cluster_args:
  cluster_description: null
  cluster_method: agglomerative
  compression_ratio: 0.3
  expert_sim: ttm
  frequency_penalty: true
  linkage_method: average
  max_cluster_size: null
  multi_layer: null
  num_clusters: null
  singleton_outlier_experts: false
  singleton_super_experts: false
  softmax_temperature: null
ds_args:
  dataset_config_name: null
  dataset_name: theblackcat102/evol-codealpaca-v1:171,Salesforce/xlam-function-calling-60k:171,open-r1/Mixture-of-Thoughts[code]:171,open-r1/Mixture-of-Thoughts[math]:171,open-r1/Mixture-of-Thoughts[science]:170,SWE-bench/SWE-smith-trajectories(tool):170
  dataset_test_split: test
  shuffle: true
  split: train
eval_args:
  evalplus_tasks:
  - mbpp
  - humaneval
  greedy: true
  lm_eval_tasks:
  - winogrande
  - arc_challenge
  - arc_easy
  - boolq
  - hellaswag
  - mmlu
  - openbookqa
  - rte
  min_p: 0.0
  parallel_tasks: 32
  results_dir: null
  run_evalplus: true
  run_livecodebench: true
  run_lm_eval: true
  run_math: false
  run_wildbench: false
  server_log_file_name: server.log
  temperature: 0.7
  top_k: 20
  top_p: 0.8
  use_server: true
  vllm_port: 8000
model_args:
  model_name: /disk1/rongxiao/hf_cache/hub/models--Qwen--Qwen3.6-35B-A3B/snapshots/995ad96eacd98c81ed38be0c5b274b04031597b0
  num_experts_per_tok_override: null
obs_args:
  batch_size: 1
  batches_per_category: 1024
  distance_measure: angular
  model_max_length: 2048
  output_file_name: observations_qwen36_pre_reap_bnb4_paper_1024_2048_standard_streaming-pre_reap-bnb_4bit-nf4-bfloat16-dq_true.pt
  overwrite_observations: false
  record_pruning_metrics_only: true
  renormalize_router_weights: true
  return_vllm_tokens_prompt: false
  select_only_categories: null
  split_by_category: false
  truncate: false
pre_reap_quant_args:
  pre_reap_bnb_4bit_compute_dtype: bfloat16
  pre_reap_bnb_4bit_quant_type: nf4
  pre_reap_bnb_4bit_use_double_quant: true
  pre_reap_quantization_method: bnb_4bit
prune_args:
  n_experts_to_prune: null
  overwrite_pruned_model: true
  perserve_outliers: false
  perserve_super_experts: false
  prune_method: reap
reap_args:
  debug: false
  do_eval: false
  plot_clusters: true
  profile: false
  run_observer_only: false
  seed: 42
  smoke_test: false