|
|
|
@ -1,6 +1,8 @@
|
|
|
|
|
import json
|
|
|
|
|
import torch
|
|
|
|
|
import pickle
|
|
|
|
|
import numpy as np
|
|
|
|
|
from tqdm import tqdm
|
|
|
|
|
from read import read_config
|
|
|
|
|
from argparse import ArgumentParser
|
|
|
|
|
from peft import PeftModelForCausalLM
|
|
|
|
@ -22,7 +24,7 @@ def setup_model(config):
|
|
|
|
|
if added_tokens > 0:
|
|
|
|
|
model.resize_token_embeddings(len(tokenizer))
|
|
|
|
|
|
|
|
|
|
if config["lora"]:
|
|
|
|
|
if 'lora' in config and config['lora']:
|
|
|
|
|
model = PeftModelForCausalLM.from_pretrained(model, config["lora_path"], device_map="auto", torch_dtype=torch.float16, return_hidden_states=True)
|
|
|
|
|
model.to(dtype=torch.float16)
|
|
|
|
|
|
|
|
|
@ -33,10 +35,8 @@ def setup_model(config):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def eval_example(model, tokenizer, example, config):
|
|
|
|
|
|
|
|
|
|
#set up data
|
|
|
|
|
prompt = example['instruction'] + ' ' + example['instances'][0]['input']
|
|
|
|
|
gt = prompt + ' ' + example['instances'][0]['output']
|
|
|
|
|
|
|
|
|
@ -45,29 +45,40 @@ def eval_example(model, tokenizer, example, config):
|
|
|
|
|
input = {k: v.to(model.device) for k, v in input.items()}
|
|
|
|
|
|
|
|
|
|
continuations = []
|
|
|
|
|
tokenized_continuations = []
|
|
|
|
|
trajectories = []
|
|
|
|
|
for i in range(5):
|
|
|
|
|
print(i)
|
|
|
|
|
outputs = model.generate(input_ids=input['input_ids'],
|
|
|
|
|
max_new_tokens=config["max_new_tokens"],
|
|
|
|
|
temperature=config["temperature"])
|
|
|
|
|
|
|
|
|
|
y = model(input_ids=outputs)
|
|
|
|
|
for i in range(3):
|
|
|
|
|
with torch.no_grad():
|
|
|
|
|
outputs = model.generate(input_ids=input['input_ids'],
|
|
|
|
|
max_new_tokens=config["max_new_tokens"],
|
|
|
|
|
min_new_tokens=5,
|
|
|
|
|
temperature=config["temperature"],
|
|
|
|
|
repetition_penalty=1.0,
|
|
|
|
|
do_sample=True)
|
|
|
|
|
decoded = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()
|
|
|
|
|
|
|
|
|
|
y = model(input_ids=outputs)
|
|
|
|
|
trajectory = y.hidden_states[0].detach().cpu().numpy()[0]
|
|
|
|
|
trajectory = trajectory / np.linalg.norm(trajectory, axis=1, keepdims=True)
|
|
|
|
|
trajectory = np.cumsum(trajectory, axis=0) / np.arange(1, trajectory.shape[0]+1).reshape(-1, 1)
|
|
|
|
|
decoded = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()
|
|
|
|
|
|
|
|
|
|
trajectories.append(trajectory)
|
|
|
|
|
continuations.append(decoded[len(prompt):])
|
|
|
|
|
continuations.append(decoded)
|
|
|
|
|
tokenized_continuations.append(tokenizer.tokenize(decoded))
|
|
|
|
|
|
|
|
|
|
#compute the ground truth perplexity
|
|
|
|
|
gt_input = tokenizer(gt, return_tensors="pt")
|
|
|
|
|
gt_input = {k: v.to(model.device) for k, v in gt_input.items()}
|
|
|
|
|
|
|
|
|
|
nlls = []
|
|
|
|
|
prev_end_loc = 0
|
|
|
|
|
for begin_loc in tqdm(range(len(prompt), len(gt), 1)):
|
|
|
|
|
end_loc = min(begin_loc + max_length, seq_len)
|
|
|
|
|
stride = 512
|
|
|
|
|
seq_len = gt_input['input_ids'].size(1)
|
|
|
|
|
|
|
|
|
|
for begin_loc in tqdm(range(input['input_ids'].size(1), gt_input['input_ids'].size(1), stride)):
|
|
|
|
|
end_loc = min(begin_loc + stride, seq_len)
|
|
|
|
|
trg_len = end_loc - prev_end_loc # may be different from stride on last loop
|
|
|
|
|
input_ids = input['input_ids'][:, begin_loc:end_loc].to(model.device)
|
|
|
|
|
input_ids = gt_input['input_ids'][:, begin_loc:end_loc].to(model.device)
|
|
|
|
|
target_ids = input_ids.clone()
|
|
|
|
|
target_ids[:, :-trg_len] = -100
|
|
|
|
|
|
|
|
|
@ -80,26 +91,36 @@ def eval_example(model, tokenizer, example, config):
|
|
|
|
|
if end_loc == seq_len:
|
|
|
|
|
break
|
|
|
|
|
|
|
|
|
|
ppl = torch.exp(torch.stack(nlls).sum() / end_loc)
|
|
|
|
|
ppl = torch.exp(torch.stack(nlls).sum() / end_loc).item()
|
|
|
|
|
print('ppl: ', ppl)
|
|
|
|
|
|
|
|
|
|
print('perplexity: ', ppl)
|
|
|
|
|
print('trajectories: ', trajectories)
|
|
|
|
|
print('continuations: ', continuations)
|
|
|
|
|
print(prompt)
|
|
|
|
|
print(80*'-')
|
|
|
|
|
for continuation in continuations:
|
|
|
|
|
print(continuation)
|
|
|
|
|
print(80*'-')
|
|
|
|
|
|
|
|
|
|
raise
|
|
|
|
|
|
|
|
|
|
return ppl, trajectories, continuations
|
|
|
|
|
return ppl, trajectories, continuations, tokenized_continuations
|
|
|
|
|
|
|
|
|
|
def do_eval(config):
|
|
|
|
|
eval_data = read_jsonl_file('eval_data/user_oriented_instructions.jsonl')
|
|
|
|
|
model, tokenizer = setup_model(config)
|
|
|
|
|
trajectories = []
|
|
|
|
|
perplexities = []
|
|
|
|
|
continuations = []
|
|
|
|
|
for example in eval_data:
|
|
|
|
|
gt_perplexity, trajectories, continuations = eval_example(model, tokenizer, example, config)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
all_trajectories = []
|
|
|
|
|
all_perplexities = []
|
|
|
|
|
all_continuations = []
|
|
|
|
|
all_tokenized_continuations = []
|
|
|
|
|
for example in tqdm(eval_data):
|
|
|
|
|
gt_perplexity, trajectories, continuations, tokenized_continuations = eval_example(model, tokenizer, example, config)
|
|
|
|
|
all_trajectories.append(trajectories)
|
|
|
|
|
all_perplexities.append(gt_perplexity)
|
|
|
|
|
all_continuations.append(continuations)
|
|
|
|
|
|
|
|
|
|
with open('eval_data/eval__model-{}__lora-{}.pkl'.format(config['model_name'].replace('/', '_'), config['lora_path'].replace('/', '_')), 'wb') as f:
|
|
|
|
|
r = {'trajectories': all_trajectories,
|
|
|
|
|
'perplexities': all_perplexities,
|
|
|
|
|
'continuations': all_continuations,
|
|
|
|
|
'tokenized_continuations': all_tokenized_continuations}
|
|
|
|
|
pickle.dump(r, f)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == '__main__':
|
|
|
|
|