apiche

Apiche's group workspace

Group: debug_actor_mcp_tir

1-3

of 3

Tags

Notes

Author

apiche

State

Crashed

Start time

August 18th, 2025 2:56:59 PM

Runtime

2h 36m 1s

Tracked hours

Run path

apiche/pipeline-rl/debug_actor_mcp_tir_finetune

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

Python version

CPython 3.11.11

Git repository

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

Git state

git checkout -b "debug_actor_mcp_tir/finetune" d2e6d09deb3a2efddd79d8e42d14ac2ca8e6101f

Command

pipelinerl/entrypoints/run_finetune.py --config-dir results/debug_actor_mcp_tir/conf --config-name exp_config output_dir=results/debug_actor_mcp_tir hydra.run.dir=results/debug_actor_mcp_tir/finetune +me.weight_update_group_init_method=tcp://localhost:9000 +me.weight_update_group_world_size=3 +me.llm_urls=http://localhost:8080+http://localhost:8081

System Hardware

CPU count	112
Logical CPU count	224
GPU count	4
GPU type	NVIDIA H100 80GB HBM3

W&B CLI Version

0.19.11

Group

debug_actor_mcp_tir

Config parameters are your model's inputs. Learn more

▶
Config parameters:{} 207 keys
- _cpu:
  "False"
- _mixed_precision:
  "no"
- accelerate_config:
  null
- actor.discount_factor:
  1
- actor.llm_max_rollouts:
  64
- actor.log_each_n_secs:
  0
- actor.problem_queue_size:
  64
- actor.result_queue_size:
  64
- actor.rollout_policy:
  "pipelinerl.domains.tir_mcp.generate_math_rollout2"
- actor.rollout_workers:
  1
- actor.shared_memory_entry_size:
  10,000,000
- actor.system_prompt:
  "Please reason step by step, and put your final answer within \boxed{}."
- actor.task_template:
  "{task}"
- actor.throughput_window_size:
  50
- agent_max_loops:
  3
- agent._target_:
  "tapeagents.agent.Agent"
- agent.max_iterations:
  5
- agent.name:
  "mcp_agent"
- ▶
  agent.nodes:[] 5 items
- agent.store_llm_calls:
  true
- agent.templates.allowed_steps:
  "You have access to the following tools: {tools_description} "
- agent.templates.allowed_tools:
  "You have access to the following tools: {tools_description} "
- agent.templates.format:
  "Output only a single JSON dict. Do not repeat the last thought again. If the last action does not change the observation, do not repeat it! DO NOT OUTPUT ANYTHING BESIDES THE JSON! DO NOT PLACE ANY COMMENTS INSIDE THE JSON. It will break the system that processes the output. "
- agent.templates.system_prompt:
  "You are an expert AI Agent trained to assist users with complex information processing tasks. Your role is to understand user queries and respond in a helpful and accurate manner. Keep your replies concise and direct. Prioritize clarity and avoid over-elaboration. Do not express emotions or opinions about user questions. You must use the python tool for computation. "
- agent.templates.thought_format:
  "Important! Respond with the plain text, do not include any JSON or code. Do not output anything besides what I asked in this message. "
- attempts:
  1
- backend:
  "nccl"
- dataset_loader:
  "pipelinerl.domains.math.load_datasets"
- debug:
  "False"
- debug.mode:
  ""
- debug.place_inference_workers:
  true
- debug.streams_from:
  null
- debug.use_existing_llms:
  false
- deepspeed_config:
  "deepspeed_stage3_bf16"
- deepspeed_plugins:
  "DeepSpeedPlugin(hf_ds_config=<accelerate.utils.deepspeed.HfDeepSpeedConfig object at 0x7ffc769b7910>, gradient_accumulation_steps=1, gradient_clipping='auto', zero_stage=3, is_train_batch_min=True, offload_optimizer_device='none', offload_param_device='none', offload_optimizer_nvme_path='none', offload_param_nvme_path='none', zero3_init_flag=True, zero3_save_16bit_model=True, transformer_moe_cls_names=None, enable_msamp=False, msamp_opt_level='O1')"
- device:
  "cuda:0"
- distributed_type:
  "DistributedType.DEEPSPEED"
- dynamo_plugin:
  "TorchDynamoPlugin(backend=<DynamoBackend.NO: 'NO'>, mode='default', fullgraph=False, dynamic=None, options=None, disable=False, use_regional_compilation=False)"
- environment._target_:
  "pipelinerl.domains.tir_mcp.env_server.MCPEnvironmentServer"
- environment.exp_path:
  "results/debug_actor_mcp_tir/env_server"
- environment.host:
  "localhost"
- environment.math_target:
  "pipelinerl.domains.math.MathEnvironment"
- environment.mcp_config_path:
  "/home/toolkit/research-now-reasoner/pipelinerl/conf/mcp/python.json"
- environment.mcp_target:
  "tapeagents.mcp.MCPEnvironment"
- ▶
  environment.mcp_tools_whitelist:[] 1 item
  - 0:
    "run_python_code"
- environment.n_envs:
  8
- world.environment_start_port:
  7,777
- world.finetune_fraction:
  4
- world.preprocessor_fraction:
  0
- world.replicas:
  1

Summary metrics are your model's outputs. Learn more

No summary metrics saved for this run.

Check the summary metrics documentation for more information.