Text Generation
Transformers
PyTorch
RefinedWeb
sft
custom_code
text-generation-inference
andreaskoepf's picture
Update README.md
e43fda3
|
raw
history blame
1.93 kB
metadata
license: apache-2.0

Supervised fine-tuning of falcon40b with a mix of OASST top-2 threads and synthetic instruction datasets. Exported at end of 2nd epoch.

Model:

falcon-40b:
  dtype: bf16
  log_dir: "falcon_log_40b"
  learning_rate: 1e-5
  model_name: "tiiuae/falcon-40b"
  deepspeed_config: configs/zero3_config_falcon.json
  output_dir: falcon
  weight_decay: 0.0
  max_length: 2048
  warmup_steps: 20
  gradient_checkpointing: true
  gradient_accumulation_steps: 1
  per_device_train_batch_size: 18
  per_device_eval_batch_size: 10
  eval_steps: 120
  #save_steps: 80
  num_train_epochs: 8
  save_total_limit: 4
  use_flash_attention: false
  residual_dropout: 0.3
  residual_dropout_lima: true
  sort_by_length: false
  save_strategy: steps

Dataset:

sft9-stage2:
  # oasst_export: 100.00% (29899)
  # vicuna: 50.00% (16963)
  # code_alpaca: 50.00% (9510)
  # oa_wiki_qa_bart_10000row: 100.00% (9434)
  # grade_school_math_instructions: 100.00% (8351)
  # dolly15k: 100.00% (14250)

  save_strategy: steps # epoch seems not to work, gets stuck with DS 0.9.1
  save_steps: 613
  use_custom_sampler: true
  datasets:
    - oasst_export:
        lang: "bg,ca,cs,da,de,en,es,fr,hr,hu,it,nl,pl,pt,ro,ru,sl,sr,sv,uk" # sft-8.0
        input_file_path: 2023-06-02_oasst_all_labels.jsonl.gz
        val_split: 0.05
        top_k: 2
    - vicuna:
        fraction: 0.5
        val_split: 0.025
        max_val_set: 250
    - code_alpaca:
        fraction: 0.5
        val_split: 0.05
        max_val_set: 250
    - oa_wiki_qa_bart_10000row:
        val_split: 0.05
        max_val_set: 250
    - grade_school_math_instructions:
        val_split: 0.05
    - dolly15k:
        val_split: 0.05
        max_val_set: 300