default.yaml

model:
  name: "unsloth/Llama-3.2-1B"
  dtype: "bfloat16"

calib:
  data_path: "./data/calib/refinedweb"
  dataset: "refinedweb"
  num_samples: 256
  seq_len: 1024
  batch_size: 8
  seed: 2023

profiling:
  module_names:
    - "self_attn.q_proj"
    - "self_attn.k_proj"
    - "self_attn.v_proj"
    - "self_attn.o_proj"
    - "mlp.gate_proj"
    - "mlp.up_proj"
    - "mlp.down_proj"
  cr_candidates: [0.05, 0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6, 0.65, 0.7]
  ks_ratios: [1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0, 2.1, 2.2, 2.3, 2.4, 2.5, 2.6, 2.7, 2.8, 2.9, 3.0]
  profile_cache: "./layer_prof_llama3_1b_v2.json"
  cr_cache: "./cr_layer_prof_llama3_1b_v4.json"

compression:
  target_kept_ratio: 0.20   
  param_precision: 50000
  dobi_like: false
  output_dir: "./llama3_1b_swift_svd_20_new"
  adam_refine_steps: 100
  method: dijekstra

evaluation:
  tasks: ["wikitext"] #["mmlu", "arc_easy", "arc_challenge", "hellaswag", "piqa", "lambada_openai", "race", "sciq"] #
  batch_size: 4
  max_batch_size: 16
  device: "cuda"

evaluation_ppl:
  dataset_name: "c4"