Spaces:
Running
Running
File size: 4,368 Bytes
6204efe |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 |
import json
import os
import re
from collections import defaultdict
from datetime import datetime, timedelta, timezone
import huggingface_hub
from huggingface_hub import ModelCard
from huggingface_hub.hf_api import ModelInfo
from transformers import AutoConfig
from transformers.models.auto.tokenization_auto import tokenizer_class_from_name, get_tokenizer_config
def check_model_card(repo_id: str) -> tuple[bool, str]:
"""Checks if the model card and license exist and have been filled"""
try:
card = ModelCard.load(repo_id)
except huggingface_hub.utils.EntryNotFoundError:
return False, "Please add a model card to your model to explain how you trained/fine-tuned it."
# Enforce license metadata
if card.data.license is None:
if not ("license_name" in card.data and "license_link" in card.data):
return False, (
"License not found. Please add a license to your model card using the `license` metadata or a"
" `license_name`/`license_link` pair."
)
# Enforce card content
if len(card.text) < 200:
return False, "Please add a description to your model card, it is too short."
return True, ""
def is_model_on_hub(model_name: str, revision: str, token: str = None, trust_remote_code=False, test_tokenizer=False) -> tuple[bool, str]:
"""Makes sure the model is on the hub, and uses a valid configuration (in the latest transformers version)"""
try:
config = AutoConfig.from_pretrained(model_name, revision=revision, trust_remote_code=trust_remote_code, token=token)
if test_tokenizer:
tokenizer_config = get_tokenizer_config(model_name)
if tokenizer_config is not None:
tokenizer_class_candidate = tokenizer_config.get("tokenizer_class", None)
else:
tokenizer_class_candidate = config.tokenizer_class
tokenizer_class = tokenizer_class_from_name(tokenizer_class_candidate)
if tokenizer_class is None:
return (
False,
f"uses {tokenizer_class_candidate}, which is not in a transformers release, therefore not supported at the moment.",
None
)
return True, None, config
except ValueError:
return (
False,
"needs to be launched with `trust_remote_code=True`. For safety reason, we do not allow these models to be automatically submitted to the leaderboard.",
None
)
except Exception as e:
return False, "was not found on hub!", None
def get_model_size(model_info: ModelInfo, precision: str):
"""Gets the model size from the configuration, or the model name if the configuration does not contain the information."""
try:
model_size = round(model_info.safetensors["total"] / 1e9, 3)
except (AttributeError, TypeError):
return 0 # Unknown model sizes are indicated as 0, see NUMERIC_INTERVALS in app.py
size_factor = 8 if (precision == "GPTQ" or "gptq" in model_info.modelId.lower()) else 1
model_size = size_factor * model_size
return model_size
def get_model_arch(model_info: ModelInfo):
"""Gets the model architecture from the configuration"""
return model_info.config.get("architectures", "Unknown")
def already_submitted_models(requested_models_dir: str) -> set[str]:
depth = 1
file_names = []
users_to_submission_dates = defaultdict(list)
for root, _, files in os.walk(requested_models_dir):
current_depth = root.count(os.sep) - requested_models_dir.count(os.sep)
if current_depth == depth:
for file in files:
if not file.endswith(".json"):
continue
with open(os.path.join(root, file), "r") as f:
info = json.load(f)
file_names.append(f"{info['model']}_{info['revision']}_{info['precision']}")
# Select organisation
if info["model"].count("/") == 0 or "submitted_time" not in info:
continue
organisation, _ = info["model"].split("/")
users_to_submission_dates[organisation].append(info["submitted_time"])
return set(file_names), users_to_submission_dates
|