LLaDA/eval_LLaDA.sh at main · preordinary/LLaDA · 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
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
export CUDA_VISIBLE_DEVICES=2

BASE_OUTPUT_PATH="./results/llada"
MODEL_PATH="your model path"
length=512
steps=512
block_length=32


task="humaneval"
OUTPUT_PATH="${BASE_OUTPUT_PATH}/${task}_${length}"
accelerate launch evaluation_script.py \
    -m dllm_eval \
    --model LLaDA \
    --tasks ${task} \
    --batch_size 1 \
    --model_args "pretrained=${MODEL_PATH},assistant_prefix=<reasoning> " \
    --gen_kwargs "block_length=${block_length},gen_length=${length},steps=${steps},temperature=0" \
    --num_fewshot 0 \
    --output_path "${OUTPUT_PATH}" \
    --log_samples \
    --confirm_run_unsafe_code

python metrics/humaneval.py \
    --res_path "${OUTPUT_PATH}"


task="gsm8k"
OUTPUT_PATH="${BASE_OUTPUT_PATH}/${task}_${length}"
accelerate launch evaluation_script.py \
    -m dllm_eval \
    --model LLaDA \
    --tasks ${task} \
    --batch_size 1 \
    --model_args "pretrained=${MODEL_PATH},assistant_prefix=<reasoning> " \
    --gen_kwargs "block_length=${block_length},gen_length=${length},steps=${steps},temperature=0" \
    --num_fewshot 0 \
    --output_path "${OUTPUT_PATH}" \
    --log_samples \
    --confirm_run_unsafe_code

python metrics/gsm8k.py \
    --res_path "${OUTPUT_PATH}"


task="mbpp"
OUTPUT_PATH="${BASE_OUTPUT_PATH}/${task}_${length}"
accelerate launch evaluation_script.py \
    -m dllm_eval \
    --model LLaDA \
    --tasks ${task} \
    --batch_size 1 \
    --model_args "pretrained=${MODEL_PATH},assistant_prefix=<reasoning> " \
    --gen_kwargs "block_length=${block_length},gen_length=${length},steps=${steps},temperature=0" \
    --num_fewshot 0 \
    --output_path "${OUTPUT_PATH}" \
    --log_samples \
    --confirm_run_unsafe_code

python metrics/mbpp.py \
    --res_path "${OUTPUT_PATH}"


task="math500"
OUTPUT_PATH="${BASE_OUTPUT_PATH}/${task}_${length}"
accelerate launch evaluation_script.py \
    -m dllm_eval \
    --model LLaDA \
    --tasks ${task} \
    --batch_size 1 \
    --model_args "pretrained=${MODEL_PATH},assistant_prefix=<reasoning> " \
    --gen_kwargs "block_length=${block_length},gen_length=${length},steps=${steps},temperature=0" \
    --num_fewshot 0 \
    --output_path "${OUTPUT_PATH}" \
    --log_samples \
    --confirm_run_unsafe_code

python metrics/math500.py \
    --res_path "${OUTPUT_PATH}"