andreaskoepf commited on
Commit
5443432
1 Parent(s): 7ec8d12

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +45 -0
README.md CHANGED
@@ -1,3 +1,48 @@
1
  ---
2
  license: apache-2.0
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
  ---
4
+ https://wandb.ai/open-assistant/supervised-finetuning/runs/7pz5n33h
5
+
6
+ datasets:
7
+ ```
8
+ oasst_export_eu:
9
+ datasets:
10
+ - oasst_export:
11
+ lang: "en,es,de,fr"
12
+ input_file_path: 2023-03-27_oasst_research_ready_synth.jsonl.gz
13
+ - alpaca
14
+ - oig_file:
15
+ source_url: https://huggingface.co/datasets/laion/OIG/resolve/main/unified_chip2.jsonl
16
+ max_count: 15000
17
+ min_length: 500
18
+ val_split: 0.2
19
+ - oig_file:
20
+ source_url: https://huggingface.co/datasets/laion/OIG/raw/main/unified_grade_school_math_instructions.jsonl
21
+ val_split: 0.1
22
+ min_length: 1000
23
+ sort_by_length: false
24
+ use_custom_sampler: false
25
+ ```
26
+
27
+ pythia:
28
+ ```
29
+ pythia-12b:
30
+ fp16: true
31
+ log_dir: "pythia_log_12b"
32
+ learning_rate: 6e-6
33
+ model_name: EleutherAI/pythia-12b-deduped
34
+ output_dir: pythia_model_12b
35
+ weight_decay: 0.0
36
+ residual_dropout: 0.2
37
+ max_length: 2048
38
+ use_flash_attention: true
39
+ warmup_steps: 100
40
+ gradient_checkpointing: false
41
+ gradient_accumulation_steps: 4
42
+ per_device_train_batch_size: 2
43
+ per_device_eval_batch_size: 5
44
+ eval_steps: 200
45
+ save_steps: 500
46
+ num_train_epochs: 16
47
+ save_total_limit: 4
48
+ ```