Skip to content

fix internvl2 eval MME #198

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
llmc-reviewer merged 1 commit into from
Nov 18, 2024
Merged

fix internvl2 eval MME #198

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 13 additions & 3 deletions llmc/eval/eval_vlm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,7 @@ def __init__(self, config):
], 'VLM eval only support MME dataset now.'
self.eval_dataset_path = self.eval_config['path']
self.eval_bs = self.eval_config['bs']
self.output_include_input = True
if self.dataset == 'MME':
self.img_qas = self.load_mme()
self.patch_datasets(config.model.type)
Expand All @@ -39,6 +40,10 @@ def patch_datasets(self, model_type):
for idx in range(len(self.img_qas)):
if '<image>\n' not in self.img_qas[idx]['question']:
self.img_qas[idx]['question'] = '<image>\n' + self.img_qas[idx]['question']
if model_type == 'InternVL2':
self.output_include_input = False
elif model_type == 'Llava':
self.output_include_input = True

def eval(self, model, tokenizer):
vlm_model = model.vlm_model
Expand All @@ -64,9 +69,14 @@ def eval(self, model, tokenizer):
for k, v in inputs.items()
}
outputs = vlm_model.generate(**inputs, max_new_tokens=32, do_sample=False)
gen_txts = vlm_tokenizer.batch_decode(
outputs[:, inputs['input_ids'].shape[1]:], skip_special_tokens=True
)
if self.output_include_input:
gen_txts = vlm_tokenizer.batch_decode(
outputs[:, inputs['input_ids'].shape[1]:], skip_special_tokens=True
)
else:
gen_txts = vlm_tokenizer.batch_decode(
outputs, skip_special_tokens=True
)
for n in range(len(batch_samples)):
result = batch_samples[n].copy()
result.update({'gen_txt': gen_txts[n]})
Expand Down