COLD_decoding/commongen.sh at main · qkaren/COLD_decoding · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
#!/bin/bash

## CommonGen

python3 cold_decoding.py \
	--seed 12 \
	--mode lexical_generation \
	--pretrained_model gpt2-xl  \
	--init-temp 1 \
    --length 10 \
	--max-length 40 \
	--num-iters 2000 \
	--min-iters 1000 \
	--constraint-weight 0.5 \
    --abductive-c2-weight 0.1 \
	--stepsize 0.1 \
	--noise-iters 1 \
	--win-anneal-iters 1000 \
	--start 0 \
	--end 5 \
	--lr-nll-portion 0.6 \
    --topk 5 \
    --output-lgt-temp 1 \
	--verbose \
    --straight-through  \
	--large-noise-iters 50,500,1000,1500 \
	--large_gs_std 1,0.5,0.1,0.05  \
	--stepsize-ratio 1  \
    --batch-size 32 \
    --repeat-batch 8 \
    --print-every 200 \
    --input-file "./data/commongen/commongen.dev.jsonl" \
	--output-dir "./data/commongen/" \