Clean the convert to torch script

This commit is contained in:
oobabooga 2023-01-07 00:04:52 -03:00
parent c7b29668a2
commit 898e12058e

View File

@ -1,38 +1,27 @@
'''
Converts a transformers model to .pt, which is faster to load.
Run with python convert.py /path/to/model/
Make sure to write /path/to/model/ with a trailing / and not
/path/to/model
Example:
python convert.py models/opt-1.3b
Output will be written to torch-dumps/name-of-the-model.pt
'''
from transformers import AutoModelForCausalLM, AutoModelForSeq2SeqLM, OPTForCausalLM, AutoTokenizer, set_seed
from transformers import GPT2Tokenizer, GPT2Model, T5Tokenizer, T5ForConditionalGeneration
from transformers import AutoModelForCausalLM, T5ForConditionalGeneration
import torch
import sys
from sys import argv
import time
import glob
import psutil
print(f"torch-dumps/{argv[1].split('/')[-2]}.pt")
if argv[1].endswith('pt'):
model = OPTForCausalLM.from_pretrained(argv[1], device_map="auto")
torch.save(model, f"torch-dumps/{argv[1].split('/')[-2]}.pt")
elif 'galactica' in argv[1].lower():
model = OPTForCausalLM.from_pretrained(argv[1], low_cpu_mem_usage=True, torch_dtype=torch.float16)
#model = OPTForCausalLM.from_pretrained(argv[1], low_cpu_mem_usage=True, load_in_8bit=True)
torch.save(model, f"torch-dumps/{argv[1].split('/')[-2]}.pt")
elif 'flan-t5' in argv[1].lower():
model = T5ForConditionalGeneration.from_pretrained(argv[1], low_cpu_mem_usage=True, torch_dtype=torch.float16)
torch.save(model, f"torch-dumps/{argv[1].split('/')[-2]}.pt")
else:
print("Loading the model")
model = AutoModelForCausalLM.from_pretrained(argv[1], low_cpu_mem_usage=True, torch_dtype=torch.float16)
print("Model loaded")
#model = AutoModelForCausalLM.from_pretrained(argv[1], device_map='auto', load_in_8bit=True)
torch.save(model, f"torch-dumps/{argv[1].split('/')[-2]}.pt")
path = argv[1]
if path[-1] != '/':
path = path+'/'
model_name = path.split('/')[-2]
print(f"Loading {model_name}...")
if model_name in ['flan-t5', 't5-large']:
model = T5ForConditionalGeneration.from_pretrained(path).cuda()
else:
model = AutoModelForCausalLM.from_pretrained(path, low_cpu_mem_usage=True, torch_dtype=torch.float16).cuda()
print("Model loaded.")
print(f"Saving to torch-dumps/{model_name}.pt")
torch.save(model, f"torch-dumps/{model_name}.pt")