forked from AGI-Edgerunners/LLM-Adapters
-
Notifications
You must be signed in to change notification settings - Fork 0
/
multi_dataset_eval.py
43 lines (30 loc) · 1.31 KB
/
multi_dataset_eval.py
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
from concurrent.futures import ProcessPoolExecutor
import queue
import subprocess
def evaluate(dataset, gpu):
print('*******dataset:', dataset)
command = f"CUDA_VISIBLE_DEVICES={gpu} python evaluate.py \
--model LLaMA-7B \
--adapter LoRA \
--dataset {dataset} \
--base_model '../LLM/models/llama-7b-hf' \
--lora_weights './trained_models/llama-lora'"
result = subprocess.run(command, shell=True, text=True, capture_output=False)
print(f"Evaluation results for dataset {dataset} on GPU {gpu}:\n{result.stdout}")
return gpu
datasets = ['AQuA', 'AddSub', 'MultiArith', 'SingleEq', 'gsm8k', 'SVAMP']
gpus = [1, 2, 3]
tasks_queue = queue.Queue()
gpu_queue = queue.Queue()
for gpu in gpus:
gpu_queue.put(gpu)
for task in datasets:
tasks_queue.put(task)
num_processes = min(len(datasets), len(gpus)) # number of processes to run in parallel
with ProcessPoolExecutor(max_workers=num_processes) as executor:
futures = [executor.submit(evaluate, tasks_queue.get(), gpu_queue.get()) for i in range(num_processes)]
for future in futures:
gpu_id = future.result()
gpu_queue.put(gpu_id)
if tasks_queue.qsize() > 0:
futures.append(executor.submit(evaluate, tasks_queue.get(), gpu_queue.get()))