apunkt/doc-to-lora
1
0
Fork 0
mirror of https://github.com/SakanaAI/doc-to-lora.git synced 2026-07-23 17:01:04 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

webui chat + result display + eval datasets + preprocessing eval + refactoring data_utils.py to data_definition.py

Browse source
This commit is contained in:
51616 2025-05-20 06:06:20 +00:00
parent cdfd42cf43
commit 0e4c1444f9
11 changed files with 891 additions and 424 deletions
Download patch file Download diff file Expand all files Collapse all files

45
webui/app.py
View file

@ -21,9 +21,14 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
modulated_model = None
# TODO: grab the results on-deman
# TODO: cache the grabed results
def get_run_data(run_path):
"""
Loads and processes data from all_results.json and *_results.json for a given run.
Also recursively searches for results files in subfolders.
Args:
run_path: Path to the directory containing the results files.
@ -38,7 +43,7 @@ def get_run_data(run_path):
except FileNotFoundError:
data = {}
# Load data from *_results.json files
# Load data from *_results.json files in the current directory
for results_json_file in glob(os.path.join(run_path, "*_results.json")):
try:
with open(results_json_file) as f:
@ -55,6 +60,24 @@ def get_run_data(run_path):
except FileNotFoundError:
print(f"Warning: {results_json_file} not found.")
# Recursively search for *_results.json files in subdirectories
for subdir in [d for d in glob(os.path.join(run_path, "*")) if os.path.isdir(d)]:
subdir_name = os.path.basename(subdir)
# Look for *_results.json files in the subdirectory
for results_json_file in glob(os.path.join(subdir, "*_results.json")):
try:
with open(results_json_file) as f:
split_data = json.load(f)
# Use a prefix that includes the subfolder name
file_prefix = f"{subdir_name}_{os.path.basename(results_json_file).replace('_results.json', '')}"
for key, value in split_data.items():
if isinstance(value, float):
value = round(value, 4)
new_key = f"{file_prefix}_{key}"
data[new_key] = value
except FileNotFoundError:
print(f"Warning: {results_json_file} not found.")
grouped_data = {}
for key, value in data.items():
if isinstance(value, dict):
@ -96,6 +119,7 @@ def get_run_data(run_path):
def get_generated_text_data(run_path):
"""
Loads generated text data from all *_generated_text.jsonl files.
Also recursively searches for generated text files in subfolders.
Args:
run_path: Path to the directory containing the .jsonl files.
@ -106,7 +130,7 @@ def get_generated_text_data(run_path):
generated_data = {}
files = sorted(glob(f"{run_path}/*_generated_text.jsonl"))
print(files)
# Find all *_generated_text.jsonl files
# Find all *_generated_text.jsonl files in the current directory
for filename in files:
# Extract split name from filename (remove _generated_text.jsonl)
split = filename.split("/")[-1].split("_generated_text.jsonl")[0]
@ -117,6 +141,23 @@ def get_generated_text_data(run_path):
generated_data[split] = data
except FileNotFoundError:
generated_data[split] = None
# Recursively search for *_generated_text.jsonl files in subdirectories
for subdir in [d for d in glob(os.path.join(run_path, "*")) if os.path.isdir(d)]:
subdir_name = os.path.basename(subdir)
subdir_files = sorted(glob(f"{subdir}/*_generated_text.jsonl"))
for filename in subdir_files:
# Extract split name from filename and include subfolder name
base_split = filename.split("/")[-1].split("_generated_text.jsonl")[0]
split = f"{subdir_name}_{base_split}"
try:
with open(filename) as f:
lines = f.readlines()
data = [json.loads(line) for line in lines]
generated_data[split] = data
except FileNotFoundError:
generated_data[split] = None
return generated_data