|
from dataclasses import dataclass, make_dataclass |
|
from enum import Enum |
|
|
|
import pandas as pd |
|
|
|
from src.about import Tasks, nc_tasks, nr_tasks, lp_tasks |
|
|
|
def fields(raw_class): |
|
return [v for k, v in raw_class.__dict__.items() if k[:2] != "__" and k[-2:] != "__"] |
|
|
|
|
|
|
|
|
|
|
|
@dataclass |
|
class ColumnContent: |
|
name: str |
|
type: str |
|
displayed_by_default: bool |
|
hidden: bool = False |
|
never_hidden: bool = False |
|
|
|
|
|
auto_eval_column_dict = [] |
|
|
|
auto_eval_column_dict.append(["model_type_symbol", ColumnContent, ColumnContent("T", "str", True, never_hidden=True)]) |
|
auto_eval_column_dict.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)]) |
|
|
|
auto_eval_column_dict.append(["average", ColumnContent, ColumnContent("Average ⬆️", "number", True)]) |
|
for task in Tasks: |
|
auto_eval_column_dict.append([task.name, ColumnContent, ColumnContent(task.value.col_name, "number", True)]) |
|
|
|
auto_eval_column_dict.append(["model_type", ColumnContent, ColumnContent("Type", "str", False)]) |
|
auto_eval_column_dict.append(["architecture", ColumnContent, ColumnContent("Architecture", "str", False)]) |
|
auto_eval_column_dict.append(["weight_type", ColumnContent, ColumnContent("Weight type", "str", False, True)]) |
|
auto_eval_column_dict.append(["precision", ColumnContent, ColumnContent("Precision", "str", False)]) |
|
auto_eval_column_dict.append(["license", ColumnContent, ColumnContent("Hub License", "str", False)]) |
|
auto_eval_column_dict.append(["params", ColumnContent, ColumnContent("#Params (B)", "number", False)]) |
|
auto_eval_column_dict.append(["likes", ColumnContent, ColumnContent("Hub ❤️", "number", False)]) |
|
auto_eval_column_dict.append(["still_on_hub", ColumnContent, ColumnContent("Available on the hub", "bool", False)]) |
|
auto_eval_column_dict.append(["revision", ColumnContent, ColumnContent("Model sha", "str", False, False)]) |
|
|
|
|
|
AutoEvalColumn = make_dataclass("AutoEvalColumn", auto_eval_column_dict, frozen=True) |
|
|
|
|
|
auto_eval_column_dict_nc = [] |
|
auto_eval_column_dict_nc.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)]) |
|
auto_eval_column_dict_nc.append(["average_rank", ColumnContent, ColumnContent("Average Rank⬆️", "number", True)]) |
|
for task in nc_tasks: |
|
auto_eval_column_dict_nc.append(['_'.join(task.value.col_name.split('-')), ColumnContent, ColumnContent(task.value.benchmark, "number", True)]) |
|
auto_eval_column_dict_nc.append(["author", ColumnContent, ColumnContent("Author", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_nc.append(["email", ColumnContent, ColumnContent("Email", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_nc.append(["Paper_URL", ColumnContent, ColumnContent("Paper URL", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_nc.append(["Github_URL", ColumnContent, ColumnContent("Github URL", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_nc.append(["Time", ColumnContent, ColumnContent("Time", "markdown", True, never_hidden=False)]) |
|
|
|
|
|
AutoEvalColumn_NodeClassification = make_dataclass("AutoEvalColumn_NodeClassification", auto_eval_column_dict_nc, frozen=True) |
|
|
|
|
|
auto_eval_column_dict_nr = [] |
|
auto_eval_column_dict_nr.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)]) |
|
auto_eval_column_dict_nr.append(["average_rank", ColumnContent, ColumnContent("Average Rank⬆️", "number", True)]) |
|
for task in nr_tasks: |
|
auto_eval_column_dict_nr.append(['_'.join(task.value.col_name.split('-')), ColumnContent, ColumnContent(task.value.benchmark, "number", True)]) |
|
auto_eval_column_dict_nr.append(["author", ColumnContent, ColumnContent("Author", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_nr.append(["email", ColumnContent, ColumnContent("Email", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_nr.append(["Paper_URL", ColumnContent, ColumnContent("Paper URL", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_nr.append(["Github_URL", ColumnContent, ColumnContent("Github URL", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_nr.append(["Time", ColumnContent, ColumnContent("Time", "markdown", True, never_hidden=False)]) |
|
|
|
|
|
AutoEvalColumn_NodeRegression = make_dataclass("AutoEvalColumn_NodeRegression", auto_eval_column_dict_nr, frozen=True) |
|
|
|
|
|
auto_eval_column_dict_lp = [] |
|
auto_eval_column_dict_lp.append(["model", ColumnContent, ColumnContent("Model", "markdown", True, never_hidden=True)]) |
|
auto_eval_column_dict_lp.append(["average_rank", ColumnContent, ColumnContent("Average Rank⬆️", "number", True)]) |
|
for task in lp_tasks: |
|
auto_eval_column_dict_lp.append(['_'.join(task.value.col_name.split('-')), ColumnContent, ColumnContent(task.value.benchmark, "number", True)]) |
|
auto_eval_column_dict_lp.append(["author", ColumnContent, ColumnContent("Author", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_lp.append(["email", ColumnContent, ColumnContent("Email", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_lp.append(["Paper_URL", ColumnContent, ColumnContent("Paper URL", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_lp.append(["Github_URL", ColumnContent, ColumnContent("Github URL", "markdown", True, never_hidden=False)]) |
|
auto_eval_column_dict_lp.append(["Time", ColumnContent, ColumnContent("Time", "markdown", True, never_hidden=False)]) |
|
|
|
|
|
AutoEvalColumn_LinkPrediction = make_dataclass("AutoEvalColumn_LinkPrediction", auto_eval_column_dict_lp, frozen=True) |
|
|
|
|
|
|
|
@dataclass(frozen=True) |
|
class EvalQueueColumn: |
|
model = ColumnContent("model", "markdown", True) |
|
revision = ColumnContent("revision", "str", True) |
|
private = ColumnContent("private", "bool", True) |
|
precision = ColumnContent("precision", "str", True) |
|
weight_type = ColumnContent("weight_type", "str", "Original") |
|
status = ColumnContent("status", "str", True) |
|
|
|
|
|
@dataclass |
|
class ModelDetails: |
|
name: str |
|
display_name: str = "" |
|
symbol: str = "" |
|
|
|
|
|
class ModelType(Enum): |
|
PT = ModelDetails(name="pretrained", symbol="🟢") |
|
FT = ModelDetails(name="fine-tuned", symbol="🔶") |
|
IFT = ModelDetails(name="instruction-tuned", symbol="⭕") |
|
RL = ModelDetails(name="RL-tuned", symbol="🟦") |
|
Unknown = ModelDetails(name="", symbol="?") |
|
|
|
def to_str(self, separator=" "): |
|
return f"{self.value.symbol}{separator}{self.value.name}" |
|
|
|
@staticmethod |
|
def from_str(type): |
|
if "fine-tuned" in type or "🔶" in type: |
|
return ModelType.FT |
|
if "pretrained" in type or "🟢" in type: |
|
return ModelType.PT |
|
if "RL-tuned" in type or "🟦" in type: |
|
return ModelType.RL |
|
if "instruction-tuned" in type or "⭕" in type: |
|
return ModelType.IFT |
|
return ModelType.Unknown |
|
|
|
class OFFICIAL(Enum): |
|
official = ModelDetails("Official") |
|
unofficial = ModelDetails("Unofficial") |
|
|
|
class HONOR(Enum): |
|
yes = ModelDetails("Yes") |
|
no = ModelDetails("No") |
|
|
|
class TASK_LIST(Enum): |
|
amazon_user_churn = ModelDetails("rel-amazon-user-churn") |
|
amazon_item_churn = ModelDetails("rel-amazon-item-churn") |
|
amazon_user_ltv = ModelDetails("rel-amazon-user-ltv") |
|
amazon_item_ltv = ModelDetails("rel-amazon-item-ltv") |
|
amazon_user_item_purchase = ModelDetails("rel-amazon-user-item-purchase") |
|
amazon_user_item_rate = ModelDetails("rel-amazon-user-item-rate") |
|
amazon_user_item_review = ModelDetails("rel-amazon-user-item-review") |
|
|
|
|
|
stack_user_engagement = ModelDetails("rel-stack-user-engagement") |
|
stack_user_badge = ModelDetails("rel-stack-user-badge") |
|
stack_post_votes = ModelDetails("rel-stack-post-votes") |
|
stack_user_post_comment = ModelDetails("rel-stack-user-post-comment") |
|
stack_user_post_related = ModelDetails("rel-stack-user-post-related") |
|
|
|
|
|
trial_study_outcome = ModelDetails("rel-trial-study-outcome") |
|
trial_study_adverse = ModelDetails("rel-trial-study-adverse") |
|
trial_site_success = ModelDetails("rel-trial-site-success") |
|
trial_condition_sponsor_run = ModelDetails("rel-trial-condition-sponsor-run") |
|
trial_site_sponsor_run = ModelDetails("rel-trial-site-sponsor-run") |
|
|
|
|
|
f1_driver_position = ModelDetails("rel-f1-driver-position") |
|
f1_driver_dnf = ModelDetails("rel-f1-driver-dnf") |
|
f1_driver_top3 = ModelDetails("rel-f1-driver-top3") |
|
|
|
|
|
hm_user_churn = ModelDetails("rel-hm-user-churn") |
|
hm_item_sales = ModelDetails("rel-hm-item-sales") |
|
hm_user_item_purchase = ModelDetails("rel-hm-user-item-purchase") |
|
|
|
|
|
event_user_repeat = ModelDetails("rel-event-user-repeat") |
|
event_user_ignore = ModelDetails("rel-event-user-ignore") |
|
event_user_attendance = ModelDetails("rel-event-user-attendance") |
|
|
|
|
|
avito_user_visits = ModelDetails("rel-avito-user-visits") |
|
avito_user_clicks = ModelDetails("rel-avito-user-clicks") |
|
avito_ads_clicks = ModelDetails("rel-avito-ads-clicks") |
|
avito_user_ad_visit = ModelDetails("rel-avito-user-ad-visit") |
|
|
|
class WeightType(Enum): |
|
Adapter = ModelDetails("Adapter") |
|
Original = ModelDetails("Original") |
|
Delta = ModelDetails("Delta") |
|
|
|
class Precision(Enum): |
|
float16 = ModelDetails("float16") |
|
bfloat16 = ModelDetails("bfloat16") |
|
float32 = ModelDetails("float32") |
|
|
|
|
|
|
|
Unknown = ModelDetails("?") |
|
|
|
def from_str(precision): |
|
if precision in ["torch.float16", "float16"]: |
|
return Precision.float16 |
|
if precision in ["torch.bfloat16", "bfloat16"]: |
|
return Precision.bfloat16 |
|
if precision in ["float32"]: |
|
return Precision.float32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
return Precision.Unknown |
|
|
|
|
|
COLS = [c.name for c in fields(AutoEvalColumn) if not c.hidden] |
|
COLS_NC = [c.name for c in fields(AutoEvalColumn_NodeClassification) if not c.hidden] |
|
COLS_NR = [c.name for c in fields(AutoEvalColumn_NodeRegression) if not c.hidden] |
|
COLS_LP = [c.name for c in fields(AutoEvalColumn_LinkPrediction) if not c.hidden] |
|
|
|
TYPES = [c.type for c in fields(AutoEvalColumn) if not c.hidden] |
|
COLS_LITE = [c.name for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.hidden] |
|
TYPES_LITE = [c.type for c in fields(AutoEvalColumn) if c.displayed_by_default and not c.hidden] |
|
|
|
EVAL_COLS = [c.name for c in fields(EvalQueueColumn)] |
|
EVAL_TYPES = [c.type for c in fields(EvalQueueColumn)] |
|
|
|
BENCHMARK_COLS = [t.value.col_name for t in Tasks] |
|
|
|
NUMERIC_INTERVALS = { |
|
"?": pd.Interval(-1, 0, closed="right"), |
|
"~1.5": pd.Interval(0, 2, closed="right"), |
|
"~3": pd.Interval(2, 4, closed="right"), |
|
"~7": pd.Interval(4, 9, closed="right"), |
|
"~13": pd.Interval(9, 20, closed="right"), |
|
"~35": pd.Interval(20, 45, closed="right"), |
|
"~60": pd.Interval(45, 70, closed="right"), |
|
"70+": pd.Interval(70, 10000, closed="right"), |
|
} |
|
|