Mirrors/text-generation-webui
1
0
Fork 0
mirror of https://github.com/oobabooga/text-generation-webui.git synced 2024-09-20 18:45:09 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Fix "perplexity evaluation" progress messages

Browse source
This commit is contained in:
oobabooga 2023-05-23 01:54:52 -03:00
parent 4d94a111d4
commit 75adc110d4
1 changed files with 9 additions and 7 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

16
modules/evaluate.py
View file

@ -28,7 +28,9 @@ past_evaluations = load_past_evaluations()
def save_past_evaluations(df): def save_past_evaluations(df):
global past_evaluations global past_evaluations
past_evaluations = df past_evaluations = df
df.to_csv(Path('logs/evaluations.csv'), index=False) filepath = Path('logs/evaluations.csv')
filepath.parent.mkdir(parents=True, exist_ok=True)
df.to_csv(filepath, index=False)
def calculate_perplexity(models, input_dataset, stride, _max_length): def calculate_perplexity(models, input_dataset, stride, _max_length):
@ -39,7 +41,7 @@ def calculate_perplexity(models, input_dataset, stride, _max_length):
global past_evaluations global past_evaluations
cumulative_log = '' cumulative_log = ''
cumulative_log += "Loading the input dataset...\n" cumulative_log += "Loading the input dataset...\n\n"
yield cumulative_log yield cumulative_log
# Copied from https://github.com/qwopqwop200/GPTQ-for-LLaMa/blob/triton/utils/datautils.py # Copied from https://github.com/qwopqwop200/GPTQ-for-LLaMa/blob/triton/utils/datautils.py
@ -58,13 +60,13 @@ def calculate_perplexity(models, input_dataset, stride, _max_length):
for model in models: for model in models:
if is_in_past_evaluations(model, input_dataset, stride, _max_length): if is_in_past_evaluations(model, input_dataset, stride, _max_length):
cumulative_log += f"{model} has already been tested. Ignoring.\n" cumulative_log += f"{model} has already been tested. Ignoring.\n\n"
yield cumulative_log yield cumulative_log
continue continue
if model != 'current model': if model != 'current model':
try: try:
yield cumulative_log + f"Loading {model}...\n" yield cumulative_log + f"Loading {model}...\n\n"
model_settings = get_model_specific_settings(model) model_settings = get_model_specific_settings(model)
shared.settings.update(model_settings) # hijacking the interface defaults shared.settings.update(model_settings) # hijacking the interface defaults
update_model_parameters(model_settings) # hijacking the command-line arguments update_model_parameters(model_settings) # hijacking the command-line arguments
@ -72,12 +74,12 @@ def calculate_perplexity(models, input_dataset, stride, _max_length):
unload_model() unload_model()
shared.model, shared.tokenizer = load_model(shared.model_name) shared.model, shared.tokenizer = load_model(shared.model_name)
except: except:
cumulative_log += f"Failed to load {model}. Moving on.\n" cumulative_log += f"Failed to load {model}. Moving on.\n\n"
yield cumulative_log yield cumulative_log
continue continue
cumulative_log += f"Processing {model}...\n" cumulative_log += f"Processing {model}...\n\n"
yield cumulative_log + "Tokenizing the input dataset...\n" yield cumulative_log + "Tokenizing the input dataset...\n\n"
encodings = encode(text, add_special_tokens=False) encodings = encode(text, add_special_tokens=False)
seq_len = encodings.shape[1] seq_len = encodings.shape[1]
max_length = _max_length or shared.model.config.max_position_embeddings max_length = _max_length or shared.model.config.max_position_embeddings