trainer.project_name=ragen_latest_qwen_25_3b_it'
+    id: ???
+    num: ???
+    config_name: _2_sokoban
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: /mnt/general/wanghy/RAGEN/verl/verl/trainer/config
+      schema: file
+      provider: command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/18-05-39
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/18-24-23/.hydra/config.yaml

@@ -0,0 +1,897 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 14400
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: 3
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 700
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 128
+    group_size: 1
+    env_configs:
+      tags:
+      - game_2048
+      n_groups:
+      - 128
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: gpt-4o-mini
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/18-24-23/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/18-24-23
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/18-46-27/.hydra/config.yaml

@@ -0,0 +1,897 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 7200
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 300
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: 3
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 300
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 8
+    group_size: 1
+    env_configs:
+      tags:
+      - game_2048
+      n_groups:
+      - 8
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: gpt-4o-mini
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/18-46-27/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/19-17-08/.hydra/config.yaml

@@ -0,0 +1,897 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 64000
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 100
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: null
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 100
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 1
+    group_size: 1
+    env_configs:
+      tags:
+      - game_2048
+      n_groups:
+      - 1
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: gpt-4o-mini
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/19-35-11/.hydra/config.yaml

@@ -0,0 +1,897 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 64000
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 400
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: null
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 400
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 1
+    group_size: 1
+    env_configs:
+      tags:
+      - game_2048
+      n_groups:
+      - 1
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: gpt-4o-mini
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/19-35-11/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/19-35-11
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/19-35-11/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/20-20-56/.hydra/config.yaml

@@ -0,0 +1,903 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 3600
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: null
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 25
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 128
+    group_size: 1
+    env_configs:
+      tags:
+      - Bandit
+      n_groups:
+      - 128
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: ark-deepseek-v3-250324
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  ark-deepseek-v3-250324:
+    provider_name: openai
+    model_name: ark-deepseek-v3-250324
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/20-20-56/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/20-20-56
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/20-20-56/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/20-20-56/eval_api.log

@@ -0,0 +1,148 @@
+[2025-12-17 20:21:03,433][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:03,727][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:03,791][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:03,851][openai._base_client][INFO] - Retrying request to /chat/completions in 0.410615 seconds
+[2025-12-17 20:21:03,851][openai._base_client][INFO] - Retrying request to /chat/completions in 0.385690 seconds
+[2025-12-17 20:21:03,856][openai._base_client][INFO] - Retrying request to /chat/completions in 0.383690 seconds
+[2025-12-17 20:21:04,216][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:04,252][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:04,297][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:04,530][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:04,622][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:04,773][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:04,807][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:05,052][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:05,414][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:06,006][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:08,641][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:08,775][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:08,897][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:08,956][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:09,030][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:09,243][openai._base_client][INFO] - Retrying request to /chat/completions in 0.937736 seconds
+[2025-12-17 20:21:09,244][openai._base_client][INFO] - Retrying request to /chat/completions in 0.751963 seconds
+[2025-12-17 20:21:09,265][openai._base_client][INFO] - Retrying request to /chat/completions in 0.778112 seconds
+[2025-12-17 20:21:09,915][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:10,251][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:10,336][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:10,478][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:10,493][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:10,994][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:11,145][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:11,736][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:13,484][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:13,923][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:14,230][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:14,411][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:15,046][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:15,057][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:15,069][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:15,848][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:15,928][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:16,016][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:16,098][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:17,363][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:18,158][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:18,829][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:18,974][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:19,056][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:19,417][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:19,608][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:20,003][openai._base_client][INFO] - Retrying request to /chat/completions in 0.462326 seconds
+[2025-12-17 20:21:20,052][openai._base_client][INFO] - Retrying request to /chat/completions in 0.377625 seconds
+[2025-12-17 20:21:20,190][openai._base_client][INFO] - Retrying request to /chat/completions in 0.479703 seconds
+[2025-12-17 20:21:20,752][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:21,124][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:21,164][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:21,264][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:21,695][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:22,356][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:22,703][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:23,392][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:23,817][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:24,745][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:24,772][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:25,195][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:25,251][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:25,434][openai._base_client][INFO] - Retrying request to /chat/completions in 0.982149 seconds
+[2025-12-17 20:21:25,447][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:25,470][openai._base_client][INFO] - Retrying request to /chat/completions in 0.961314 seconds
+[2025-12-17 20:21:25,674][openai._base_client][INFO] - Retrying request to /chat/completions in 0.963079 seconds
+[2025-12-17 20:21:25,779][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:25,904][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:26,153][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:26,459][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:26,771][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:28,561][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:28,827][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:28,946][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:29,427][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:29,627][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:30,340][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:30,509][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:30,818][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:31,067][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:31,316][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:31,785][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:31,990][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:32,046][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:32,228][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:33,694][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:33,799][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:34,010][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:34,306][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:34,905][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:35,312][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:36,234][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:36,424][openai._base_client][INFO] - Retrying request to /chat/completions in 0.438620 seconds
+[2025-12-17 20:21:36,440][openai._base_client][INFO] - Retrying request to /chat/completions in 0.474563 seconds
+[2025-12-17 20:21:36,598][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:36,720][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:37,074][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:37,170][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:37,692][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:37,706][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:37,919][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:37,934][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:39,344][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:39,457][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:39,989][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:40,374][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:40,909][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:40,933][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:41,871][openai._base_client][INFO] - Retrying request to /chat/completions in 0.881132 seconds
+[2025-12-17 20:21:42,060][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:42,125][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:42,488][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:42,732][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:42,825][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:42,971][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:43,242][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:43,246][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:43,262][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:43,987][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:44,916][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:45,341][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:45,547][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:46,561][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:46,695][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:47,156][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:47,197][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:47,239][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:48,196][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:48,211][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:48,424][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:48,887][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:49,036][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:49,217][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:21:54,228][openai._base_client][INFO] - Retrying request to /chat/completions in 0.442995 seconds
+[2025-12-17 20:21:59,248][openai._base_client][INFO] - Retrying request to /chat/completions in 0.482630 seconds
+[2025-12-17 20:21:59,249][openai._base_client][INFO] - Retrying request to /chat/completions in 0.486855 seconds
+[2025-12-17 20:21:59,249][openai._base_client][INFO] - Retrying request to /chat/completions in 0.486917 seconds
+[2025-12-17 20:21:59,627][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:22:00,055][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:22:00,139][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:22:04,740][openai._base_client][INFO] - Retrying request to /chat/completions in 0.930248 seconds
+[2025-12-17 20:22:05,182][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:22:05,283][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:22:12,992][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"

outputs/2025-12-17/20-24-27/.hydra/config.yaml

@@ -0,0 +1,903 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 64000
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 400
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: null
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 400
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 1
+    group_size: 1
+    env_configs:
+      tags:
+      - game_2048
+      n_groups:
+      - 1
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: ark-deepseek-v3-250324
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  ark-deepseek-v3-250324:
+    provider_name: openai
+    model_name: ark-deepseek-v3-250324
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/20-24-27/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/20-24-27
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/20-24-27/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/20-24-27/eval_api.log

@@ -0,0 +1,673 @@
+[2025-12-17 20:24:37,702][huggingface_hub.utils._http][WARNING] - '(MaxRetryError("HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /Qwen/Qwen2.5-0.5B-Instruct/resolve/main/tokenizer_config.json (Caused by NewConnectionError('<urllib3.connection.HTTPSConnection object at 0x7f7551274880>: Failed to establish a new connection: [Errno 101] Network is unreachable'))"), '(Request ID: 0a6b4e7a-9344-47b9-8660-d93003be9e05)')' thrown while requesting HEAD https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct/resolve/main/tokenizer_config.json
+[2025-12-17 20:24:37,703][huggingface_hub.utils._http][WARNING] - Retrying in 1s [Retry 1/5].
+[2025-12-17 20:24:48,719][huggingface_hub.utils._http][WARNING] - '(MaxRetryError("HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /Qwen/Qwen2.5-0.5B-Instruct/resolve/main/tokenizer_config.json (Caused by NewConnectionError('<urllib3.connection.HTTPSConnection object at 0x7f7551274bb0>: Failed to establish a new connection: [Errno 101] Network is unreachable'))"), '(Request ID: 8a9560d5-7f8d-4009-b17c-e8f33d83c67d)')' thrown while requesting HEAD https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct/resolve/main/tokenizer_config.json
+[2025-12-17 20:24:48,720][huggingface_hub.utils._http][WARNING] - Retrying in 2s [Retry 2/5].
+[2025-12-17 20:24:57,833][openai._base_client][INFO] - Retrying request to /chat/completions in 0.376132 seconds
+[2025-12-17 20:25:06,847][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:25:06,870][openai._base_client][INFO] - Retrying request to /chat/completions in 0.452581 seconds
+[2025-12-17 20:25:15,239][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:25:15,261][openai._base_client][INFO] - Retrying request to /chat/completions in 0.411661 seconds
+[2025-12-17 20:25:20,683][openai._base_client][INFO] - Retrying request to /chat/completions in 0.828438 seconds
+[2025-12-17 20:25:41,793][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:25:41,806][openai._base_client][INFO] - Retrying request to /chat/completions in 0.425828 seconds
+[2025-12-17 20:25:49,814][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:25:49,831][openai._base_client][INFO] - Retrying request to /chat/completions in 0.408170 seconds
+[2025-12-17 20:25:55,250][openai._base_client][INFO] - Retrying request to /chat/completions in 0.776128 seconds
+[2025-12-17 20:26:15,324][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:26:15,342][openai._base_client][INFO] - Retrying request to /chat/completions in 0.433717 seconds
+[2025-12-17 20:26:24,024][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:26:24,046][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413058 seconds
+[2025-12-17 20:26:29,470][openai._base_client][INFO] - Retrying request to /chat/completions in 0.767873 seconds
+[2025-12-17 20:26:37,395][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:26:37,427][openai._base_client][INFO] - Retrying request to /chat/completions in 0.445893 seconds
+[2025-12-17 20:26:45,926][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:26:45,945][openai._base_client][INFO] - Retrying request to /chat/completions in 0.401360 seconds
+[2025-12-17 20:26:56,738][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:26:56,771][openai._base_client][INFO] - Retrying request to /chat/completions in 0.417793 seconds
+[2025-12-17 20:27:02,196][openai._base_client][INFO] - Retrying request to /chat/completions in 0.862126 seconds
+[2025-12-17 20:27:28,757][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:27:28,796][openai._base_client][INFO] - Retrying request to /chat/completions in 0.387259 seconds
+[2025-12-17 20:27:39,017][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:27:39,047][openai._base_client][INFO] - Retrying request to /chat/completions in 0.469462 seconds
+[2025-12-17 20:27:49,617][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:27:49,667][openai._base_client][INFO] - Retrying request to /chat/completions in 0.452411 seconds
+[2025-12-17 20:27:58,664][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:27:58,707][openai._base_client][INFO] - Retrying request to /chat/completions in 0.426367 seconds
+[2025-12-17 20:28:09,831][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:28:09,879][openai._base_client][INFO] - Retrying request to /chat/completions in 0.453611 seconds
+[2025-12-17 20:28:15,340][openai._base_client][INFO] - Retrying request to /chat/completions in 0.803844 seconds
+[2025-12-17 20:28:31,172][openai._base_client][INFO] - Retrying request to /chat/completions in 0.498329 seconds
+[2025-12-17 20:28:36,680][openai._base_client][INFO] - Retrying request to /chat/completions in 0.937599 seconds
+[2025-12-17 20:28:52,649][openai._base_client][INFO] - Retrying request to /chat/completions in 0.480890 seconds
+[2025-12-17 20:28:58,139][openai._base_client][INFO] - Retrying request to /chat/completions in 0.753952 seconds
+[2025-12-17 20:29:13,922][openai._base_client][INFO] - Retrying request to /chat/completions in 0.444222 seconds
+[2025-12-17 20:29:19,375][openai._base_client][INFO] - Retrying request to /chat/completions in 0.938501 seconds
+[2025-12-17 20:29:32,363][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:29:32,407][openai._base_client][INFO] - Retrying request to /chat/completions in 0.487606 seconds
+[2025-12-17 20:29:43,744][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:29:43,783][openai._base_client][INFO] - Retrying request to /chat/completions in 0.393820 seconds
+[2025-12-17 20:29:55,387][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:29:55,439][openai._base_client][INFO] - Retrying request to /chat/completions in 0.417551 seconds
+[2025-12-17 20:30:06,532][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:30:06,586][openai._base_client][INFO] - Retrying request to /chat/completions in 0.490828 seconds
+[2025-12-17 20:30:12,084][openai._base_client][INFO] - Retrying request to /chat/completions in 0.906120 seconds
+[2025-12-17 20:30:24,870][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:30:24,927][openai._base_client][INFO] - Retrying request to /chat/completions in 0.496967 seconds
+[2025-12-17 20:30:36,117][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:30:36,187][openai._base_client][INFO] - Retrying request to /chat/completions in 0.458074 seconds
+[2025-12-17 20:30:48,509][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:30:48,592][openai._base_client][INFO] - Retrying request to /chat/completions in 0.497333 seconds
+[2025-12-17 20:31:00,269][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:31:00,369][openai._base_client][INFO] - Retrying request to /chat/completions in 0.437797 seconds
+[2025-12-17 20:31:10,200][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:31:10,291][openai._base_client][INFO] - Retrying request to /chat/completions in 0.465757 seconds
+[2025-12-17 20:31:22,015][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:31:22,064][openai._base_client][INFO] - Retrying request to /chat/completions in 0.446729 seconds
+[2025-12-17 20:41:23,655][openai._base_client][INFO] - Retrying request to /chat/completions in 0.914445 seconds
+[2025-12-17 20:41:45,315][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:41:45,370][openai._base_client][INFO] - Retrying request to /chat/completions in 0.403432 seconds
+[2025-12-17 20:41:57,481][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:41:57,586][openai._base_client][INFO] - Retrying request to /chat/completions in 0.451162 seconds
+[2025-12-17 20:42:03,046][openai._base_client][INFO] - Retrying request to /chat/completions in 0.852434 seconds
+[2025-12-17 20:42:14,386][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:42:14,485][openai._base_client][INFO] - Retrying request to /chat/completions in 0.380134 seconds
+[2025-12-17 20:42:25,157][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:42:25,242][openai._base_client][INFO] - Retrying request to /chat/completions in 0.416152 seconds
+[2025-12-17 20:42:30,679][openai._base_client][INFO] - Retrying request to /chat/completions in 0.994369 seconds
+[2025-12-17 20:42:43,452][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:42:43,536][openai._base_client][INFO] - Retrying request to /chat/completions in 0.475820 seconds
+[2025-12-17 20:42:54,385][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:42:54,497][openai._base_client][INFO] - Retrying request to /chat/completions in 0.461195 seconds
+[2025-12-17 20:43:05,263][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:43:05,374][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413379 seconds
+[2025-12-17 20:43:10,797][openai._base_client][INFO] - Retrying request to /chat/completions in 0.769987 seconds
+[2025-12-17 20:43:22,953][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:43:23,041][openai._base_client][INFO] - Retrying request to /chat/completions in 0.419698 seconds
+[2025-12-17 20:43:28,470][openai._base_client][INFO] - Retrying request to /chat/completions in 0.857271 seconds
+[2025-12-17 20:43:44,371][openai._base_client][INFO] - Retrying request to /chat/completions in 0.406330 seconds
+[2025-12-17 20:43:58,602][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:43:58,691][openai._base_client][INFO] - Retrying request to /chat/completions in 0.471129 seconds
+[2025-12-17 20:44:11,029][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:44:11,107][openai._base_client][INFO] - Retrying request to /chat/completions in 0.432727 seconds
+[2025-12-17 20:44:16,548][openai._base_client][INFO] - Retrying request to /chat/completions in 0.773695 seconds
+[2025-12-17 20:44:32,539][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:44:32,669][openai._base_client][INFO] - Retrying request to /chat/completions in 0.444900 seconds
+[2025-12-17 20:44:45,487][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:44:45,559][openai._base_client][INFO] - Retrying request to /chat/completions in 0.446846 seconds
+[2025-12-17 20:44:56,906][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:44:57,005][openai._base_client][INFO] - Retrying request to /chat/completions in 0.428509 seconds
+[2025-12-17 20:45:09,826][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:45:09,948][openai._base_client][INFO] - Retrying request to /chat/completions in 0.404253 seconds
+[2025-12-17 20:45:22,299][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:45:22,427][openai._base_client][INFO] - Retrying request to /chat/completions in 0.429378 seconds
+[2025-12-17 20:45:35,589][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:45:35,695][openai._base_client][INFO] - Retrying request to /chat/completions in 0.473082 seconds
+[2025-12-17 20:45:47,724][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:45:47,856][openai._base_client][INFO] - Retrying request to /chat/completions in 0.382924 seconds
+[2025-12-17 20:46:00,402][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:46:00,509][openai._base_client][INFO] - Retrying request to /chat/completions in 0.433030 seconds
+[2025-12-17 20:46:20,941][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:46:21,035][openai._base_client][INFO] - Retrying request to /chat/completions in 0.467294 seconds
+[2025-12-17 20:46:33,286][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:46:33,413][openai._base_client][INFO] - Retrying request to /chat/completions in 0.405314 seconds
+[2025-12-17 20:46:48,204][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:46:48,305][openai._base_client][INFO] - Retrying request to /chat/completions in 0.492102 seconds
+[2025-12-17 20:47:02,430][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:47:02,559][openai._base_client][INFO] - Retrying request to /chat/completions in 0.450762 seconds
+[2025-12-17 20:47:14,805][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:47:14,980][openai._base_client][INFO] - Retrying request to /chat/completions in 0.440584 seconds
+[2025-12-17 20:47:20,430][openai._base_client][INFO] - Retrying request to /chat/completions in 0.992334 seconds
+[2025-12-17 20:47:35,252][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:47:35,386][openai._base_client][INFO] - Retrying request to /chat/completions in 0.420410 seconds
+[2025-12-17 20:47:40,817][openai._base_client][INFO] - Retrying request to /chat/completions in 0.895212 seconds
+[2025-12-17 20:47:54,563][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:47:54,720][openai._base_client][INFO] - Retrying request to /chat/completions in 0.380347 seconds
+[2025-12-17 20:48:05,405][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:48:05,500][openai._base_client][INFO] - Retrying request to /chat/completions in 0.381711 seconds
+[2025-12-17 20:48:20,026][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:48:20,154][openai._base_client][INFO] - Retrying request to /chat/completions in 0.415074 seconds
+[2025-12-17 20:48:25,580][openai._base_client][INFO] - Retrying request to /chat/completions in 0.939974 seconds
+[2025-12-17 20:48:37,550][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:48:37,728][openai._base_client][INFO] - Retrying request to /chat/completions in 0.447852 seconds
+[2025-12-17 20:48:50,476][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:48:50,637][openai._base_client][INFO] - Retrying request to /chat/completions in 0.487941 seconds
+[2025-12-17 20:49:07,468][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:49:07,645][openai._base_client][INFO] - Retrying request to /chat/completions in 0.433020 seconds
+[2025-12-17 20:49:19,805][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:49:19,955][openai._base_client][INFO] - Retrying request to /chat/completions in 0.453066 seconds
+[2025-12-17 20:49:33,619][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:49:33,778][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413746 seconds
+[2025-12-17 20:49:44,616][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:49:44,750][openai._base_client][INFO] - Retrying request to /chat/completions in 0.499446 seconds
+[2025-12-17 20:49:57,458][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:49:57,665][openai._base_client][INFO] - Retrying request to /chat/completions in 0.396918 seconds
+[2025-12-17 20:50:09,774][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:50:09,965][openai._base_client][INFO] - Retrying request to /chat/completions in 0.394980 seconds
+[2025-12-17 20:50:23,560][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:50:23,697][openai._base_client][INFO] - Retrying request to /chat/completions in 0.403515 seconds
+[2025-12-17 20:50:40,642][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:50:40,845][openai._base_client][INFO] - Retrying request to /chat/completions in 0.440986 seconds
+[2025-12-17 20:50:53,122][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:50:53,259][openai._base_client][INFO] - Retrying request to /chat/completions in 0.463452 seconds
+[2025-12-17 20:50:53,802][openai._base_client][INFO] - Retrying request to /chat/completions in 0.925208 seconds
+[2025-12-17 20:51:07,341][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:51:07,484][openai._base_client][INFO] - Retrying request to /chat/completions in 0.489543 seconds
+[2025-12-17 20:51:08,075][openai._base_client][INFO] - Retrying request to /chat/completions in 0.759774 seconds
+[2025-12-17 20:51:20,904][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:51:21,092][openai._base_client][INFO] - Retrying request to /chat/completions in 0.454143 seconds
+[2025-12-17 20:51:34,510][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:51:34,689][openai._base_client][INFO] - Retrying request to /chat/completions in 0.396712 seconds
+[2025-12-17 20:51:40,095][openai._base_client][INFO] - Retrying request to /chat/completions in 0.861944 seconds
+[2025-12-17 20:51:55,765][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:51:55,950][openai._base_client][INFO] - Retrying request to /chat/completions in 0.460006 seconds
+[2025-12-17 20:52:08,339][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:52:08,562][openai._base_client][INFO] - Retrying request to /chat/completions in 0.487827 seconds
+[2025-12-17 20:52:20,540][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:52:20,763][openai._base_client][INFO] - Retrying request to /chat/completions in 0.414528 seconds
+[2025-12-17 20:52:34,885][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:52:35,068][openai._base_client][INFO] - Retrying request to /chat/completions in 0.423166 seconds
+[2025-12-17 20:52:35,570][openai._base_client][INFO] - Retrying request to /chat/completions in 0.982388 seconds
+[2025-12-17 20:52:51,455][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:52:51,628][openai._base_client][INFO] - Retrying request to /chat/completions in 0.377388 seconds
+[2025-12-17 20:52:52,084][openai._base_client][INFO] - Retrying request to /chat/completions in 0.966061 seconds
+[2025-12-17 20:53:07,787][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:53:07,973][openai._base_client][INFO] - Retrying request to /chat/completions in 0.407450 seconds
+[2025-12-17 20:53:23,041][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:53:23,201][openai._base_client][INFO] - Retrying request to /chat/completions in 0.456685 seconds
+[2025-12-17 20:53:36,895][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:53:37,110][openai._base_client][INFO] - Retrying request to /chat/completions in 0.388334 seconds
+[2025-12-17 20:53:54,315][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:53:54,531][openai._base_client][INFO] - Retrying request to /chat/completions in 0.408942 seconds
+[2025-12-17 20:54:11,079][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:54:11,302][openai._base_client][INFO] - Retrying request to /chat/completions in 0.461260 seconds
+[2025-12-17 20:54:16,773][openai._base_client][INFO] - Retrying request to /chat/completions in 0.876180 seconds
+[2025-12-17 20:54:32,695][openai._base_client][INFO] - Retrying request to /chat/completions in 0.479562 seconds
+[2025-12-17 20:54:46,770][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:54:46,984][openai._base_client][INFO] - Retrying request to /chat/completions in 0.392750 seconds
+[2025-12-17 20:54:52,385][openai._base_client][INFO] - Retrying request to /chat/completions in 0.914224 seconds
+[2025-12-17 20:55:08,348][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413191 seconds
+[2025-12-17 20:55:23,677][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:55:23,860][openai._base_client][INFO] - Retrying request to /chat/completions in 0.480641 seconds
+[2025-12-17 20:55:29,349][openai._base_client][INFO] - Retrying request to /chat/completions in 0.894725 seconds
+[2025-12-17 20:55:45,298][openai._base_client][INFO] - Retrying request to /chat/completions in 0.458802 seconds
+[2025-12-17 20:56:00,128][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:56:00,302][openai._base_client][INFO] - Retrying request to /chat/completions in 0.472512 seconds
+[2025-12-17 20:56:05,786][openai._base_client][INFO] - Retrying request to /chat/completions in 0.865616 seconds
+[2025-12-17 20:56:20,512][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:56:20,742][openai._base_client][INFO] - Retrying request to /chat/completions in 0.386785 seconds
+[2025-12-17 20:56:26,138][openai._base_client][INFO] - Retrying request to /chat/completions in 0.949151 seconds
+[2025-12-17 20:56:41,314][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:56:41,545][openai._base_client][INFO] - Retrying request to /chat/completions in 0.458188 seconds
+[2025-12-17 20:56:55,474][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:56:55,696][openai._base_client][INFO] - Retrying request to /chat/completions in 0.376590 seconds
+[2025-12-17 20:57:11,189][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:57:11,433][openai._base_client][INFO] - Retrying request to /chat/completions in 0.432420 seconds
+[2025-12-17 20:57:16,874][openai._base_client][INFO] - Retrying request to /chat/completions in 0.841450 seconds
+[2025-12-17 20:57:30,956][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:57:31,229][openai._base_client][INFO] - Retrying request to /chat/completions in 0.380599 seconds
+[2025-12-17 20:57:36,619][openai._base_client][INFO] - Retrying request to /chat/completions in 0.888379 seconds
+[2025-12-17 20:58:01,217][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:58:01,486][openai._base_client][INFO] - Retrying request to /chat/completions in 0.432111 seconds
+[2025-12-17 20:58:17,199][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:58:17,505][openai._base_client][INFO] - Retrying request to /chat/completions in 0.427784 seconds
+[2025-12-17 20:58:31,459][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:58:31,728][openai._base_client][INFO] - Retrying request to /chat/completions in 0.429981 seconds
+[2025-12-17 20:58:37,167][openai._base_client][INFO] - Retrying request to /chat/completions in 0.800965 seconds
+[2025-12-17 20:58:56,872][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:58:57,219][openai._base_client][INFO] - Retrying request to /chat/completions in 0.457042 seconds
+[2025-12-17 20:59:02,687][openai._base_client][INFO] - Retrying request to /chat/completions in 0.990057 seconds
+[2025-12-17 20:59:17,378][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:59:17,699][openai._base_client][INFO] - Retrying request to /chat/completions in 0.467380 seconds
+[2025-12-17 20:59:32,338][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:59:32,733][openai._base_client][INFO] - Retrying request to /chat/completions in 0.428658 seconds
+[2025-12-17 20:59:53,061][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 20:59:53,261][openai._base_client][INFO] - Retrying request to /chat/completions in 0.448933 seconds
+[2025-12-17 21:00:06,884][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:00:07,067][openai._base_client][INFO] - Retrying request to /chat/completions in 0.403251 seconds
+[2025-12-17 21:00:20,468][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:00:20,681][openai._base_client][INFO] - Retrying request to /chat/completions in 0.406592 seconds
+[2025-12-17 21:00:26,096][openai._base_client][INFO] - Retrying request to /chat/completions in 0.971805 seconds
+[2025-12-17 21:00:44,100][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:00:44,340][openai._base_client][INFO] - Retrying request to /chat/completions in 0.468274 seconds
+[2025-12-17 21:00:58,511][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:00:58,800][openai._base_client][INFO] - Retrying request to /chat/completions in 0.387688 seconds
+[2025-12-17 21:01:15,089][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:01:15,355][openai._base_client][INFO] - Retrying request to /chat/completions in 0.461001 seconds
+[2025-12-17 21:01:20,824][openai._base_client][INFO] - Retrying request to /chat/completions in 0.922049 seconds
+[2025-12-17 21:01:36,803][openai._base_client][INFO] - Retrying request to /chat/completions in 0.398734 seconds
+[2025-12-17 21:01:42,212][openai._base_client][INFO] - Retrying request to /chat/completions in 0.808116 seconds
+[2025-12-17 21:02:08,639][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:02:08,911][openai._base_client][INFO] - Retrying request to /chat/completions in 0.450400 seconds
+[2025-12-17 21:02:24,490][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:02:24,809][openai._base_client][INFO] - Retrying request to /chat/completions in 0.454115 seconds
+[2025-12-17 21:02:38,523][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:02:38,879][openai._base_client][INFO] - Retrying request to /chat/completions in 0.430310 seconds
+[2025-12-17 21:02:55,567][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:02:55,801][openai._base_client][INFO] - Retrying request to /chat/completions in 0.483469 seconds
+[2025-12-17 21:03:01,294][openai._base_client][INFO] - Retrying request to /chat/completions in 0.887115 seconds
+[2025-12-17 21:03:17,464][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:03:17,738][openai._base_client][INFO] - Retrying request to /chat/completions in 0.384835 seconds
+[2025-12-17 21:03:23,134][openai._base_client][INFO] - Retrying request to /chat/completions in 0.985838 seconds
+[2025-12-17 21:03:39,207][openai._base_client][INFO] - Retrying request to /chat/completions in 0.465483 seconds
+[2025-12-17 21:03:44,684][openai._base_client][INFO] - Retrying request to /chat/completions in 0.904536 seconds
+[2025-12-17 21:04:00,654][openai._base_client][INFO] - Retrying request to /chat/completions in 0.383806 seconds
+[2025-12-17 21:04:15,255][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:04:15,579][openai._base_client][INFO] - Retrying request to /chat/completions in 0.468508 seconds
+[2025-12-17 21:04:21,057][openai._base_client][INFO] - Retrying request to /chat/completions in 0.753483 seconds
+[2025-12-17 21:04:47,305][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:04:47,580][openai._base_client][INFO] - Retrying request to /chat/completions in 0.492165 seconds
+[2025-12-17 21:05:04,597][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:05:04,925][openai._base_client][INFO] - Retrying request to /chat/completions in 0.475751 seconds
+[2025-12-17 21:05:10,412][openai._base_client][INFO] - Retrying request to /chat/completions in 0.891718 seconds
+[2025-12-17 21:05:35,671][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:05:36,059][openai._base_client][INFO] - Retrying request to /chat/completions in 0.469969 seconds
+[2025-12-17 21:05:53,641][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:05:53,990][openai._base_client][INFO] - Retrying request to /chat/completions in 0.379917 seconds
+[2025-12-17 21:06:10,497][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:06:10,788][openai._base_client][INFO] - Retrying request to /chat/completions in 0.429431 seconds
+[2025-12-17 21:06:24,119][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:06:24,474][openai._base_client][INFO] - Retrying request to /chat/completions in 0.420752 seconds
+[2025-12-17 21:06:39,910][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:06:40,239][openai._base_client][INFO] - Retrying request to /chat/completions in 0.458853 seconds
+[2025-12-17 21:06:57,152][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:06:57,494][openai._base_client][INFO] - Retrying request to /chat/completions in 0.403034 seconds
+[2025-12-17 21:07:12,508][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:07:12,805][openai._base_client][INFO] - Retrying request to /chat/completions in 0.482101 seconds
+[2025-12-17 21:07:29,632][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:07:29,952][openai._base_client][INFO] - Retrying request to /chat/completions in 0.410732 seconds
+[2025-12-17 21:07:43,022][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:07:43,284][openai._base_client][INFO] - Retrying request to /chat/completions in 0.382039 seconds
+[2025-12-17 21:08:00,686][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:08:00,979][openai._base_client][INFO] - Retrying request to /chat/completions in 0.377921 seconds
+[2025-12-17 21:08:16,565][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:08:16,962][openai._base_client][INFO] - Retrying request to /chat/completions in 0.422265 seconds
+[2025-12-17 21:08:34,129][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:08:34,490][openai._base_client][INFO] - Retrying request to /chat/completions in 0.444491 seconds
+[2025-12-17 21:08:53,056][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:08:53,466][openai._base_client][INFO] - Retrying request to /chat/completions in 0.497551 seconds
+[2025-12-17 21:09:10,951][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:09:11,297][openai._base_client][INFO] - Retrying request to /chat/completions in 0.394048 seconds
+[2025-12-17 21:09:11,797][openai._base_client][INFO] - Retrying request to /chat/completions in 0.755152 seconds
+[2025-12-17 21:09:29,847][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:09:30,182][openai._base_client][INFO] - Retrying request to /chat/completions in 0.386199 seconds
+[2025-12-17 21:09:47,530][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:09:47,863][openai._base_client][INFO] - Retrying request to /chat/completions in 0.396109 seconds
+[2025-12-17 21:09:48,273][openai._base_client][INFO] - Retrying request to /chat/completions in 0.961823 seconds
+[2025-12-17 21:10:04,666][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:10:04,998][openai._base_client][INFO] - Retrying request to /chat/completions in 0.451887 seconds
+[2025-12-17 21:10:22,287][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:10:22,606][openai._base_client][INFO] - Retrying request to /chat/completions in 0.424598 seconds
+[2025-12-17 21:10:38,280][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:10:38,616][openai._base_client][INFO] - Retrying request to /chat/completions in 0.415812 seconds
+[2025-12-17 21:10:56,309][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:10:56,662][openai._base_client][INFO] - Retrying request to /chat/completions in 0.482973 seconds
+[2025-12-17 21:11:15,521][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:11:15,774][openai._base_client][INFO] - Retrying request to /chat/completions in 0.393013 seconds
+[2025-12-17 21:11:29,849][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:11:30,198][openai._base_client][INFO] - Retrying request to /chat/completions in 0.442469 seconds
+[2025-12-17 21:11:47,747][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:11:48,129][openai._base_client][INFO] - Retrying request to /chat/completions in 0.395388 seconds
+[2025-12-17 21:12:03,245][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:12:03,595][openai._base_client][INFO] - Retrying request to /chat/completions in 0.432279 seconds
+[2025-12-17 21:12:23,676][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:12:24,064][openai._base_client][INFO] - Retrying request to /chat/completions in 0.448249 seconds
+[2025-12-17 21:12:40,888][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:12:41,384][openai._base_client][INFO] - Retrying request to /chat/completions in 0.430356 seconds
+[2025-12-17 21:12:58,396][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:12:58,747][openai._base_client][INFO] - Retrying request to /chat/completions in 0.463586 seconds
+[2025-12-17 21:13:19,522][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:13:19,924][openai._base_client][INFO] - Retrying request to /chat/completions in 0.439477 seconds
+[2025-12-17 21:13:20,664][openai._base_client][INFO] - Retrying request to /chat/completions in 0.947315 seconds
+[2025-12-17 21:13:26,717][openai._base_client][INFO] - Retrying request to /chat/completions in 0.437089 seconds
+[2025-12-17 21:13:43,318][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:13:43,694][openai._base_client][INFO] - Retrying request to /chat/completions in 0.375097 seconds
+[2025-12-17 21:13:57,772][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:13:58,133][openai._base_client][INFO] - Retrying request to /chat/completions in 0.467417 seconds
+[2025-12-17 21:13:58,679][openai._base_client][INFO] - Retrying request to /chat/completions in 0.756745 seconds
+[2025-12-17 21:14:05,589][openai._base_client][INFO] - Retrying request to /chat/completions in 0.444211 seconds
+[2025-12-17 21:14:22,794][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:14:23,148][openai._base_client][INFO] - Retrying request to /chat/completions in 0.377687 seconds
+[2025-12-17 21:14:23,557][openai._base_client][INFO] - Retrying request to /chat/completions in 0.892988 seconds
+[2025-12-17 21:14:39,245][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:14:39,697][openai._base_client][INFO] - Retrying request to /chat/completions in 0.416374 seconds
+[2025-12-17 21:14:57,350][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:14:57,765][openai._base_client][INFO] - Retrying request to /chat/completions in 0.459352 seconds
+[2025-12-17 21:15:16,489][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:15:16,909][openai._base_client][INFO] - Retrying request to /chat/completions in 0.446536 seconds
+[2025-12-17 21:15:17,371][openai._base_client][INFO] - Retrying request to /chat/completions in 0.841225 seconds
+[2025-12-17 21:15:31,470][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:15:31,956][openai._base_client][INFO] - Retrying request to /chat/completions in 0.461583 seconds
+[2025-12-17 21:15:54,247][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:15:54,630][openai._base_client][INFO] - Retrying request to /chat/completions in 0.480728 seconds
+[2025-12-17 21:16:13,654][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:16:14,153][openai._base_client][INFO] - Retrying request to /chat/completions in 0.435732 seconds
+[2025-12-17 21:16:14,605][openai._base_client][INFO] - Retrying request to /chat/completions in 0.751333 seconds
+[2025-12-17 21:16:31,926][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:16:32,327][openai._base_client][INFO] - Retrying request to /chat/completions in 0.467176 seconds
+[2025-12-17 21:16:46,760][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:16:47,230][openai._base_client][INFO] - Retrying request to /chat/completions in 0.379605 seconds
+[2025-12-17 21:17:05,378][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:17:05,786][openai._base_client][INFO] - Retrying request to /chat/completions in 0.429707 seconds
+[2025-12-17 21:17:23,201][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:17:23,584][openai._base_client][INFO] - Retrying request to /chat/completions in 0.375990 seconds
+[2025-12-17 21:17:43,849][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:17:44,253][openai._base_client][INFO] - Retrying request to /chat/completions in 0.460394 seconds
+[2025-12-17 21:18:03,075][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:18:03,494][openai._base_client][INFO] - Retrying request to /chat/completions in 0.469362 seconds
+[2025-12-17 21:18:26,717][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:18:27,114][openai._base_client][INFO] - Retrying request to /chat/completions in 0.404565 seconds
+[2025-12-17 21:18:32,528][openai._base_client][INFO] - Retrying request to /chat/completions in 0.985175 seconds
+[2025-12-17 21:18:50,837][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:18:51,293][openai._base_client][INFO] - Retrying request to /chat/completions in 0.439535 seconds
+[2025-12-17 21:19:08,065][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:19:08,479][openai._base_client][INFO] - Retrying request to /chat/completions in 0.431964 seconds
+[2025-12-17 21:19:25,378][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:19:25,912][openai._base_client][INFO] - Retrying request to /chat/completions in 0.478424 seconds
+[2025-12-17 21:19:43,584][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:19:44,055][openai._base_client][INFO] - Retrying request to /chat/completions in 0.417621 seconds
+[2025-12-17 21:20:04,688][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:20:05,173][openai._base_client][INFO] - Retrying request to /chat/completions in 0.460573 seconds
+[2025-12-17 21:20:24,378][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:20:24,872][openai._base_client][INFO] - Retrying request to /chat/completions in 0.478597 seconds
+[2025-12-17 21:20:42,973][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:20:43,376][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63735).
+[2025-12-17 21:20:43,634][openai._base_client][INFO] - Retrying request to /chat/completions in 0.484888 seconds
+[2025-12-17 21:21:06,397][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:21:06,751][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63806).
+[2025-12-17 21:21:07,056][openai._base_client][INFO] - Retrying request to /chat/completions in 0.461664 seconds
+[2025-12-17 21:21:31,491][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:21:31,923][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63834).
+[2025-12-17 21:21:32,196][openai._base_client][INFO] - Retrying request to /chat/completions in 0.403047 seconds
+[2025-12-17 21:21:37,610][openai._base_client][INFO] - Retrying request to /chat/completions in 0.899892 seconds
+[2025-12-17 21:22:02,262][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:22:02,560][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63904).
+[2025-12-17 21:22:02,851][openai._base_client][INFO] - Retrying request to /chat/completions in 0.463071 seconds
+[2025-12-17 21:22:26,554][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:22:26,931][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63930).
+[2025-12-17 21:22:27,230][openai._base_client][INFO] - Retrying request to /chat/completions in 0.423216 seconds
+[2025-12-17 21:22:32,665][openai._base_client][INFO] - Retrying request to /chat/completions in 0.752884 seconds
+[2025-12-17 21:23:09,457][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:23:09,851][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63997).
+[2025-12-17 21:23:10,131][openai._base_client][INFO] - Retrying request to /chat/completions in 0.480049 seconds
+[2025-12-17 21:23:15,622][openai._base_client][INFO] - Retrying request to /chat/completions in 0.940709 seconds
+[2025-12-17 21:23:31,646][openai._base_client][INFO] - Retrying request to /chat/completions in 0.483123 seconds
+[2025-12-17 21:23:37,142][openai._base_client][INFO] - Retrying request to /chat/completions in 0.815025 seconds
+[2025-12-17 21:23:53,044][openai._base_client][INFO] - Retrying request to /chat/completions in 0.433259 seconds
+[2025-12-17 21:24:17,022][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:24:17,518][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63705).
+[2025-12-17 21:24:17,780][openai._base_client][INFO] - Retrying request to /chat/completions in 0.494958 seconds
+[2025-12-17 21:24:41,217][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:24:41,555][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63730).
+[2025-12-17 21:24:41,789][openai._base_client][INFO] - Retrying request to /chat/completions in 0.494123 seconds
+[2025-12-17 21:24:47,293][openai._base_client][INFO] - Retrying request to /chat/completions in 0.784485 seconds
+[2025-12-17 21:25:10,534][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:25:10,815][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63769).
+[2025-12-17 21:25:11,088][openai._base_client][INFO] - Retrying request to /chat/completions in 0.436701 seconds
+[2025-12-17 21:25:36,107][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:25:36,444][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63808).
+[2025-12-17 21:25:36,659][openai._base_client][INFO] - Retrying request to /chat/completions in 0.417857 seconds
+[2025-12-17 21:25:59,967][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:26:00,299][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63849).
+[2025-12-17 21:26:00,598][openai._base_client][INFO] - Retrying request to /chat/completions in 0.415725 seconds
+[2025-12-17 21:26:23,762][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:26:24,057][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63925).
+[2025-12-17 21:26:24,285][openai._base_client][INFO] - Retrying request to /chat/completions in 0.428787 seconds
+[2025-12-17 21:26:29,724][openai._base_client][INFO] - Retrying request to /chat/completions in 0.945815 seconds
+[2025-12-17 21:27:05,255][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:27:05,578][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63978).
+[2025-12-17 21:27:05,869][openai._base_client][INFO] - Retrying request to /chat/completions in 0.414837 seconds
+[2025-12-17 21:27:11,295][openai._base_client][INFO] - Retrying request to /chat/completions in 0.887559 seconds
+[2025-12-17 21:27:37,417][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:27:37,847][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63671).
+[2025-12-17 21:27:38,115][openai._base_client][INFO] - Retrying request to /chat/completions in 0.393680 seconds
+[2025-12-17 21:28:04,870][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:28:05,174][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63708).
+[2025-12-17 21:28:05,487][openai._base_client][INFO] - Retrying request to /chat/completions in 0.381150 seconds
+[2025-12-17 21:28:10,879][openai._base_client][INFO] - Retrying request to /chat/completions in 0.781893 seconds
+[2025-12-17 21:28:34,765][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:28:35,092][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63767).
+[2025-12-17 21:28:35,394][openai._base_client][INFO] - Retrying request to /chat/completions in 0.441034 seconds
+[2025-12-17 21:28:57,035][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:28:57,280][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63784).
+[2025-12-17 21:28:57,552][openai._base_client][INFO] - Retrying request to /chat/completions in 0.375908 seconds
+[2025-12-17 21:29:19,882][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:29:20,134][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63832).
+[2025-12-17 21:29:20,422][openai._base_client][INFO] - Retrying request to /chat/completions in 0.467810 seconds
+[2025-12-17 21:29:47,600][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:29:47,875][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63895).
+[2025-12-17 21:29:48,144][openai._base_client][INFO] - Retrying request to /chat/completions in 0.433355 seconds
+[2025-12-17 21:30:12,099][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:30:12,294][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63944).
+[2025-12-17 21:30:12,464][openai._base_client][INFO] - Retrying request to /chat/completions in 0.486132 seconds
+[2025-12-17 21:30:17,960][openai._base_client][INFO] - Retrying request to /chat/completions in 0.803581 seconds
+[2025-12-17 21:30:44,323][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:30:44,766][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63621).
+[2025-12-17 21:30:45,026][openai._base_client][INFO] - Retrying request to /chat/completions in 0.375178 seconds
+[2025-12-17 21:30:50,412][openai._base_client][INFO] - Retrying request to /chat/completions in 0.877246 seconds
+[2025-12-17 21:31:14,026][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:31:14,313][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63658).
+[2025-12-17 21:31:14,600][openai._base_client][INFO] - Retrying request to /chat/completions in 0.395608 seconds
+[2025-12-17 21:31:38,863][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:31:39,102][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63700).
+[2025-12-17 21:31:39,378][openai._base_client][INFO] - Retrying request to /chat/completions in 0.394765 seconds
+[2025-12-17 21:32:02,842][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:32:03,108][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63718).
+[2025-12-17 21:32:03,367][openai._base_client][INFO] - Retrying request to /chat/completions in 0.440122 seconds
+[2025-12-17 21:32:08,818][openai._base_client][INFO] - Retrying request to /chat/completions in 0.976106 seconds
+[2025-12-17 21:32:34,002][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:32:34,336][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63766).
+[2025-12-17 21:32:34,578][openai._base_client][INFO] - Retrying request to /chat/completions in 0.407178 seconds
+[2025-12-17 21:32:39,996][openai._base_client][INFO] - Retrying request to /chat/completions in 0.910550 seconds
+[2025-12-17 21:33:02,730][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:33:02,993][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63743).
+[2025-12-17 21:33:03,253][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413284 seconds
+[2025-12-17 21:33:33,288][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:33:33,587][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63808).
+[2025-12-17 21:33:33,803][openai._base_client][INFO] - Retrying request to /chat/completions in 0.453051 seconds
+[2025-12-17 21:33:57,098][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:33:57,369][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63823).
+[2025-12-17 21:33:57,601][openai._base_client][INFO] - Retrying request to /chat/completions in 0.428246 seconds
+[2025-12-17 21:34:20,797][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:34:21,005][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63841).
+[2025-12-17 21:34:21,288][openai._base_client][INFO] - Retrying request to /chat/completions in 0.379805 seconds
+[2025-12-17 21:34:44,648][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:34:44,882][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63837).
+[2025-12-17 21:34:45,110][openai._base_client][INFO] - Retrying request to /chat/completions in 0.418437 seconds
+[2025-12-17 21:34:47,146][openai._base_client][INFO] - Retrying request to /chat/completions in 0.869629 seconds
+[2025-12-17 21:35:11,693][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:35:11,953][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63868).
+[2025-12-17 21:35:12,209][openai._base_client][INFO] - Retrying request to /chat/completions in 0.474489 seconds
+[2025-12-17 21:35:14,049][openai._base_client][INFO] - Retrying request to /chat/completions in 0.755838 seconds
+[2025-12-17 21:35:41,258][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:35:41,556][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63837).
+[2025-12-17 21:35:41,794][openai._base_client][INFO] - Retrying request to /chat/completions in 0.388767 seconds
+[2025-12-17 21:36:04,864][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:36:05,199][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63850).
+[2025-12-17 21:36:05,452][openai._base_client][INFO] - Retrying request to /chat/completions in 0.443235 seconds
+[2025-12-17 21:36:32,194][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:36:32,411][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63834).
+[2025-12-17 21:36:32,643][openai._base_client][INFO] - Retrying request to /chat/completions in 0.438105 seconds
+[2025-12-17 21:36:56,152][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:36:56,412][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63793).
+[2025-12-17 21:36:56,674][openai._base_client][INFO] - Retrying request to /chat/completions in 0.427387 seconds
+[2025-12-17 21:37:20,167][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:37:20,420][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63815).
+[2025-12-17 21:37:20,677][openai._base_client][INFO] - Retrying request to /chat/completions in 0.384808 seconds
+[2025-12-17 21:37:45,345][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:37:45,587][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63862).
+[2025-12-17 21:37:45,802][openai._base_client][INFO] - Retrying request to /chat/completions in 0.395270 seconds
+[2025-12-17 21:37:51,206][openai._base_client][INFO] - Retrying request to /chat/completions in 0.908562 seconds
+[2025-12-17 21:38:07,192][openai._base_client][INFO] - Retrying request to /chat/completions in 0.384472 seconds
+[2025-12-17 21:38:30,058][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:38:30,242][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63880).
+[2025-12-17 21:38:30,382][openai._base_client][INFO] - Retrying request to /chat/completions in 0.392893 seconds
+[2025-12-17 21:38:53,082][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:38:53,325][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63855).
+[2025-12-17 21:38:53,556][openai._base_client][INFO] - Retrying request to /chat/completions in 0.457561 seconds
+[2025-12-17 21:39:17,985][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:39:18,203][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63873).
+[2025-12-17 21:39:18,432][openai._base_client][INFO] - Retrying request to /chat/completions in 0.388298 seconds
+[2025-12-17 21:39:42,226][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:39:42,453][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63903).
+[2025-12-17 21:39:42,713][openai._base_client][INFO] - Retrying request to /chat/completions in 0.404799 seconds
+[2025-12-17 21:39:48,127][openai._base_client][INFO] - Retrying request to /chat/completions in 0.814009 seconds
+[2025-12-17 21:40:15,205][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:40:15,444][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63925).
+[2025-12-17 21:40:15,695][openai._base_client][INFO] - Retrying request to /chat/completions in 0.498713 seconds
+[2025-12-17 21:40:21,203][openai._base_client][INFO] - Retrying request to /chat/completions in 0.936308 seconds
+[2025-12-17 21:40:47,527][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:40:47,786][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63925).
+[2025-12-17 21:40:47,984][openai._base_client][INFO] - Retrying request to /chat/completions in 0.388824 seconds
+[2025-12-17 21:41:10,876][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:41:11,130][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63953).
+[2025-12-17 21:41:11,369][openai._base_client][INFO] - Retrying request to /chat/completions in 0.463830 seconds
+[2025-12-17 21:41:33,682][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:41:33,920][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63994).
+[2025-12-17 21:41:34,105][openai._base_client][INFO] - Retrying request to /chat/completions in 0.436674 seconds
+[2025-12-17 21:41:39,550][openai._base_client][INFO] - Retrying request to /chat/completions in 0.774222 seconds
+[2025-12-17 21:41:55,403][openai._base_client][INFO] - Retrying request to /chat/completions in 0.384192 seconds
+[2025-12-17 21:42:18,636][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:42:18,992][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63594).
+[2025-12-17 21:42:19,244][openai._base_client][INFO] - Retrying request to /chat/completions in 0.480574 seconds
+[2025-12-17 21:42:24,734][openai._base_client][INFO] - Retrying request to /chat/completions in 0.842342 seconds
+[2025-12-17 21:42:49,757][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:42:50,026][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63601).
+[2025-12-17 21:42:50,281][openai._base_client][INFO] - Retrying request to /chat/completions in 0.457240 seconds
+[2025-12-17 21:42:55,747][openai._base_client][INFO] - Retrying request to /chat/completions in 0.891600 seconds
+[2025-12-17 21:43:18,584][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:43:18,845][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63620).
+[2025-12-17 21:43:19,114][openai._base_client][INFO] - Retrying request to /chat/completions in 0.377966 seconds
+[2025-12-17 21:43:41,989][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:43:42,254][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63645).
+[2025-12-17 21:43:42,453][openai._base_client][INFO] - Retrying request to /chat/completions in 0.492798 seconds
+[2025-12-17 21:44:06,057][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:44:06,281][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63675).
+[2025-12-17 21:44:06,520][openai._base_client][INFO] - Retrying request to /chat/completions in 0.380775 seconds
+[2025-12-17 21:44:31,011][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:44:31,293][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63730).
+[2025-12-17 21:44:31,523][openai._base_client][INFO] - Retrying request to /chat/completions in 0.419322 seconds
+[2025-12-17 21:44:36,952][openai._base_client][INFO] - Retrying request to /chat/completions in 0.784631 seconds
+[2025-12-17 21:45:09,944][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:45:10,167][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63780).
+[2025-12-17 21:45:10,384][openai._base_client][INFO] - Retrying request to /chat/completions in 0.397136 seconds
+[2025-12-17 21:45:33,547][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:45:33,748][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63818).
+[2025-12-17 21:45:33,978][openai._base_client][INFO] - Retrying request to /chat/completions in 0.406804 seconds
+[2025-12-17 21:45:39,394][openai._base_client][INFO] - Retrying request to /chat/completions in 0.998786 seconds
+[2025-12-17 21:45:55,459][openai._base_client][INFO] - Retrying request to /chat/completions in 0.462323 seconds
+[2025-12-17 21:46:00,931][openai._base_client][INFO] - Retrying request to /chat/completions in 0.750444 seconds
+[2025-12-17 21:46:16,766][openai._base_client][INFO] - Retrying request to /chat/completions in 0.493543 seconds
+[2025-12-17 21:46:22,269][openai._base_client][INFO] - Retrying request to /chat/completions in 0.877412 seconds
+[2025-12-17 21:46:38,218][openai._base_client][INFO] - Retrying request to /chat/completions in 0.470898 seconds
+[2025-12-17 21:47:02,428][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:47:02,658][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63847).
+[2025-12-17 21:47:02,845][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413937 seconds
+[2025-12-17 21:47:08,269][openai._base_client][INFO] - Retrying request to /chat/completions in 0.829970 seconds
+[2025-12-17 21:47:34,938][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:47:35,147][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63873).
+[2025-12-17 21:47:35,365][openai._base_client][INFO] - Retrying request to /chat/completions in 0.492797 seconds
+[2025-12-17 21:47:59,096][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:47:59,377][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63854).
+[2025-12-17 21:47:59,623][openai._base_client][INFO] - Retrying request to /chat/completions in 0.430073 seconds
+[2025-12-17 21:48:05,062][openai._base_client][INFO] - Retrying request to /chat/completions in 0.927004 seconds
+[2025-12-17 21:48:40,644][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:48:40,899][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63853).
+[2025-12-17 21:48:41,129][openai._base_client][INFO] - Retrying request to /chat/completions in 0.431160 seconds
+[2025-12-17 21:48:46,570][openai._base_client][INFO] - Retrying request to /chat/completions in 0.938753 seconds
+[2025-12-17 21:49:14,875][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:49:15,116][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63885).
+[2025-12-17 21:49:15,322][openai._base_client][INFO] - Retrying request to /chat/completions in 0.436692 seconds
+[2025-12-17 21:49:41,057][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:49:41,316][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63887).
+[2025-12-17 21:49:41,558][openai._base_client][INFO] - Retrying request to /chat/completions in 0.461283 seconds
+[2025-12-17 21:49:47,029][openai._base_client][INFO] - Retrying request to /chat/completions in 0.907892 seconds
+[2025-12-17 21:50:22,886][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:50:23,112][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63900).
+[2025-12-17 21:50:23,355][openai._base_client][INFO] - Retrying request to /chat/completions in 0.392431 seconds
+[2025-12-17 21:50:28,756][openai._base_client][INFO] - Retrying request to /chat/completions in 0.976465 seconds
+[2025-12-17 21:50:54,039][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:50:54,312][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63913).
+[2025-12-17 21:50:54,522][openai._base_client][INFO] - Retrying request to /chat/completions in 0.458333 seconds
+[2025-12-17 21:50:59,990][openai._base_client][INFO] - Retrying request to /chat/completions in 0.962774 seconds
+[2025-12-17 21:51:16,021][openai._base_client][INFO] - Retrying request to /chat/completions in 0.467453 seconds
+[2025-12-17 21:51:39,311][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:51:39,540][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63928).
+[2025-12-17 21:51:39,742][openai._base_client][INFO] - Retrying request to /chat/completions in 0.482920 seconds
+[2025-12-17 21:51:45,233][openai._base_client][INFO] - Retrying request to /chat/completions in 0.953730 seconds
+[2025-12-17 21:52:13,501][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:52:13,753][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63935).
+[2025-12-17 21:52:13,943][openai._base_client][INFO] - Retrying request to /chat/completions in 0.498707 seconds
+[2025-12-17 21:52:35,447][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:52:35,679][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63967).
+[2025-12-17 21:52:35,893][openai._base_client][INFO] - Retrying request to /chat/completions in 0.449679 seconds
+[2025-12-17 21:52:36,780][openai._base_client][INFO] - Retrying request to /chat/completions in 0.848735 seconds
+[2025-12-17 21:52:48,091][openai._base_client][INFO] - Retrying request to /chat/completions in 0.428443 seconds
+[2025-12-17 21:53:10,948][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:53:11,267][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63981).
+[2025-12-17 21:53:11,487][openai._base_client][INFO] - Retrying request to /chat/completions in 0.393188 seconds
+[2025-12-17 21:53:12,274][openai._base_client][INFO] - Retrying request to /chat/completions in 0.870777 seconds
+[2025-12-17 21:53:50,093][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:53:50,339][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63984).
+[2025-12-17 21:53:50,543][openai._base_client][INFO] - Retrying request to /chat/completions in 0.383355 seconds
+[2025-12-17 21:54:14,328][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:54:14,504][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 64000).
+[2025-12-17 21:54:14,696][openai._base_client][INFO] - Retrying request to /chat/completions in 0.497212 seconds
+[2025-12-17 21:54:38,117][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:54:38,340][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63983).
+[2025-12-17 21:54:38,520][openai._base_client][INFO] - Retrying request to /chat/completions in 0.378823 seconds
+[2025-12-17 21:55:01,816][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:55:02,029][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63981).
+[2025-12-17 21:55:02,211][openai._base_client][INFO] - Retrying request to /chat/completions in 0.436061 seconds
+[2025-12-17 21:55:28,087][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:55:28,357][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63969).
+[2025-12-17 21:55:28,568][openai._base_client][INFO] - Retrying request to /chat/completions in 0.394212 seconds
+[2025-12-17 21:55:51,874][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:55:52,091][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63952).
+[2025-12-17 21:55:52,313][openai._base_client][INFO] - Retrying request to /chat/completions in 0.400072 seconds
+[2025-12-17 21:56:14,942][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:56:15,190][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63979).
+[2025-12-17 21:56:15,411][openai._base_client][INFO] - Retrying request to /chat/completions in 0.486208 seconds
+[2025-12-17 21:56:20,907][openai._base_client][INFO] - Retrying request to /chat/completions in 0.835961 seconds
+[2025-12-17 21:56:55,068][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:56:55,455][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63621).
+[2025-12-17 21:56:55,634][openai._base_client][INFO] - Retrying request to /chat/completions in 0.408000 seconds
+[2025-12-17 21:57:17,911][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:57:18,168][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63629).
+[2025-12-17 21:57:18,377][openai._base_client][INFO] - Retrying request to /chat/completions in 0.383857 seconds
+[2025-12-17 21:57:42,132][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:57:42,381][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63660).
+[2025-12-17 21:57:42,574][openai._base_client][INFO] - Retrying request to /chat/completions in 0.495062 seconds
+[2025-12-17 21:57:48,079][openai._base_client][INFO] - Retrying request to /chat/completions in 0.769231 seconds
+[2025-12-17 21:58:03,927][openai._base_client][INFO] - Retrying request to /chat/completions in 0.378371 seconds
+[2025-12-17 21:58:09,317][openai._base_client][INFO] - Retrying request to /chat/completions in 0.837954 seconds
+[2025-12-17 21:58:45,409][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:58:45,717][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63687).
+[2025-12-17 21:58:45,956][openai._base_client][INFO] - Retrying request to /chat/completions in 0.422383 seconds
+[2025-12-17 21:58:51,387][openai._base_client][INFO] - Retrying request to /chat/completions in 0.790698 seconds
+[2025-12-17 21:59:24,069][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:59:24,345][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63677).
+[2025-12-17 21:59:24,571][openai._base_client][INFO] - Retrying request to /chat/completions in 0.480772 seconds
+[2025-12-17 21:59:47,603][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 21:59:47,786][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63692).
+[2025-12-17 21:59:47,981][openai._base_client][INFO] - Retrying request to /chat/completions in 0.453736 seconds
+[2025-12-17 22:00:16,087][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:00:16,278][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63683).
+[2025-12-17 22:00:16,493][openai._base_client][INFO] - Retrying request to /chat/completions in 0.419782 seconds
+[2025-12-17 22:00:21,920][openai._base_client][INFO] - Retrying request to /chat/completions in 0.902689 seconds
+[2025-12-17 22:00:43,816][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:00:44,075][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63690).
+[2025-12-17 22:00:44,264][openai._base_client][INFO] - Retrying request to /chat/completions in 0.435081 seconds
+[2025-12-17 22:01:08,006][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:01:08,246][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63700).
+[2025-12-17 22:01:08,450][openai._base_client][INFO] - Retrying request to /chat/completions in 0.435999 seconds
+[2025-12-17 22:01:35,991][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:01:36,213][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63700).
+[2025-12-17 22:01:36,431][openai._base_client][INFO] - Retrying request to /chat/completions in 0.469198 seconds
+[2025-12-17 22:02:02,773][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:02:02,990][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63730).
+[2025-12-17 22:02:03,223][openai._base_client][INFO] - Retrying request to /chat/completions in 0.392258 seconds
+[2025-12-17 22:02:08,629][openai._base_client][INFO] - Retrying request to /chat/completions in 0.780852 seconds
+[2025-12-17 22:02:35,710][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:02:35,987][root][WARNING] - Truncated history for env 0 to keep prompt within 64000 tokens (current 63737).
+[2025-12-17 22:02:36,212][openai._base_client][INFO] - Retrying request to /chat/completions in 0.471201 seconds
+[2025-12-17 22:02:41,695][openai._base_client][INFO] - Retrying request to /chat/completions in 0.761189 seconds

outputs/2025-12-17/22-05-23/.hydra/config.yaml

@@ -0,0 +1,903 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 3600
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: null
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 25
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 128
+    group_size: 1
+    env_configs:
+      tags:
+      - CoordFrozenLake
+      n_groups:
+      - 128
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: ark-deepseek-v3-250324
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  ark-deepseek-v3-250324:
+    provider_name: openai
+    model_name: ark-deepseek-v3-250324
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/22-05-23/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/22-05-23
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/22-05-23/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/22-05-23/eval_api.log

@@ -0,0 +1,439 @@
+[2025-12-17 22:05:31,725][openai._base_client][INFO] - Retrying request to /chat/completions in 0.491558 seconds
+[2025-12-17 22:05:31,727][openai._base_client][INFO] - Retrying request to /chat/completions in 0.407830 seconds
+[2025-12-17 22:05:31,727][openai._base_client][INFO] - Retrying request to /chat/completions in 0.458426 seconds
+[2025-12-17 22:05:31,727][openai._base_client][INFO] - Retrying request to /chat/completions in 0.433003 seconds
+[2025-12-17 22:05:31,728][openai._base_client][INFO] - Retrying request to /chat/completions in 0.462197 seconds
+[2025-12-17 22:05:32,259][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:32,474][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:32,581][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:33,207][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:33,468][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:33,496][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:33,550][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:33,608][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:34,501][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:34,678][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:34,840][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:37,140][openai._base_client][INFO] - Retrying request to /chat/completions in 0.789049 seconds
+[2025-12-17 22:05:37,164][openai._base_client][INFO] - Retrying request to /chat/completions in 0.921024 seconds
+[2025-12-17 22:05:37,189][openai._base_client][INFO] - Retrying request to /chat/completions in 0.832440 seconds
+[2025-12-17 22:05:37,199][openai._base_client][INFO] - Retrying request to /chat/completions in 0.974576 seconds
+[2025-12-17 22:05:37,222][openai._base_client][INFO] - Retrying request to /chat/completions in 0.750014 seconds
+[2025-12-17 22:05:38,288][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:38,687][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:38,776][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:38,784][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:39,014][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:39,878][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:39,951][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:40,134][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:40,530][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:43,015][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:43,729][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:43,791][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:44,406][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:44,488][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:44,841][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:44,958][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:45,850][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:46,126][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:46,185][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:47,397][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:47,939][openai._base_client][INFO] - Retrying request to /chat/completions in 0.434967 seconds
+[2025-12-17 22:05:47,981][openai._base_client][INFO] - Retrying request to /chat/completions in 0.491912 seconds
+[2025-12-17 22:05:48,037][openai._base_client][INFO] - Retrying request to /chat/completions in 0.483886 seconds
+[2025-12-17 22:05:48,091][openai._base_client][INFO] - Retrying request to /chat/completions in 0.455813 seconds
+[2025-12-17 22:05:48,181][openai._base_client][INFO] - Retrying request to /chat/completions in 0.421189 seconds
+[2025-12-17 22:05:48,316][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:48,535][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:50,264][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:50,501][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:50,820][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:50,949][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:51,412][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:51,757][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:52,224][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:52,749][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:53,252][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:53,255][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:53,379][openai._base_client][INFO] - Retrying request to /chat/completions in 0.833622 seconds
+[2025-12-17 22:05:53,476][openai._base_client][INFO] - Retrying request to /chat/completions in 0.994456 seconds
+[2025-12-17 22:05:53,524][openai._base_client][INFO] - Retrying request to /chat/completions in 0.827713 seconds
+[2025-12-17 22:05:53,553][openai._base_client][INFO] - Retrying request to /chat/completions in 0.946853 seconds
+[2025-12-17 22:05:53,606][openai._base_client][INFO] - Retrying request to /chat/completions in 0.807286 seconds
+[2025-12-17 22:05:53,628][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:55,341][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:56,682][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:56,904][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:57,440][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:57,546][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:58,249][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:58,697][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:58,717][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:59,595][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:05:59,926][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:00,089][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:00,591][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:00,684][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:00,918][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:00,993][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:01,190][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:01,825][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:02,422][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:03,363][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:03,833][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:04,443][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:04,591][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:04,998][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:05,086][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:05,718][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:05,918][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:06,097][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:06,259][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:06,501][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:06,892][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:06,933][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:07,280][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:08,218][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:08,554][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:08,970][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:10,031][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:10,445][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:10,791][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:10,827][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:10,863][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:11,422][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:11,782][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:11,891][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:12,188][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:12,860][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:13,078][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:13,181][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:13,496][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:13,954][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:14,056][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:14,746][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:14,882][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:15,563][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:16,825][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:16,947][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:16,978][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:17,506][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:17,529][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:17,628][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:18,326][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:18,630][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:18,831][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:18,852][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:19,012][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:19,464][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:20,306][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:20,688][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:21,452][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:22,109][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:22,196][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:22,370][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:22,802][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:22,979][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:23,227][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:24,391][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:24,640][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:25,032][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:26,545][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:31,556][openai._base_client][INFO] - Retrying request to /chat/completions in 0.425657 seconds
+[2025-12-17 22:06:36,570][openai._base_client][INFO] - Retrying request to /chat/completions in 0.486874 seconds
+[2025-12-17 22:06:36,572][openai._base_client][INFO] - Retrying request to /chat/completions in 0.386743 seconds
+[2025-12-17 22:06:37,131][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:37,881][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:38,303][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:38,515][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:42,385][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:43,958][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:44,223][openai._base_client][INFO] - Retrying request to /chat/completions in 0.381577 seconds
+[2025-12-17 22:06:44,226][openai._base_client][INFO] - Retrying request to /chat/completions in 0.440298 seconds
+[2025-12-17 22:06:48,967][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:49,227][openai._base_client][INFO] - Retrying request to /chat/completions in 0.471569 seconds
+[2025-12-17 22:06:49,236][openai._base_client][INFO] - Retrying request to /chat/completions in 0.444679 seconds
+[2025-12-17 22:06:49,250][openai._base_client][INFO] - Retrying request to /chat/completions in 0.433829 seconds
+[2025-12-17 22:06:49,258][openai._base_client][INFO] - Retrying request to /chat/completions in 0.497646 seconds
+[2025-12-17 22:06:49,259][openai._base_client][INFO] - Retrying request to /chat/completions in 0.405379 seconds
+[2025-12-17 22:06:49,260][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:49,479][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:49,710][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:50,211][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:50,242][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:51,072][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:51,199][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:51,464][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:52,096][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:52,317][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:52,926][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:54,175][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:54,203][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:54,233][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:54,304][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:54,361][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:54,725][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:54,977][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:55,388][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:55,699][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:55,722][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:55,768][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:55,951][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:56,353][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:57,554][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:58,248][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:58,364][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:58,367][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:58,480][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:59,174][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:59,332][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:59,638][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:06:59,893][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:00,111][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:00,168][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:00,427][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:00,555][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:01,223][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:01,410][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:02,024][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:02,529][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:02,753][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:03,219][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:03,290][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:03,487][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:03,619][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:04,150][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:04,584][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:04,954][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:05,253][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:05,618][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:06,072][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:06,094][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:06,253][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:06,398][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:06,468][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:07,495][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:07,714][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:07,818][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:08,577][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:08,774][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:09,055][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:09,529][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:09,791][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:09,916][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:10,491][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:10,835][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:11,686][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:11,816][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:11,924][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:12,078][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:12,297][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:12,649][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:12,771][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:12,782][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:13,078][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:13,261][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:13,302][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:13,762][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:14,005][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:14,101][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:15,142][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:15,603][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:15,722][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:15,751][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:16,138][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:16,187][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:16,939][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:17,284][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:17,744][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:17,754][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:17,935][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:18,763][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:19,309][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:19,620][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:20,203][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:20,446][openai._base_client][INFO] - Retrying request to /chat/completions in 0.421019 seconds
+[2025-12-17 22:07:20,449][openai._base_client][INFO] - Retrying request to /chat/completions in 0.430619 seconds
+[2025-12-17 22:07:20,452][openai._base_client][INFO] - Retrying request to /chat/completions in 0.449461 seconds
+[2025-12-17 22:07:20,455][openai._base_client][INFO] - Retrying request to /chat/completions in 0.437412 seconds
+[2025-12-17 22:07:20,456][openai._base_client][INFO] - Retrying request to /chat/completions in 0.465990 seconds
+[2025-12-17 22:07:20,458][openai._base_client][INFO] - Retrying request to /chat/completions in 0.381520 seconds
+[2025-12-17 22:07:20,459][openai._base_client][INFO] - Retrying request to /chat/completions in 0.479633 seconds
+[2025-12-17 22:07:20,460][openai._base_client][INFO] - Retrying request to /chat/completions in 0.436579 seconds
+[2025-12-17 22:07:20,461][openai._base_client][INFO] - Retrying request to /chat/completions in 0.460265 seconds
+[2025-12-17 22:07:20,462][openai._base_client][INFO] - Retrying request to /chat/completions in 0.491518 seconds
+[2025-12-17 22:07:20,463][openai._base_client][INFO] - Retrying request to /chat/completions in 0.397991 seconds
+[2025-12-17 22:07:20,463][openai._base_client][INFO] - Retrying request to /chat/completions in 0.448540 seconds
+[2025-12-17 22:07:20,464][openai._base_client][INFO] - Retrying request to /chat/completions in 0.422551 seconds
+[2025-12-17 22:07:20,465][openai._base_client][INFO] - Retrying request to /chat/completions in 0.480339 seconds
+[2025-12-17 22:07:25,278][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:25,454][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:25,465][openai._base_client][INFO] - Retrying request to /chat/completions in 0.430430 seconds
+[2025-12-17 22:07:25,476][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:25,690][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:25,823][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:25,941][openai._base_client][INFO] - Retrying request to /chat/completions in 0.960822 seconds
+[2025-12-17 22:07:26,384][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:26,456][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:26,705][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:26,777][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:27,008][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:27,018][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:27,426][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:28,938][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:29,363][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:29,518][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:29,721][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:30,048][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:30,098][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:30,167][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:30,647][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:31,258][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:31,620][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:31,683][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:32,096][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:32,139][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:32,802][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:33,090][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:33,514][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:34,281][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:34,731][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:34,770][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:34,842][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:35,600][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:35,603][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:35,669][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:35,681][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:35,889][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:36,171][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:36,687][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:37,072][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:37,449][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:37,464][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:37,670][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:37,967][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:38,184][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:38,850][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:38,949][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:39,529][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:39,562][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:39,649][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:39,778][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:40,092][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:40,388][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:40,515][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:40,666][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:41,432][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:41,692][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:42,053][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:42,565][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:43,003][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:43,141][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:43,224][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:43,315][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:43,609][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:43,987][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:44,381][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:44,820][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:45,079][openai._base_client][INFO] - Retrying request to /chat/completions in 0.460560 seconds
+[2025-12-17 22:07:45,081][openai._base_client][INFO] - Retrying request to /chat/completions in 0.424450 seconds
+[2025-12-17 22:07:45,084][openai._base_client][INFO] - Retrying request to /chat/completions in 0.390722 seconds
+[2025-12-17 22:07:45,086][openai._base_client][INFO] - Retrying request to /chat/completions in 0.405809 seconds
+[2025-12-17 22:07:45,088][openai._base_client][INFO] - Retrying request to /chat/completions in 0.439058 seconds
+[2025-12-17 22:07:45,090][openai._base_client][INFO] - Retrying request to /chat/completions in 0.473577 seconds
+[2025-12-17 22:07:45,092][openai._base_client][INFO] - Retrying request to /chat/completions in 0.425851 seconds
+[2025-12-17 22:07:45,093][openai._base_client][INFO] - Retrying request to /chat/completions in 0.486488 seconds
+[2025-12-17 22:07:45,094][openai._base_client][INFO] - Retrying request to /chat/completions in 0.455646 seconds
+[2025-12-17 22:07:45,095][openai._base_client][INFO] - Retrying request to /chat/completions in 0.481686 seconds
+[2025-12-17 22:07:45,097][openai._base_client][INFO] - Retrying request to /chat/completions in 0.450651 seconds
+[2025-12-17 22:07:45,098][openai._base_client][INFO] - Retrying request to /chat/completions in 0.408203 seconds
+[2025-12-17 22:07:45,099][openai._base_client][INFO] - Retrying request to /chat/completions in 0.442139 seconds
+[2025-12-17 22:07:45,100][openai._base_client][INFO] - Retrying request to /chat/completions in 0.480236 seconds
+[2025-12-17 22:07:45,100][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413266 seconds
+[2025-12-17 22:07:45,101][openai._base_client][INFO] - Retrying request to /chat/completions in 0.457115 seconds
+[2025-12-17 22:07:49,103][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:49,541][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:49,708][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:49,977][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:50,261][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:50,437][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:50,510][openai._base_client][INFO] - Retrying request to /chat/completions in 0.781889 seconds
+[2025-12-17 22:07:50,694][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:50,739][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:50,939][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:51,021][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:51,073][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:51,343][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:51,353][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:51,505][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:52,301][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:53,194][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:53,223][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:53,429][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:53,549][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:54,781][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:54,984][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:55,018][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:55,181][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:55,220][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:55,298][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:55,387][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:55,450][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:56,083][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:56,086][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:56,766][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:56,831][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:57,199][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:57,508][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:58,083][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:58,732][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:58,998][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:59,723][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:59,833][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:07:59,963][openai._base_client][INFO] - Retrying request to /chat/completions in 0.378121 seconds
+[2025-12-17 22:07:59,966][openai._base_client][INFO] - Retrying request to /chat/completions in 0.457314 seconds
+[2025-12-17 22:07:59,970][openai._base_client][INFO] - Retrying request to /chat/completions in 0.388884 seconds
+[2025-12-17 22:07:59,973][openai._base_client][INFO] - Retrying request to /chat/completions in 0.391069 seconds
+[2025-12-17 22:07:59,976][openai._base_client][INFO] - Retrying request to /chat/completions in 0.437831 seconds
+[2025-12-17 22:07:59,978][openai._base_client][INFO] - Retrying request to /chat/completions in 0.499115 seconds
+[2025-12-17 22:07:59,980][openai._base_client][INFO] - Retrying request to /chat/completions in 0.496874 seconds
+[2025-12-17 22:07:59,982][openai._base_client][INFO] - Retrying request to /chat/completions in 0.490802 seconds
+[2025-12-17 22:07:59,985][openai._base_client][INFO] - Retrying request to /chat/completions in 0.393528 seconds
+[2025-12-17 22:07:59,987][openai._base_client][INFO] - Retrying request to /chat/completions in 0.442447 seconds
+[2025-12-17 22:07:59,989][openai._base_client][INFO] - Retrying request to /chat/completions in 0.470796 seconds
+[2025-12-17 22:07:59,990][openai._base_client][INFO] - Retrying request to /chat/completions in 0.469399 seconds
+[2025-12-17 22:07:59,991][openai._base_client][INFO] - Retrying request to /chat/completions in 0.456937 seconds
+[2025-12-17 22:07:59,992][openai._base_client][INFO] - Retrying request to /chat/completions in 0.468436 seconds
+[2025-12-17 22:07:59,992][openai._base_client][INFO] - Retrying request to /chat/completions in 0.414766 seconds
+[2025-12-17 22:07:59,993][openai._base_client][INFO] - Retrying request to /chat/completions in 0.484587 seconds
+[2025-12-17 22:08:04,349][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:04,471][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:04,560][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:04,945][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:05,193][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:05,351][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:05,366][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:05,451][openai._base_client][INFO] - Retrying request to /chat/completions in 0.874872 seconds
+[2025-12-17 22:08:05,463][openai._base_client][INFO] - Retrying request to /chat/completions in 0.751168 seconds
+[2025-12-17 22:08:05,464][openai._base_client][INFO] - Retrying request to /chat/completions in 0.999065 seconds
+[2025-12-17 22:08:05,648][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:05,799][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:05,904][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:06,215][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:06,255][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:06,714][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:08,274][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:10,177][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:10,496][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:11,192][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:11,224][openai._base_client][INFO] - Retrying request to /chat/completions in 0.417513 seconds
+[2025-12-17 22:08:11,229][openai._base_client][INFO] - Retrying request to /chat/completions in 0.428893 seconds
+[2025-12-17 22:08:11,230][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413421 seconds
+[2025-12-17 22:08:11,646][openai._base_client][INFO] - Retrying request to /chat/completions in 0.785642 seconds
+[2025-12-17 22:08:11,647][openai._base_client][INFO] - Retrying request to /chat/completions in 0.946303 seconds
+[2025-12-17 22:08:11,660][openai._base_client][INFO] - Retrying request to /chat/completions in 0.897845 seconds
+[2025-12-17 22:08:18,077][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:23,089][openai._base_client][INFO] - Retrying request to /chat/completions in 0.457206 seconds
+[2025-12-17 22:08:27,709][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:08:28,847][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"

outputs/2025-12-17/22-10-17/.hydra/config.yaml

@@ -0,0 +1,903 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 3600
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: null
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 25
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 128
+    group_size: 1
+    env_configs:
+      tags:
+      - CoordFrozenLake
+      n_groups:
+      - 128
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: ark-deepseek-v3-250324
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  ark-deepseek-v3-250324:
+    provider_name: openai
+    model_name: ark-deepseek-v3-250324
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/22-10-17/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/22-10-17
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/22-10-17/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/22-10-17/eval_api.log

@@ -0,0 +1,710 @@
+[2025-12-17 22:10:18,662][huggingface_hub.utils._http][WARNING] - '(MaxRetryError('HTTPSConnectionPool(host=\'huggingface.co\', port=443): Max retries exceeded with url: /Qwen/Qwen2.5-0.5B-Instruct/resolve/main/tokenizer_config.json (Caused by SSLError(SSLCertVerificationError(1, "[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: Hostname mismatch, certificate is not valid for \'huggingface.co\'. (_ssl.c:1017)")))'), '(Request ID: dfbbee7e-875b-497b-a947-fc2a57f8c6d8)')' thrown while requesting HEAD https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct/resolve/main/tokenizer_config.json
+[2025-12-17 22:10:18,663][huggingface_hub.utils._http][WARNING] - Retrying in 1s [Retry 1/5].
+[2025-12-17 22:10:27,186][openai._base_client][INFO] - Retrying request to /chat/completions in 0.475078 seconds
+[2025-12-17 22:10:27,188][openai._base_client][INFO] - Retrying request to /chat/completions in 0.375608 seconds
+[2025-12-17 22:10:27,191][openai._base_client][INFO] - Retrying request to /chat/completions in 0.392040 seconds
+[2025-12-17 22:10:27,202][openai._base_client][INFO] - Retrying request to /chat/completions in 0.397899 seconds
+[2025-12-17 22:10:28,581][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:28,852][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:29,155][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:29,422][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:29,928][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:30,031][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:30,054][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:30,204][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:31,028][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:31,430][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:31,740][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:32,099][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:32,567][openai._base_client][INFO] - Retrying request to /chat/completions in 0.922617 seconds
+[2025-12-17 22:10:32,605][openai._base_client][INFO] - Retrying request to /chat/completions in 0.837427 seconds
+[2025-12-17 22:10:32,666][openai._base_client][INFO] - Retrying request to /chat/completions in 0.810971 seconds
+[2025-12-17 22:10:34,364][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:34,747][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:35,038][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:35,606][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:36,010][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:36,076][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:36,882][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:37,009][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:37,063][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:37,108][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:37,986][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:38,283][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:39,119][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:39,410][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:40,196][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:40,778][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:40,889][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:41,037][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:41,728][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:41,839][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:43,067][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:43,177][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:43,455][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:43,483][openai._base_client][INFO] - Retrying request to /chat/completions in 0.453279 seconds
+[2025-12-17 22:10:43,839][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:44,020][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:44,794][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:44,972][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:45,059][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:45,536][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:46,034][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:46,065][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:46,271][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:46,929][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:48,258][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:48,492][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:48,939][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:48,943][openai._base_client][INFO] - Retrying request to /chat/completions in 0.868502 seconds
+[2025-12-17 22:10:49,114][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:49,272][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:50,032][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:50,215][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:50,342][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:50,377][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:51,030][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:51,653][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:52,836][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:53,376][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:53,636][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:53,977][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:54,845][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:55,326][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:55,542][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:55,731][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:55,780][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:55,791][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:56,149][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:56,273][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:56,732][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:57,265][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:58,291][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:59,374][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:10:59,607][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:00,432][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:00,453][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:00,631][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:01,264][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:01,360][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:01,441][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:01,763][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:02,190][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:02,397][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:03,120][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:03,152][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:03,767][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:04,232][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:05,243][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:05,370][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:06,035][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:06,040][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:06,211][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:06,772][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:06,929][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:07,776][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:08,172][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:08,664][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:09,188][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:09,404][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:09,726][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:09,975][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:10,329][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:10,681][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:11,009][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:11,492][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:11,966][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:13,108][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:13,396][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:13,726][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:13,914][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:14,142][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:14,195][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:14,741][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:15,110][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:15,461][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:15,879][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:15,916][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:16,465][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:16,598][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:16,818][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:16,917][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:18,105][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:18,612][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:19,218][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:19,628][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:20,000][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:25,011][openai._base_client][INFO] - Retrying request to /chat/completions in 0.420404 seconds
+[2025-12-17 22:11:30,020][openai._base_client][INFO] - Retrying request to /chat/completions in 0.477561 seconds
+[2025-12-17 22:11:30,021][openai._base_client][INFO] - Retrying request to /chat/completions in 0.491321 seconds
+[2025-12-17 22:11:30,436][openai._base_client][INFO] - Retrying request to /chat/completions in 0.760238 seconds
+[2025-12-17 22:11:35,364][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:38,596][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:11:43,606][openai._base_client][INFO] - Retrying request to /chat/completions in 0.489904 seconds
+[2025-12-17 22:11:49,113][openai._base_client][INFO] - Retrying request to /chat/completions in 0.888540 seconds
+[2025-12-17 22:12:06,365][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:06,643][openai._base_client][INFO] - Retrying request to /chat/completions in 0.499044 seconds
+[2025-12-17 22:12:11,343][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:11,647][openai._base_client][INFO] - Retrying request to /chat/completions in 0.398364 seconds
+[2025-12-17 22:12:11,648][openai._base_client][INFO] - Retrying request to /chat/completions in 0.398762 seconds
+[2025-12-17 22:12:11,649][openai._base_client][INFO] - Retrying request to /chat/completions in 0.398745 seconds
+[2025-12-17 22:12:11,650][openai._base_client][INFO] - Retrying request to /chat/completions in 0.468907 seconds
+[2025-12-17 22:12:11,650][openai._base_client][INFO] - Retrying request to /chat/completions in 0.439208 seconds
+[2025-12-17 22:12:11,650][openai._base_client][INFO] - Retrying request to /chat/completions in 0.450010 seconds
+[2025-12-17 22:12:11,651][openai._base_client][INFO] - Retrying request to /chat/completions in 0.378317 seconds
+[2025-12-17 22:12:11,736][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:12,049][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:12,145][openai._base_client][INFO] - Retrying request to /chat/completions in 0.930230 seconds
+[2025-12-17 22:12:12,170][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:12,189][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:13,002][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:13,459][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:15,707][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:16,737][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:17,060][openai._base_client][INFO] - Retrying request to /chat/completions in 0.763625 seconds
+[2025-12-17 22:12:17,060][openai._base_client][INFO] - Retrying request to /chat/completions in 0.818201 seconds
+[2025-12-17 22:12:17,093][openai._base_client][INFO] - Retrying request to /chat/completions in 0.957747 seconds
+[2025-12-17 22:12:17,103][openai._base_client][INFO] - Retrying request to /chat/completions in 0.894779 seconds
+[2025-12-17 22:12:17,122][openai._base_client][INFO] - Retrying request to /chat/completions in 0.832084 seconds
+[2025-12-17 22:12:17,409][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:18,079][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:18,506][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:18,586][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:18,660][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:18,821][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:18,841][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:19,777][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:22,047][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:22,653][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:22,793][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:22,808][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:22,840][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:23,085][openai._base_client][INFO] - Retrying request to /chat/completions in 0.488046 seconds
+[2025-12-17 22:12:23,354][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:23,397][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:23,920][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:24,587][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:25,009][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:26,303][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:27,045][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:27,101][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:27,109][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:27,232][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:27,348][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:27,831][openai._base_client][INFO] - Retrying request to /chat/completions in 0.421525 seconds
+[2025-12-17 22:12:27,962][openai._base_client][INFO] - Retrying request to /chat/completions in 0.447741 seconds
+[2025-12-17 22:12:28,057][openai._base_client][INFO] - Retrying request to /chat/completions in 0.414256 seconds
+[2025-12-17 22:12:28,236][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:28,295][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:28,577][openai._base_client][INFO] - Retrying request to /chat/completions in 0.777026 seconds
+[2025-12-17 22:12:29,035][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:29,169][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:29,844][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:29,885][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:31,239][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:31,256][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:31,637][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:32,004][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:32,551][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:32,887][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:32,895][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:32,982][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:33,258][openai._base_client][INFO] - Retrying request to /chat/completions in 0.985926 seconds
+[2025-12-17 22:12:33,590][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:33,836][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:34,263][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:34,338][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:34,385][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:34,701][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:35,111][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:35,979][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:36,691][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:37,122][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:37,788][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:38,036][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:38,205][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:38,519][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:38,797][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:38,810][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:38,820][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:39,060][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:39,341][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:39,551][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:40,414][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:40,589][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:40,867][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:42,105][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:42,227][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:42,303][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:42,377][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:42,531][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:43,131][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:43,196][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:43,628][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:43,743][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:43,810][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:43,812][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:44,086][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:44,330][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:44,795][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:44,849][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:45,210][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:45,323][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:46,009][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:46,066][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:47,050][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:48,280][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:53,289][openai._base_client][INFO] - Retrying request to /chat/completions in 0.486726 seconds
+[2025-12-17 22:12:58,301][openai._base_client][INFO] - Retrying request to /chat/completions in 0.452051 seconds
+[2025-12-17 22:12:58,302][openai._base_client][INFO] - Retrying request to /chat/completions in 0.411771 seconds
+[2025-12-17 22:12:58,418][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:12:59,297][openai._base_client][INFO] - Retrying request to /chat/completions in 0.785537 seconds
+[2025-12-17 22:12:59,553][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:02,720][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:06,164][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:06,374][openai._base_client][INFO] - Retrying request to /chat/completions in 0.392103 seconds
+[2025-12-17 22:13:06,376][openai._base_client][INFO] - Retrying request to /chat/completions in 0.488766 seconds
+[2025-12-17 22:13:06,863][openai._base_client][INFO] - Retrying request to /chat/completions in 0.463152 seconds
+[2025-12-17 22:13:06,871][openai._base_client][INFO] - Retrying request to /chat/completions in 0.391973 seconds
+[2025-12-17 22:13:06,872][openai._base_client][INFO] - Retrying request to /chat/completions in 0.389787 seconds
+[2025-12-17 22:13:06,875][openai._base_client][INFO] - Retrying request to /chat/completions in 0.455291 seconds
+[2025-12-17 22:13:06,882][openai._base_client][INFO] - Retrying request to /chat/completions in 0.493787 seconds
+[2025-12-17 22:13:06,915][openai._base_client][INFO] - Retrying request to /chat/completions in 0.394976 seconds
+[2025-12-17 22:13:07,768][openai._base_client][INFO] - Retrying request to /chat/completions in 0.818449 seconds
+[2025-12-17 22:13:07,829][openai._base_client][INFO] - Retrying request to /chat/completions in 0.786717 seconds
+[2025-12-17 22:13:07,919][openai._base_client][INFO] - Retrying request to /chat/completions in 0.891454 seconds
+[2025-12-17 22:13:09,694][openai._base_client][INFO] - Retrying request to /chat/completions in 0.393567 seconds
+[2025-12-17 22:13:09,801][openai._base_client][INFO] - Retrying request to /chat/completions in 0.470791 seconds
+[2025-12-17 22:13:09,819][openai._base_client][INFO] - Retrying request to /chat/completions in 0.450295 seconds
+[2025-12-17 22:13:10,628][openai._base_client][INFO] - Retrying request to /chat/completions in 0.843196 seconds
+[2025-12-17 22:13:10,784][openai._base_client][INFO] - Retrying request to /chat/completions in 0.913686 seconds
+[2025-12-17 22:13:10,885][openai._base_client][INFO] - Retrying request to /chat/completions in 0.973925 seconds
+[2025-12-17 22:13:11,254][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:11,725][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:11,762][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:11,986][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:12,171][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:12,193][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:12,290][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:12,297][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:12,487][openai._base_client][INFO] - Retrying request to /chat/completions in 0.391575 seconds
+[2025-12-17 22:13:12,907][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:13,062][openai._base_client][INFO] - Retrying request to /chat/completions in 0.399038 seconds
+[2025-12-17 22:13:13,075][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:13,434][openai._base_client][INFO] - Retrying request to /chat/completions in 0.872143 seconds
+[2025-12-17 22:13:13,959][openai._base_client][INFO] - Retrying request to /chat/completions in 0.990413 seconds
+[2025-12-17 22:13:14,616][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:14,788][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:14,863][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:15,306][openai._base_client][INFO] - Retrying request to /chat/completions in 0.479567 seconds
+[2025-12-17 22:13:15,709][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:15,839][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:15,911][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:16,294][openai._base_client][INFO] - Retrying request to /chat/completions in 0.964564 seconds
+[2025-12-17 22:13:16,463][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:16,983][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:17,760][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:18,008][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:18,152][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:18,311][openai._base_client][INFO] - Retrying request to /chat/completions in 0.401399 seconds
+[2025-12-17 22:13:18,658][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:18,885][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:18,929][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:19,094][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:19,375][openai._base_client][INFO] - Retrying request to /chat/completions in 0.833866 seconds
+[2025-12-17 22:13:19,509][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:19,807][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:20,036][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:21,278][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:21,321][openai._base_client][INFO] - Retrying request to /chat/completions in 0.441944 seconds
+[2025-12-17 22:13:22,027][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:22,283][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:22,521][openai._base_client][INFO] - Retrying request to /chat/completions in 0.982776 seconds
+[2025-12-17 22:13:22,753][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:22,782][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:22,801][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:23,025][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:23,060][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:23,296][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:23,828][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:23,837][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:24,311][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:24,402][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:24,552][openai._base_client][INFO] - Retrying request to /chat/completions in 0.476497 seconds
+[2025-12-17 22:13:24,629][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:25,726][openai._base_client][INFO] - Retrying request to /chat/completions in 0.992546 seconds
+[2025-12-17 22:13:26,086][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:26,547][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:26,555][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:26,671][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:26,974][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:27,253][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:27,591][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:27,872][openai._base_client][INFO] - Retrying request to /chat/completions in 0.472380 seconds
+[2025-12-17 22:13:28,427][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:28,546][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:28,650][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:28,998][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:29,090][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:29,495][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:30,295][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:30,423][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:30,948][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:30,951][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:31,210][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:31,543][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:32,037][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:32,427][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:32,433][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:32,788][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:33,242][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:33,580][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:34,950][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:39,958][openai._base_client][INFO] - Retrying request to /chat/completions in 0.412724 seconds
+[2025-12-17 22:13:40,469][openai._base_client][INFO] - Retrying request to /chat/completions in 0.421660 seconds
+[2025-12-17 22:13:40,478][openai._base_client][INFO] - Retrying request to /chat/completions in 0.376600 seconds
+[2025-12-17 22:13:40,493][openai._base_client][INFO] - Retrying request to /chat/completions in 0.379375 seconds
+[2025-12-17 22:13:40,503][openai._base_client][INFO] - Retrying request to /chat/completions in 0.467010 seconds
+[2025-12-17 22:13:40,551][openai._base_client][INFO] - Retrying request to /chat/completions in 0.379720 seconds
+[2025-12-17 22:13:40,584][openai._base_client][INFO] - Retrying request to /chat/completions in 0.499449 seconds
+[2025-12-17 22:13:40,589][openai._base_client][INFO] - Retrying request to /chat/completions in 0.402081 seconds
+[2025-12-17 22:13:40,647][openai._base_client][INFO] - Retrying request to /chat/completions in 0.486434 seconds
+[2025-12-17 22:13:40,694][openai._base_client][INFO] - Retrying request to /chat/completions in 0.426023 seconds
+[2025-12-17 22:13:40,768][openai._base_client][INFO] - Retrying request to /chat/completions in 0.483882 seconds
+[2025-12-17 22:13:41,381][openai._base_client][INFO] - Retrying request to /chat/completions in 0.870733 seconds
+[2025-12-17 22:13:41,414][openai._base_client][INFO] - Retrying request to /chat/completions in 0.981753 seconds
+[2025-12-17 22:13:41,518][openai._base_client][INFO] - Retrying request to /chat/completions in 0.909140 seconds
+[2025-12-17 22:13:41,567][openai._base_client][INFO] - Retrying request to /chat/completions in 0.829048 seconds
+[2025-12-17 22:13:41,613][openai._base_client][INFO] - Retrying request to /chat/completions in 0.938879 seconds
+[2025-12-17 22:13:41,646][openai._base_client][INFO] - Retrying request to /chat/completions in 0.894225 seconds
+[2025-12-17 22:13:41,833][openai._base_client][INFO] - Retrying request to /chat/completions in 0.779200 seconds
+[2025-12-17 22:13:45,436][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:46,065][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:47,097][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:47,426][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:48,320][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:53,331][openai._base_client][INFO] - Retrying request to /chat/completions in 0.475203 seconds
+[2025-12-17 22:13:53,835][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413323 seconds
+[2025-12-17 22:13:53,842][openai._base_client][INFO] - Retrying request to /chat/completions in 0.493395 seconds
+[2025-12-17 22:13:53,888][openai._base_client][INFO] - Retrying request to /chat/completions in 0.443205 seconds
+[2025-12-17 22:13:53,952][openai._base_client][INFO] - Retrying request to /chat/completions in 0.448484 seconds
+[2025-12-17 22:13:58,534][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:13:58,812][openai._base_client][INFO] - Retrying request to /chat/completions in 0.812502 seconds
+[2025-12-17 22:13:59,257][openai._base_client][INFO] - Retrying request to /chat/completions in 0.793379 seconds
+[2025-12-17 22:14:00,211][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:02,577][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:02,935][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:04,730][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:05,896][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:06,172][openai._base_client][INFO] - Retrying request to /chat/completions in 0.447226 seconds
+[2025-12-17 22:14:06,174][openai._base_client][INFO] - Retrying request to /chat/completions in 0.442895 seconds
+[2025-12-17 22:14:06,176][openai._base_client][INFO] - Retrying request to /chat/completions in 0.418267 seconds
+[2025-12-17 22:14:06,180][openai._base_client][INFO] - Retrying request to /chat/completions in 0.493748 seconds
+[2025-12-17 22:14:11,095][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:11,178][openai._base_client][INFO] - Retrying request to /chat/completions in 0.464254 seconds
+[2025-12-17 22:14:11,179][openai._base_client][INFO] - Retrying request to /chat/completions in 0.409329 seconds
+[2025-12-17 22:14:11,180][openai._base_client][INFO] - Retrying request to /chat/completions in 0.407777 seconds
+[2025-12-17 22:14:11,181][openai._base_client][INFO] - Retrying request to /chat/completions in 0.413046 seconds
+[2025-12-17 22:14:11,182][openai._base_client][INFO] - Retrying request to /chat/completions in 0.471643 seconds
+[2025-12-17 22:14:11,182][openai._base_client][INFO] - Retrying request to /chat/completions in 0.457442 seconds
+[2025-12-17 22:14:11,183][openai._base_client][INFO] - Retrying request to /chat/completions in 0.448894 seconds
+[2025-12-17 22:14:11,183][openai._base_client][INFO] - Retrying request to /chat/completions in 0.437169 seconds
+[2025-12-17 22:14:11,218][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:11,585][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:11,598][openai._base_client][INFO] - Retrying request to /chat/completions in 0.810601 seconds
+[2025-12-17 22:14:11,622][openai._base_client][INFO] - Retrying request to /chat/completions in 0.791737 seconds
+[2025-12-17 22:14:11,675][openai._base_client][INFO] - Retrying request to /chat/completions in 0.998483 seconds
+[2025-12-17 22:14:12,043][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:14,306][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:15,755][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:16,055][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:16,085][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:16,590][openai._base_client][INFO] - Retrying request to /chat/completions in 0.870561 seconds
+[2025-12-17 22:14:16,596][openai._base_client][INFO] - Retrying request to /chat/completions in 0.776345 seconds
+[2025-12-17 22:14:16,634][openai._base_client][INFO] - Retrying request to /chat/completions in 0.815641 seconds
+[2025-12-17 22:14:16,642][openai._base_client][INFO] - Retrying request to /chat/completions in 0.895587 seconds
+[2025-12-17 22:14:16,645][openai._base_client][INFO] - Retrying request to /chat/completions in 0.829714 seconds
+[2025-12-17 22:14:16,655][openai._base_client][INFO] - Retrying request to /chat/completions in 0.872324 seconds
+[2025-12-17 22:14:17,589][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:17,851][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:17,879][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:17,913][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:18,721][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:19,430][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:19,581][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:19,989][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:22,572][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:22,634][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:22,742][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:22,854][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:23,167][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:23,252][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:23,426][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:23,893][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:24,042][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:24,392][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:24,558][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:25,932][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:26,533][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:27,044][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:27,320][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:27,382][openai._base_client][INFO] - Retrying request to /chat/completions in 0.497674 seconds
+[2025-12-17 22:14:27,456][openai._base_client][INFO] - Retrying request to /chat/completions in 0.475314 seconds
+[2025-12-17 22:14:27,467][openai._base_client][INFO] - Retrying request to /chat/completions in 0.462640 seconds
+[2025-12-17 22:14:27,479][openai._base_client][INFO] - Retrying request to /chat/completions in 0.479634 seconds
+[2025-12-17 22:14:28,394][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:28,633][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:29,078][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:29,416][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:29,466][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:30,554][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:30,955][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:31,078][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:31,399][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:32,796][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:32,883][openai._base_client][INFO] - Retrying request to /chat/completions in 0.988495 seconds
+[2025-12-17 22:14:32,932][openai._base_client][INFO] - Retrying request to /chat/completions in 0.827484 seconds
+[2025-12-17 22:14:32,934][openai._base_client][INFO] - Retrying request to /chat/completions in 0.810069 seconds
+[2025-12-17 22:14:32,943][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:33,041][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:33,288][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:33,781][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:34,029][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:34,257][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:34,485][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:37,583][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:37,780][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:37,949][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:38,555][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:39,732][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:44,743][openai._base_client][INFO] - Retrying request to /chat/completions in 0.414785 seconds
+[2025-12-17 22:14:49,510][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:49,754][openai._base_client][INFO] - Retrying request to /chat/completions in 0.457812 seconds
+[2025-12-17 22:14:49,754][openai._base_client][INFO] - Retrying request to /chat/completions in 0.407761 seconds
+[2025-12-17 22:14:49,755][openai._base_client][INFO] - Retrying request to /chat/completions in 0.431865 seconds
+[2025-12-17 22:14:50,161][openai._base_client][INFO] - Retrying request to /chat/completions in 0.857610 seconds
+[2025-12-17 22:14:52,303][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:55,192][openai._base_client][INFO] - Retrying request to /chat/completions in 0.942711 seconds
+[2025-12-17 22:14:55,367][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:55,895][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:14:56,060][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:02,925][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:03,086][openai._base_client][INFO] - Retrying request to /chat/completions in 0.399513 seconds
+[2025-12-17 22:15:06,708][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:08,088][openai._base_client][INFO] - Retrying request to /chat/completions in 0.477652 seconds
+[2025-12-17 22:15:08,089][openai._base_client][INFO] - Retrying request to /chat/completions in 0.461572 seconds
+[2025-12-17 22:15:08,235][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:08,236][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:08,444][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:08,488][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:08,615][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:08,765][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:08,804][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:08,922][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:09,315][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:10,209][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:10,683][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:10,689][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:10,780][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:12,112][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:12,539][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:12,906][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:12,949][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:12,970][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:13,213][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:13,569][openai._base_client][INFO] - Retrying request to /chat/completions in 0.810229 seconds
+[2025-12-17 22:15:13,743][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:14,175][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:14,764][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:14,942][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:15,355][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:15,463][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:15,696][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:16,407][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:16,849][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:16,937][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:17,549][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:17,574][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:17,829][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:17,979][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:18,001][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:18,596][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:20,032][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:20,662][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:20,763][openai._base_client][INFO] - Retrying request to /chat/completions in 0.376250 seconds
+[2025-12-17 22:15:20,801][openai._base_client][INFO] - Retrying request to /chat/completions in 0.412889 seconds
+[2025-12-17 22:15:20,803][openai._base_client][INFO] - Retrying request to /chat/completions in 0.392015 seconds
+[2025-12-17 22:15:20,805][openai._base_client][INFO] - Retrying request to /chat/completions in 0.469682 seconds
+[2025-12-17 22:15:20,806][openai._base_client][INFO] - Retrying request to /chat/completions in 0.494774 seconds
+[2025-12-17 22:15:20,807][openai._base_client][INFO] - Retrying request to /chat/completions in 0.414310 seconds
+[2025-12-17 22:15:20,808][openai._base_client][INFO] - Retrying request to /chat/completions in 0.491533 seconds
+[2025-12-17 22:15:20,810][openai._base_client][INFO] - Retrying request to /chat/completions in 0.476448 seconds
+[2025-12-17 22:15:20,810][openai._base_client][INFO] - Retrying request to /chat/completions in 0.498342 seconds
+[2025-12-17 22:15:20,811][openai._base_client][INFO] - Retrying request to /chat/completions in 0.474165 seconds
+[2025-12-17 22:15:20,812][openai._base_client][INFO] - Retrying request to /chat/completions in 0.460884 seconds
+[2025-12-17 22:15:20,813][openai._base_client][INFO] - Retrying request to /chat/completions in 0.446331 seconds
+[2025-12-17 22:15:25,709][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:25,814][openai._base_client][INFO] - Retrying request to /chat/completions in 0.412399 seconds
+[2025-12-17 22:15:25,814][openai._base_client][INFO] - Retrying request to /chat/completions in 0.428744 seconds
+[2025-12-17 22:15:26,032][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:26,143][openai._base_client][INFO] - Retrying request to /chat/completions in 0.816205 seconds
+[2025-12-17 22:15:26,198][openai._base_client][INFO] - Retrying request to /chat/completions in 0.950454 seconds
+[2025-12-17 22:15:26,224][openai._base_client][INFO] - Retrying request to /chat/completions in 0.826182 seconds
+[2025-12-17 22:15:26,263][openai._base_client][INFO] - Retrying request to /chat/completions in 0.760577 seconds
+[2025-12-17 22:15:26,278][openai._base_client][INFO] - Retrying request to /chat/completions in 0.813327 seconds
+[2025-12-17 22:15:26,289][openai._base_client][INFO] - Retrying request to /chat/completions in 0.840216 seconds
+[2025-12-17 22:15:26,290][openai._base_client][INFO] - Retrying request to /chat/completions in 0.953939 seconds
+[2025-12-17 22:15:26,777][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:27,185][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:28,584][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:28,862][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:30,182][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:30,251][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:30,893][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:31,230][openai._base_client][INFO] - Retrying request to /chat/completions in 0.967631 seconds
+[2025-12-17 22:15:31,246][openai._base_client][INFO] - Retrying request to /chat/completions in 0.838430 seconds
+[2025-12-17 22:15:31,571][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:31,574][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:31,769][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:32,864][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:35,609][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:38,125][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:43,136][openai._base_client][INFO] - Retrying request to /chat/completions in 0.429120 seconds
+[2025-12-17 22:15:48,153][openai._base_client][INFO] - Retrying request to /chat/completions in 0.384359 seconds
+[2025-12-17 22:15:48,154][openai._base_client][INFO] - Retrying request to /chat/completions in 0.499837 seconds
+[2025-12-17 22:15:48,154][openai._base_client][INFO] - Retrying request to /chat/completions in 0.440875 seconds
+[2025-12-17 22:15:48,154][openai._base_client][INFO] - Retrying request to /chat/completions in 0.487382 seconds
+[2025-12-17 22:15:48,155][openai._base_client][INFO] - Retrying request to /chat/completions in 0.443106 seconds
+[2025-12-17 22:15:49,360][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:53,105][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:53,601][openai._base_client][INFO] - Retrying request to /chat/completions in 0.889291 seconds
+[2025-12-17 22:15:54,207][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:54,934][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:55,944][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:58,260][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:15:58,409][openai._base_client][INFO] - Retrying request to /chat/completions in 0.498314 seconds
+[2025-12-17 22:15:58,410][openai._base_client][INFO] - Retrying request to /chat/completions in 0.472199 seconds
+[2025-12-17 22:15:58,412][openai._base_client][INFO] - Retrying request to /chat/completions in 0.394043 seconds
+[2025-12-17 22:16:02,983][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:03,413][openai._base_client][INFO] - Retrying request to /chat/completions in 0.408972 seconds
+[2025-12-17 22:16:03,413][openai._base_client][INFO] - Retrying request to /chat/completions in 0.444027 seconds
+[2025-12-17 22:16:03,414][openai._base_client][INFO] - Retrying request to /chat/completions in 0.449493 seconds
+[2025-12-17 22:16:03,415][openai._base_client][INFO] - Retrying request to /chat/completions in 0.470035 seconds
+[2025-12-17 22:16:03,438][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:03,454][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:03,811][openai._base_client][INFO] - Retrying request to /chat/completions in 0.865244 seconds
+[2025-12-17 22:16:03,885][openai._base_client][INFO] - Retrying request to /chat/completions in 0.942014 seconds
+[2025-12-17 22:16:03,911][openai._base_client][INFO] - Retrying request to /chat/completions in 0.852332 seconds
+[2025-12-17 22:16:04,774][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:04,885][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:04,983][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:05,090][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:06,061][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:06,670][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:07,346][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:09,080][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:09,161][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:11,191][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:12,141][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:17,155][openai._base_client][INFO] - Retrying request to /chat/completions in 0.391676 seconds
+[2025-12-17 22:16:21,671][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:22,341][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:22,432][openai._base_client][INFO] - Retrying request to /chat/completions in 0.447669 seconds
+[2025-12-17 22:16:22,434][openai._base_client][INFO] - Retrying request to /chat/completions in 0.497385 seconds
+[2025-12-17 22:16:26,633][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:27,436][openai._base_client][INFO] - Retrying request to /chat/completions in 0.495072 seconds
+[2025-12-17 22:16:27,436][openai._base_client][INFO] - Retrying request to /chat/completions in 0.421126 seconds
+[2025-12-17 22:16:27,658][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:28,124][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:28,533][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:28,768][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:29,116][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:29,601][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:31,114][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:31,751][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:32,634][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:32,720][openai._base_client][INFO] - Retrying request to /chat/completions in 0.417702 seconds
+[2025-12-17 22:16:32,721][openai._base_client][INFO] - Retrying request to /chat/completions in 0.451000 seconds
+[2025-12-17 22:16:32,723][openai._base_client][INFO] - Retrying request to /chat/completions in 0.459882 seconds
+[2025-12-17 22:16:32,724][openai._base_client][INFO] - Retrying request to /chat/completions in 0.490006 seconds
+[2025-12-17 22:16:32,724][openai._base_client][INFO] - Retrying request to /chat/completions in 0.385726 seconds
+[2025-12-17 22:16:32,725][openai._base_client][INFO] - Retrying request to /chat/completions in 0.407908 seconds
+[2025-12-17 22:16:32,726][openai._base_client][INFO] - Retrying request to /chat/completions in 0.460291 seconds
+[2025-12-17 22:16:32,726][openai._base_client][INFO] - Retrying request to /chat/completions in 0.477924 seconds
+[2025-12-17 22:16:37,468][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:37,727][openai._base_client][INFO] - Retrying request to /chat/completions in 0.482476 seconds
+[2025-12-17 22:16:38,061][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:38,140][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:38,248][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:38,710][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:39,846][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:40,003][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:40,384][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:42,857][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:42,923][openai._base_client][INFO] - Retrying request to /chat/completions in 0.431873 seconds
+[2025-12-17 22:16:42,925][openai._base_client][INFO] - Retrying request to /chat/completions in 0.467349 seconds
+[2025-12-17 22:16:42,926][openai._base_client][INFO] - Retrying request to /chat/completions in 0.432511 seconds
+[2025-12-17 22:16:42,927][openai._base_client][INFO] - Retrying request to /chat/completions in 0.484359 seconds
+[2025-12-17 22:16:43,357][openai._base_client][INFO] - Retrying request to /chat/completions in 0.941255 seconds
+[2025-12-17 22:16:43,361][openai._base_client][INFO] - Retrying request to /chat/completions in 0.867093 seconds
+[2025-12-17 22:16:48,508][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:49,720][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:51,943][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:52,761][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:52,798][openai._base_client][INFO] - Retrying request to /chat/completions in 0.444682 seconds
+[2025-12-17 22:16:52,799][openai._base_client][INFO] - Retrying request to /chat/completions in 0.396482 seconds
+[2025-12-17 22:16:52,800][openai._base_client][INFO] - Retrying request to /chat/completions in 0.478047 seconds
+[2025-12-17 22:16:53,198][openai._base_client][INFO] - Retrying request to /chat/completions in 0.884646 seconds
+[2025-12-17 22:16:56,834][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:16:58,416][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:02,045][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:02,089][openai._base_client][INFO] - Retrying request to /chat/completions in 0.481064 seconds
+[2025-12-17 22:17:02,089][openai._base_client][INFO] - Retrying request to /chat/completions in 0.410141 seconds
+[2025-12-17 22:17:08,775][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:11,495][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:11,523][openai._base_client][INFO] - Retrying request to /chat/completions in 0.395923 seconds
+[2025-12-17 22:17:11,922][openai._base_client][INFO] - Retrying request to /chat/completions in 0.813119 seconds
+[2025-12-17 22:17:19,561][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:19,595][openai._base_client][INFO] - Retrying request to /chat/completions in 0.427838 seconds
+[2025-12-17 22:17:28,248][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:28,270][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3502).
+[2025-12-17 22:17:28,292][openai._base_client][INFO] - Retrying request to /chat/completions in 0.466019 seconds
+[2025-12-17 22:17:35,825][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:35,841][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3494).
+[2025-12-17 22:17:35,853][openai._base_client][INFO] - Retrying request to /chat/completions in 0.428544 seconds
+[2025-12-17 22:17:44,037][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:44,057][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3504).
+[2025-12-17 22:17:44,069][openai._base_client][INFO] - Retrying request to /chat/completions in 0.430254 seconds
+[2025-12-17 22:17:51,677][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:51,693][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3504).
+[2025-12-17 22:17:51,707][openai._base_client][INFO] - Retrying request to /chat/completions in 0.494348 seconds
+[2025-12-17 22:17:59,134][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:17:59,155][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3490).
+[2025-12-17 22:17:59,169][openai._base_client][INFO] - Retrying request to /chat/completions in 0.456097 seconds
+[2025-12-17 22:18:06,915][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:18:06,936][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3490).
+[2025-12-17 22:18:06,950][openai._base_client][INFO] - Retrying request to /chat/completions in 0.414924 seconds
+[2025-12-17 22:18:14,944][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:18:14,964][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3439).
+[2025-12-17 22:18:14,982][openai._base_client][INFO] - Retrying request to /chat/completions in 0.440316 seconds
+[2025-12-17 22:18:20,432][openai._base_client][INFO] - Retrying request to /chat/completions in 0.843996 seconds
+[2025-12-17 22:18:28,899][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:18:28,913][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3425).
+[2025-12-17 22:18:28,924][openai._base_client][INFO] - Retrying request to /chat/completions in 0.433639 seconds
+[2025-12-17 22:18:34,363][openai._base_client][INFO] - Retrying request to /chat/completions in 0.873585 seconds
+[2025-12-17 22:18:50,263][openai._base_client][INFO] - Retrying request to /chat/completions in 0.442815 seconds
+[2025-12-17 22:18:55,713][openai._base_client][INFO] - Retrying request to /chat/completions in 0.932272 seconds
+[2025-12-17 22:19:11,674][openai._base_client][INFO] - Retrying request to /chat/completions in 0.416373 seconds
+[2025-12-17 22:19:17,098][openai._base_client][INFO] - Retrying request to /chat/completions in 0.806183 seconds
+[2025-12-17 22:19:26,700][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:19:26,723][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3422).
+[2025-12-17 22:19:26,742][openai._base_client][INFO] - Retrying request to /chat/completions in 0.434780 seconds
+[2025-12-17 22:19:32,184][openai._base_client][INFO] - Retrying request to /chat/completions in 0.757621 seconds
+[2025-12-17 22:19:49,822][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:19:49,844][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3422).
+[2025-12-17 22:19:49,864][openai._base_client][INFO] - Retrying request to /chat/completions in 0.478620 seconds
+[2025-12-17 22:19:58,450][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"
+[2025-12-17 22:19:58,464][root][WARNING] - Truncated history for env 26 to keep prompt within 3600 tokens (current 3421).
+[2025-12-17 22:19:58,475][openai._base_client][INFO] - Retrying request to /chat/completions in 0.436341 seconds
+[2025-12-17 22:20:03,920][openai._base_client][INFO] - Retrying request to /chat/completions in 0.928057 seconds
+[2025-12-17 22:20:12,263][httpx][INFO] - HTTP Request: POST https://api.ohmygpt.com/v1/chat/completions "HTTP/1.1 200 OK"

outputs/2025-12-17/22-28-10/.hydra/config.yaml

@@ -0,0 +1,903 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 3600
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: null
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 25
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 128
+    group_size: 1
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 128
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: ark-deepseek-v3-250324
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  ark-deepseek-v3-250324:
+    provider_name: openai
+    model_name: ark-deepseek-v3-250324
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/22-28-10/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/22-28-10
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/22-28-10/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/22-45-41/.hydra/config.yaml

@@ -0,0 +1,903 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 3600
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: null
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 25
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 128
+    group_size: 1
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 128
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: ark-deepseek-v3-250324
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  ark-deepseek-v3-250324:
+    provider_name: openai
+    model_name: ark-deepseek-v3-250324
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512

outputs/2025-12-17/22-45-41/.hydra/hydra.yaml

@@ -0,0 +1,171 @@
+hydra:
+  run:
+    dir: outputs/${now:%Y-%m-%d}/${now:%H-%M-%S}
+  sweep:
+    dir: multirun/${now:%Y-%m-%d}/${now:%H-%M-%S}
+    subdir: ${hydra.job.num}
+  launcher:
+    _target_: hydra._internal.core_plugins.basic_launcher.BasicLauncher
+  sweeper:
+    _target_: hydra._internal.core_plugins.basic_sweeper.BasicSweeper
+    max_batch_size: null
+    params: null
+  help:
+    app_name: ${hydra.job.name}
+    header: '${hydra.help.app_name} is powered by Hydra.
+
+      '
+    footer: 'Powered by Hydra (https://hydra.cc)
+
+      Use --hydra-help to view Hydra specific help
+
+      '
+    template: '${hydra.help.header}
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (group=option)
+
+
+      $APP_CONFIG_GROUPS
+
+
+      == Config ==
+
+      Override anything in the config (foo.bar=value)
+
+
+      $CONFIG
+
+
+      ${hydra.help.footer}
+
+      '
+  hydra_help:
+    template: 'Hydra (${hydra.runtime.version})
+
+      See https://hydra.cc for more info.
+
+
+      == Flags ==
+
+      $FLAGS_HELP
+
+
+      == Configuration groups ==
+
+      Compose your configuration from those groups (For example, append hydra/job_logging=disabled
+      to command line)
+
+
+      $HYDRA_CONFIG_GROUPS
+
+
+      Use ''--cfg hydra'' to Show the Hydra config.
+
+      '
+    hydra_help: ???
+  hydra_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][HYDRA] %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+    root:
+      level: INFO
+      handlers:
+      - console
+    loggers:
+      logging_example:
+        level: DEBUG
+    disable_existing_loggers: false
+  job_logging:
+    version: 1
+    formatters:
+      simple:
+        format: '[%(asctime)s][%(name)s][%(levelname)s] - %(message)s'
+    handlers:
+      console:
+        class: logging.StreamHandler
+        formatter: simple
+        stream: ext://sys.stdout
+      file:
+        class: logging.FileHandler
+        formatter: simple
+        filename: ${hydra.runtime.output_dir}/${hydra.job.name}.log
+    root:
+      level: INFO
+      handlers:
+      - console
+      - file
+    disable_existing_loggers: false
+  env: {}
+  mode: RUN
+  searchpath:
+  - file://./verl/verl/trainer/config
+  callbacks: {}
+  output_subdir: .hydra
+  overrides:
+    hydra:
+    - hydra.searchpath=[file://./verl/verl/trainer/config]
+    - hydra.mode=RUN
+    task: []
+  job:
+    name: eval_api
+    chdir: null
+    override_dirname: ''
+    id: ???
+    num: ???
+    config_name: evaluate_api_llm
+    env_set: {}
+    env_copy: []
+    config:
+      override_dirname:
+        kv_sep: '='
+        item_sep: ','
+        exclude_keys: []
+  runtime:
+    version: 1.3.2
+    version_base: '1.3'
+    cwd: /mnt/general/wanghy/RAGEN
+    config_sources:
+    - path: hydra.conf
+      schema: pkg
+      provider: hydra
+    - path: /mnt/general/wanghy/RAGEN/config
+      schema: file
+      provider: main
+    - path: ./verl/verl/trainer/config
+      schema: file
+      provider: hydra.searchpath in command-line
+    - path: ''
+      schema: structured
+      provider: schema
+    output_dir: /mnt/general/wanghy/RAGEN/outputs/2025-12-17/22-45-41
+    choices:
+      reward_model: dp_reward_model
+      critic: dp_critic
+      critic/../engine@critic.model.fsdp_config: fsdp
+      critic/../optim@critic.optim: fsdp
+      model@actor_rollout_ref.model: hf_model
+      rollout@actor_rollout_ref.rollout: rollout
+      ref@actor_rollout_ref.ref: dp_ref
+      ref/../engine@actor_rollout_ref.ref.fsdp_config: fsdp
+      data: legacy_data
+      actor@actor_rollout_ref.actor: dp_actor
+      actor/../engine@actor_rollout_ref.actor.fsdp_config: fsdp
+      actor/../optim@actor_rollout_ref.actor.optim: fsdp
+      hydra/env: default
+      hydra/callbacks: null
+      hydra/job_logging: default
+      hydra/hydra_logging: default
+      hydra/hydra_help: default
+      hydra/help: default
+      hydra/sweeper: basic
+      hydra/launcher: basic
+      hydra/output: default
+  verbose: false

outputs/2025-12-17/22-45-41/.hydra/overrides.yaml

@@ -0,0 +1 @@
+[]

outputs/2025-12-17/23-03-15/.hydra/config.yaml

@@ -0,0 +1,903 @@
+actor_rollout_ref:
+  actor:
+    optim:
+      _target_: verl.workers.config.FSDPOptimizerConfig
+      lr: 1.0e-06
+      lr_warmup_steps_ratio: 0.0
+      total_training_steps: -1
+      weight_decay: 0.01
+      lr_warmup_steps: -1
+      betas:
+      - 0.9
+      - 0.999
+      clip_grad: 1.0
+      min_lr_ratio: 0.0
+      num_cycles: 0.5
+      warmup_style: constant
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    _target_: verl.workers.config.FSDPActorConfig
+    strategy: fsdp
+    ppo_mini_batch_size: ${ppo_mini_batch_size}
+    ppo_micro_batch_size: null
+    ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    use_dynamic_bsz: false
+    ppo_max_token_len_per_gpu: 16384
+    clip_ratio: 0.2
+    clip_ratio_low: 0.2
+    clip_ratio_high: 0.28
+    freeze_vision_tower: false
+    policy_loss:
+      _target_: verl.workers.config.PolicyLossConfig
+      loss_mode: vanilla
+      clip_cov_ratio: 0.0002
+      clip_cov_lb: 1.0
+      clip_cov_ub: 5.0
+      kl_cov_ratio: 0.0002
+      ppo_kl_coef: 0.1
+    clip_ratio_c: 3.0
+    loss_agg_mode: token-mean
+    entropy_coeff: 0.001
+    tis_imp_ratio_cap: -1
+    use_kl_loss: false
+    use_torch_compile: true
+    kl_loss_coef: 0.0
+    kl_loss_type: kl
+    ppo_epochs: 1
+    shuffle: false
+    checkpoint:
+      _target_: verl.trainer.config.CheckpointConfig
+      save_contents:
+      - model
+      - optimizer
+      - extra
+      load_contents: ${.save_contents}
+      async_save: false
+    use_fused_kernels: ${oc.select:actor_rollout_ref.model.use_fused_kernels,false}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    grad_clip: 1.0
+    ulysses_sequence_parallel_size: 1
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+    use_remove_padding: ${oc.select:actor_rollout_ref.model.use_remove_padding,false}
+    use_ref: true
+    grpo_advantage_length_weight: ${grpo_advantage_length_weight}
+  ref:
+    strategy: ${actor_rollout_ref.actor.strategy}
+    use_torch_compile: ${oc.select:actor_rollout_ref.actor.use_torch_compile,true}
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: false
+      all_ranks: false
+      ranks: []
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config:
+        nsys:
+          _target_: verl.utils.profiler.config.NsightToolConfig
+          discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+        npu:
+          _target_: verl.utils.profiler.config.NPUToolConfig
+          contents: []
+          level: level1
+          analysis: true
+          discrete: false
+        torch:
+          _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+          step_start: 0
+          step_end: null
+        torch_memory:
+          _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+          trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+          stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    model: null
+    ulysses_sequence_parallel_size: ${oc.select:actor_rollout_ref.actor.ulysses_sequence_parallel_size,1}
+    entropy_from_logits_with_chunking: false
+    entropy_checkpointing: false
+  rollout:
+    _target_: verl.workers.config.RolloutConfig
+    name: vllm
+    mode: sync
+    temperature: 1
+    top_k: -1
+    top_p: 1
+    prompt_length: 1
+    response_length: 400
+    dtype: bfloat16
+    gpu_memory_utilization: 0.7
+    ignore_eos: false
+    enforce_eager: true
+    cudagraph_capture_sizes: null
+    free_cache_engine: true
+    tensor_model_parallel_size: 1
+    data_parallel_size: 1
+    expert_parallel_size: 1
+    max_num_batched_tokens: 8192
+    max_model_len: 3600
+    max_num_seqs: 1024
+    enable_chunked_prefill: true
+    enable_prefix_caching: true
+    load_format: dummy
+    log_prob_micro_batch_size: null
+    log_prob_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+    log_prob_use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+    log_prob_max_token_len_per_gpu: ${oc.select:actor_rollout_ref.actor.ppo_max_token_len_per_gpu,16384}
+    disable_log_stats: true
+    do_sample: true
+    'n': 1
+    over_sample_rate: 0
+    multi_stage_wake_up: false
+    engine_kwargs:
+      vllm: {}
+      sglang: {}
+    val_kwargs:
+      _target_: verl.workers.config.SamplingConfig
+      top_k: -1
+      top_p: 1.0
+      temperature: 0.5
+      'n': 1
+      do_sample: true
+    multi_turn:
+      _target_: verl.workers.config.MultiTurnConfig
+      enable: false
+      max_assistant_turns: null
+      tool_config_path: null
+      max_user_turns: null
+      max_parallel_calls: 1
+      max_tool_response_length: 256
+      tool_response_truncate_side: middle
+      interaction_config_path: null
+      use_inference_chat_template: false
+      tokenization_sanity_check_mode: strict
+      format: hermes
+      num_repeat_rollouts: null
+    calculate_log_probs: false
+    agent:
+      _target_: verl.workers.config.AgentLoopConfig
+      num_workers: 8
+      agent_loop_config_path: null
+      custom_async_server:
+        _target_: verl.workers.config.CustomAsyncServerConfig
+        path: null
+        name: null
+    update_weights_bucket_megabytes: 512
+    trace:
+      _target_: verl.workers.config.TraceConfig
+      backend: null
+      token2text: false
+    skip_rollout: false
+    skip_dump_dir: /tmp/rollout_dump
+    skip_tokenizer_init: true
+    profiler:
+      _target_: verl.utils.profiler.ProfilerConfig
+      tool: ${oc.select:global_profiler.tool,null}
+      enable: ${oc.select:actor_rollout_ref.actor.profiler.enable,false}
+      all_ranks: ${oc.select:actor_rollout_ref.actor.profiler.all_ranks,false}
+      ranks: ${oc.select:actor_rollout_ref.actor.profiler.ranks,[]}
+      save_path: ${oc.select:global_profiler.save_path,null}
+      tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+    layered_summon: false
+    rollout_filter_ratio: 0.25
+    rollout_filter_type: largest
+    rollout_filter_metric: reward_variance
+  model:
+    _target_: verl.workers.config.HFModelConfig
+    path: ${model_path}
+    hf_config_path: null
+    tokenizer_path: null
+    use_shm: false
+    trust_remote_code: false
+    custom_chat_template: null
+    external_lib: null
+    override_config: {}
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+    exclude_modules: null
+    use_liger: false
+    use_fused_kernels: false
+    fused_kernel_options:
+      impl_backend: torch
+  hybrid_engine: true
+  nccl_timeout: 600
+data:
+  tokenizer: null
+  use_shm: false
+  train_files: ~/data/rlhf/gsm8k/train.parquet
+  val_files: ~/data/rlhf/gsm8k/test.parquet
+  prompt_key: prompt
+  reward_fn_key: data_source
+  max_prompt_length: null
+  max_response_length: null
+  train_batch_size: null
+  val_batch_size: null
+  return_raw_input_ids: false
+  return_raw_chat: false
+  return_full_prompt: false
+  shuffle: true
+  dataloader_num_workers: 8
+  validation_shuffle: false
+  filter_overlong_prompts: false
+  filter_overlong_prompts_workers: 1
+  truncation: error
+  image_key: images
+  video_key: videos
+  trust_remote_code: false
+  custom_cls:
+    path: null
+    name: null
+  return_multi_modal_inputs: true
+  sampler:
+    class_path: null
+    class_name: null
+  datagen:
+    path: null
+    name: null
+  apply_chat_template_kwargs: {}
+critic:
+  optim:
+    _target_: verl.workers.config.FSDPOptimizerConfig
+    lr: 1.0e-05
+    lr_warmup_steps_ratio: 0.0
+    total_training_steps: -1
+    weight_decay: 0.01
+    lr_warmup_steps: -1
+    betas:
+    - 0.9
+    - 0.999
+    clip_grad: 1.0
+    min_lr_ratio: 0.0
+    num_cycles: 0.5
+    warmup_style: constant
+  model:
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      optimizer_offload: false
+      offload_policy: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+      model_dtype: fp32
+      use_orig_params: false
+      ulysses_sequence_parallel_size: 1
+      entropy_from_logits_with_chunking: false
+      use_torch_compile: true
+      entropy_checkpointing: false
+      forward_only: false
+      strategy: fsdp
+    path: ${model_path}
+    tokenizer_path: ${oc.select:actor_rollout_ref.model.path,"~/models/deepseek-llm-7b-chat"}
+    override_config: {}
+    external_lib: ${oc.select:actor_rollout_ref.model.external_lib,null}
+    trust_remote_code: ${oc.select:actor_rollout_ref.model.trust_remote_code,false}
+    _target_: verl.workers.config.FSDPCriticModelCfg
+    use_shm: false
+    enable_gradient_checkpointing: true
+    enable_activation_offload: false
+    use_remove_padding: false
+    lora_rank: ${lora.rank}
+    lora_alpha: ${lora.alpha}
+    target_modules: ${lora.target_modules}
+  _target_: verl.workers.config.FSDPCriticConfig
+  rollout_n: ${oc.select:actor_rollout_ref.rollout.n,1}
+  strategy: fsdp
+  enable: null
+  ppo_mini_batch_size: ${ppo_mini_batch_size}
+  ppo_micro_batch_size: null
+  ppo_micro_batch_size_per_gpu: ${micro_batch_size_per_gpu}
+  use_dynamic_bsz: ${oc.select:actor_rollout_ref.actor.use_dynamic_bsz,false}
+  ppo_max_token_len_per_gpu: 32768
+  forward_max_token_len_per_gpu: ${.ppo_max_token_len_per_gpu}
+  ppo_epochs: ${oc.select:actor_rollout_ref.actor.ppo_epochs,1}
+  shuffle: ${oc.select:actor_rollout_ref.actor.shuffle,false}
+  cliprange_value: 0.5
+  loss_agg_mode: ${oc.select:actor_rollout_ref.actor.loss_agg_mode,token-mean}
+  checkpoint:
+    _target_: verl.trainer.config.CheckpointConfig
+    save_contents:
+    - model
+    - optimizer
+    - extra
+    load_contents: ${.save_contents}
+    async_save: false
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config:
+      nsys:
+        _target_: verl.utils.profiler.config.NsightToolConfig
+        discrete: ${oc.select:global_profiler.global_tool_config.nsys.discrete}
+      npu:
+        _target_: verl.utils.profiler.config.NPUToolConfig
+        contents: []
+        level: level1
+        analysis: true
+        discrete: false
+      torch:
+        _target_: verl.utils.profiler.config.TorchProfilerToolConfig
+        step_start: 0
+        step_end: null
+      torch_memory:
+        _target_: verl.utils.profiler.config.TorchMemoryToolConfig
+        trace_alloc_max_entries: ${oc.select:global_profiler.global_tool_config.torch_memory.trace_alloc_max_entries,100000}
+        stack_depth: ${oc.select:global_profiler.global_tool_config.torch_memory.stack_depth,32}
+  forward_micro_batch_size: ${oc.select:.ppo_micro_batch_size,null}
+  forward_micro_batch_size_per_gpu: ${oc.select:.ppo_micro_batch_size_per_gpu,null}
+  ulysses_sequence_parallel_size: 1
+  grad_clip: 1.0
+reward_model:
+  enable: false
+  enable_resource_pool: false
+  n_gpus_per_node: 0
+  nnodes: 0
+  strategy: fsdp
+  model:
+    input_tokenizer: ${actor_rollout_ref.model.path}
+    path: ~/models/FsfairX-LLaMA3-RM-v0.1
+    external_lib: ${actor_rollout_ref.model.external_lib}
+    trust_remote_code: false
+    use_shm: false
+    use_remove_padding: false
+    use_fused_kernels: ${actor_rollout_ref.model.use_fused_kernels}
+    fsdp_config:
+      _target_: verl.workers.config.FSDPEngineConfig
+      wrap_policy:
+        min_num_params: 0
+      param_offload: false
+      reshard_after_forward: true
+      fsdp_size: -1
+      forward_prefetch: false
+  micro_batch_size: null
+  micro_batch_size_per_gpu: null
+  max_length: null
+  use_dynamic_bsz: ${critic.use_dynamic_bsz}
+  forward_max_token_len_per_gpu: ${critic.forward_max_token_len_per_gpu}
+  reward_manager: naive
+  launch_reward_fn_async: false
+  sandbox_fusion:
+    url: null
+    max_concurrent: 64
+    memory_limit_mb: 1024
+  profiler:
+    _target_: verl.utils.profiler.ProfilerConfig
+    tool: ${oc.select:global_profiler.tool,null}
+    enable: false
+    all_ranks: false
+    ranks: []
+    save_path: ${oc.select:global_profiler.save_path,null}
+    tool_config: ${oc.select:actor_rollout_ref.actor.profiler.tool_config,null}
+  ulysses_sequence_parallel_size: 1
+custom_reward_function:
+  path: null
+  name: compute_score
+algorithm:
+  _target_: verl.trainer.config.AlgoConfig
+  gamma: 1.0
+  lam: 1.0
+  adv_estimator: gae
+  norm_adv_by_std_in_grpo: true
+  use_kl_in_reward: false
+  kl_penalty: kl
+  kl_ctrl:
+    _target_: verl.trainer.config.KLControlConfig
+    type: fixed
+    kl_coef: 0.0
+    horizon: 10000
+    target_kl: 0.1
+  use_pf_ppo: false
+  pf_ppo:
+    reweight_method: pow
+    weight_pow: 2.0
+  high_level_gamma: 0.95
+  bi_level_gae: false
+trainer:
+  balance_batch: true
+  total_epochs: 30
+  total_training_steps: 200
+  project_name: null
+  experiment_name: test
+  logger:
+  - console
+  - wandb
+  log_val_generations: 0
+  rollout_data_dir: null
+  validation_data_dir: null
+  nnodes: 1
+  n_gpus_per_node: 1
+  save_freq: -1
+  esi_redundant_time: 0
+  resume_mode: auto
+  resume_from_path: null
+  val_before_train: true
+  val_only: false
+  test_freq: 10
+  critic_warmup: 0
+  default_hdfs_dir: null
+  del_local_ckpt_after_load: false
+  default_local_dir: /mnt/general/wanghy/RAGEN/saves/
+  max_actor_ckpt_to_keep: 1
+  max_critic_ckpt_to_keep: 1
+  ray_wait_register_center_timeout: 300
+  device: cuda
+  use_legacy_worker_impl: auto
+  local_log_dir: results/
+  validation_steps: 1
+  generations_to_log_to_wandb:
+    train: 128
+    val: 20
+global_profiler:
+  _target_: verl.utils.profiler.ProfilerConfig
+  tool: null
+  steps: null
+  profile_continuous_steps: false
+  save_path: outputs/profile
+  global_tool_config:
+    nsys:
+      _target_: verl.utils.profiler.config.NsightToolConfig
+      discrete: false
+      controller_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+      worker_nsight_options:
+        trace: cuda,nvtx,cublas,ucx
+        cuda-memory-usage: 'true'
+        cuda-graph-trace: graph
+        capture-range: cudaProfilerApi
+        capture-range-end: null
+        kill: none
+    torch_memory:
+      trace_alloc_max_entries: 100000
+      stack_depth: 32
+      context: all
+      stacks: all
+      kw_args: {}
+ray_kwargs:
+  ray_init:
+    num_cpus: null
+  timeline_json_file: null
+custom_envs:
+  SimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+  LargerSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      dim_x: 8
+      dim_y: 8
+      num_boxes: 2
+      max_steps: 100
+      search_depth: 10
+  SokobanDifferentGridVocab:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. \nYou are the player and\
+      \ you need to push all boxes to targets. \nWhen you are right next to a box,\
+      \ you can push it by moving in the same direction. \nYou cannot push a box through\
+      \ a wall, and you cannot pull a box. \nThe answer should be a sequence of actions,\
+      \ like <answer>Right || Right || Up</answer>\n"
+    max_tokens: 100
+    env_config:
+      search_depth: 30
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      grid_lookup:
+        0: W
+        1: .
+        2: G
+        3: C
+        4: B
+        5: A
+        6: '@'
+      grid_vocab:
+        W: wall
+        .: empty
+        G: target
+        C: box on target
+        B: box
+        A: player
+        '@': player on target
+  CoordSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: "You are solving the Sokoban puzzle. You are the player and you\
+      \ need to push all boxes to targets.\nYou are provided with a symbol grid and\
+      \ the zero-indexed coordinates of the player, each box, and each target. \n\
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner\
+      \ (5, 5). \nWhen you are exactly next to a box, you can push it by moving in\
+      \ the same direction. \nYou cannot push a box through a wall, and you cannot\
+      \ pull a box.\nThe answer should be a sequence of actions, like <answer>Right\
+      \ || Right || Up</answer>.\n"
+    max_tokens: 120
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 2
+      max_steps: 100
+      observation_format: grid_coord
+  VisualSimpleSokoban:
+    env_type: sokoban
+    max_actions_per_traj: 10
+    env_instruction: You are solving the Sokoban puzzle. You are the player and you
+      need to push all boxes to targets. When you are right next to a box, you can
+      push it by moving in the same direction. You cannot push a box through a wall,
+      and you cannot pull a box. The answer should be a sequence of actions, like
+      <answer>Right || Right || Up</answer>
+    max_tokens: 100
+    env_config:
+      dim_x: 6
+      dim_y: 6
+      num_boxes: 1
+      max_steps: 100
+      render_mode: rgb_array
+  Countdown:
+    env_type: countdown
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving the Countdown puzzle. You should use the num
+      list to create an equation that equals the target. Example answer format: <think>
+      To find an equation using [3, 5, 2] to get 4. Let''s check 2 + 5 = 7, 7 - 3
+      = 4. So the answer is 2 + 5 - 3 = 4. </think><answer>2 + 5 - 3</answer>'
+    max_tokens: 100
+    env_config: null
+  Bandit:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: train
+  BanditTest:
+    env_type: bandit
+    max_actions_per_traj: 1
+    env_instruction: ''
+    max_tokens: 100
+    env_config:
+      split: test
+  FrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. Forbid the whole and
+      go to the target. You may move to the unintended direction due to the slippery
+      ice. Example answer format: <think>To forbid the hole and go to the target,
+      I should go left then go up.</think><answer>Left || Up</answer>'
+    max_tokens: 100
+    env_config: null
+  CoordFrozenLake:
+    env_type: frozen_lake
+    max_actions_per_traj: 25
+    env_instruction: 'You are solving the FrozenLake puzzle. The observation includes
+      both a symbol grid and zero-indexed coordinates for the start, goal, player,
+      and any holes.
+
+      Coordinates range from the top-left corner (0, 0) to the bottom-right corner
+      (5, 5).
+
+      Beware that the ice is slippery, so the agent might slide and end up in an unintended
+      tile.
+
+      Respond with a sequence of actions such as <answer>Left || Up || Up</answer>.
+
+      '
+    max_tokens: 120
+    env_config:
+      observation_format: grid_coord
+  MetamathQA:
+    env_type: metamathqa
+    max_actions_per_traj: 1
+    env_instruction: 'You are solving Math problems. '
+    max_tokens: 100
+    env_config: null
+  WebShop:
+    env_type: webshop
+    max_actions_per_traj: 9
+    env_instruction: 'You are browsing an online shop. Based on the instruction, buy
+      a product that close to the production description. You need to search, read
+      the search results, pick a product, choose the size and color and buy. You should
+      only choose action from the available actions list provided later.  Example
+      process: I need a gingko light and 20x20 pillow cover that is hand painted.
+      First search[gingko light 20x20 pillow cover hand painted], answer format: <answer>search[blanket
+      with fleece throw]</answer>. Valid answer is search[<keywords>] or click[<clickable>].'
+    max_tokens: 200
+    env_config:
+      dataset: small
+  Lean:
+    env_type: lean
+    max_actions_per_traj: 30
+    env_instruction: You are a Lean theorem prover. Given a Lean theorem statement,
+      propose a sequence of tactics that completes the proof. Think step by step about
+      which tactics to apply next. Provide tactics separated by '||', for example
+      <answer>intro || simp || rfl</answer>.
+    max_tokens: 512
+    env_config: null
+  game_2048:
+    env_type: game_2048
+    max_actions_per_traj: 700
+    env_instruction: 'You are playing the 2048 game on a 4x4 grid. Merge equal tiles
+      by sliding Up, Right, Down, or Left.
+
+      If a move is invalid (no tiles move), a small penalty is applied. Respond with
+      a single action.
+
+      Example: <answer>Up</answer>
+
+      '
+    max_tokens: 8192
+    env_config: null
+  blackjack:
+    env_type: blackjack
+    max_actions_per_traj: 10
+    env_instruction: 'You are playing Blackjack against a dealer. The dealer must
+      hit on 16 or less and stand on 17 or more.
+
+      Choose either Stick or Hit. Respond with a single action.
+
+      Example: <answer>Hit</answer>
+
+      '
+    max_tokens: 64
+    env_config: null
+  rubikscube:
+    env_type: rubikscube
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a 2x2 Rubik''s Cube (Pocket Cube). The goal
+      is to restore the cube so that each of the faces consists of a single, unique
+      color.
+
+      Available actions use standard Singmaster notation for face rotations: U, U'',
+      D, D'', L, L'', R, R'', F, F'', B, B''.
+
+      - Faces: U (Up), D (Down), L (Left), R (Right), F (Front), B (Back).
+
+      - Modifiers: A letter alone means 90° clockwise (e.g., ''R''). A letter with
+      prime ('') means 90° counter-clockwise (e.g., "R''").
+
+      Respond with a sequence of actions separated by "||".
+
+      Example: <answer>U</answer>
+
+      '
+    max_tokens: 96
+    env_config:
+      scramble_depth: 1
+      max_steps: 20
+      render_mode: text
+  SimpleSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 20
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 2x2 box contains the numbers 1-4 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format, for example: <answer>place 1 at
+      row 2 col 3</answer> or <answer>1,2,3</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 4
+      difficulty: easy
+      render_format: with_feedback
+      show_conflicts: false
+      show_valid_numbers: false
+      max_steps: 20
+  MediumSudoku:
+    env_type: sudoku
+    max_actions_per_traj: 30
+    env_instruction: 'You are solving a Sudoku puzzle. Fill in the grid so that every
+      row, column, and 3x3 box contains the numbers 1-9 without repetition.
+
+      Initial cells are shown in [brackets] and cannot be modified. Empty cells are
+      shown as dots (.).
+
+      Place numbers one at a time using the format: <answer>place 5 at row 2 col 3</answer>
+      or <answer>2,3,5</answer>
+
+      The environment will provide feedback on valid/invalid moves and show conflicts
+      if any occur.
+
+      '
+    max_tokens: 150
+    parallel_friendly: false
+    max_workers: 32
+    env_config:
+      grid_size: 9
+      difficulty: medium
+      render_format: with_feedback
+      show_conflicts: true
+      show_valid_numbers: true
+      max_steps: 81
+system:
+  CUDA_VISIBLE_DEVICES: '0'
+seed:
+  train: 10000
+  val: 123
+micro_batch_size_per_gpu: 1
+ppo_mini_batch_size: 32
+model_path: Qwen/Qwen2.5-0.5B-Instruct
+enable_response_mask: true
+grpo_advantage_length_weight: false
+lora:
+  rank: 0
+  alpha: 16
+  target_modules: all-linear
+agent_proxy:
+  max_context_window: -1
+  max_turn: 25
+  action_sep: '||'
+  max_actions_per_turn: 1
+  use_turn_scores: false
+  enable_think: true
+  reward_normalization:
+    grouping: state
+    method: identity
+es_manager:
+  format_penalty: -0.1
+  train:
+    env_groups: 8
+    group_size: 16
+    env_configs:
+      tags:
+      - CoordSokoban
+      n_groups:
+      - 8
+  val:
+    env_groups: 128
+    group_size: 1
+    env_configs:
+      tags:
+      - rubikscube
+      n_groups:
+      - 128
+ctx_manager:
+  generation:
+    gen_config:
+      response_length: ${actor_rollout_ref.rollout.response_length}
+      temperature: ${actor_rollout_ref.rollout.temperature}
+      top_p: ${actor_rollout_ref.rollout.top_p}
+      top_k: ${actor_rollout_ref.rollout.top_k}
+      kwargs: null
+model_config:
+  model_name: ark-deepseek-v3-250324
+  max_concurrency: 16
+model_info:
+  Qwen2.5-7B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-7B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  Qwen2.5-72B-Instruct:
+    provider_name: together
+    model_name: Qwen/Qwen2.5-72B-Instruct-Turbo
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  claude-3.7:
+    provider_name: anthropic
+    model_name: claude-3-7-sonnet-20250219
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  gpt-4o-mini:
+    provider_name: openai
+    model_name: gpt-4o-mini
+    generation_kwargs:
+      temperature: 0
+      max_tokens: 512
+  deepseek-r1:
+    provider_name: deepseek
+    model_name: deepseek-reasoner
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  ark-deepseek-v3-250324:
+    provider_name: openai
+    model_name: ark-deepseek-v3-250324
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512
+  deepseek-v3:
+    provider_name: deepseek
+    model_name: deepseek-chat
+    generation_kwargs:
+      temperature: 0
+      max_completion_tokens: 512