File size: 5,026 Bytes
df66f6e
 
2a5f9fb
657d04f
df66f6e
657d04f
2a5f9fb
 
df66f6e
 
 
2a5f9fb
 
657d04f
 
 
 
 
976f398
 
2a5f9fb
ec84a57
2a5f9fb
657d04f
 
2a5f9fb
657d04f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
976f398
 
 
 
 
9d22eee
 
 
 
 
976f398
2a5f9fb
 
 
 
 
 
 
 
 
 
 
 
ec84a57
 
 
2a5f9fb
 
 
 
7302987
2a5f9fb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
976f398
 
 
 
2a5f9fb
 
 
9833cdb
2a5f9fb
 
 
 
 
 
 
 
 
 
 
 
 
 
 
657d04f
2a5f9fb
 
 
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
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
import json
import os
from datetime import datetime, timezone
import numpy as np
from src.display.formatting import styled_error, styled_message, styled_warning
from src.envs import API, EVAL_REQUESTS_PATH, TOKEN, QUEUE_REPO, RESULTS_REPO
from src.submission.check_validity import (
    already_submitted_models,
    check_model_card,
    get_model_size,
    is_model_on_hub,
)

from src.display.utils import (
    BENCHMARK_COLS,
    COLS
)

REQUESTED_MODELS = None
USERS_TO_SUBMISSION_DATES = None


def add_new_eval(
    eval_name: str,
    upload: object
):
    with open(upload, mode="r") as f:
        data = json.load(f)    
    
    results = data['results']

    ret = {"eval_name": eval_name}
    # TODO add complex validation
    print(results.keys())
    print(BENCHMARK_COLS)
    for input_col in results.keys():
        if input_col not in BENCHMARK_COLS:
            print(input_col)
            return styled_error(f'Missing: {input_col}')
    #ret.update({i:j['acc,none'] for i,j in results.items()})
    ret.update({i:round(np.random.normal(1, 0.5, 1)[0], 2) for i,j in results.items()})
    

    user_name = "davidadamczyk"
    OUT_DIR = f"{EVAL_REQUESTS_PATH}/{user_name}"

    out_path = f"{OUT_DIR}/{eval_name}_eval_request.json"

    with open(out_path, "w") as f:
        f.write(json.dumps(ret))


    print("Uploading eval file")

    print("path_or_fileobj: ", out_path)
    print("path_in_repo: ",out_path.split("eval-queue/")[1])
    print("repo_id: ", RESULTS_REPO)
    print("repo_type: ", "dataset")

    API.upload_file(
        path_or_fileobj=out_path,
        path_in_repo=out_path.split("eval-queue/")[1],
        repo_id=RESULTS_REPO,
        repo_type="dataset",
        commit_message=f"Add {eval_name} to eval queue",
    )

    """
    global REQUESTED_MODELS
    global USERS_TO_SUBMISSION_DATES
    if not REQUESTED_MODELS:
        REQUESTED_MODELS, USERS_TO_SUBMISSION_DATES = already_submitted_models(EVAL_REQUESTS_PATH)

    user_name = ""
    model_path = model
    if "/" in model:
        user_name = model.split("/")[0]
        model_path = model.split("/")[1]

    precision = precision.split(" ")[0]
    current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")

    if model_type is None or model_type == "":
        return styled_error("Please select a model type.")

    # Does the model actually exist?
    if revision == "":
        revision = "main"

    # Is the model on the hub?
    if weight_type in ["Delta", "Adapter"]:
        base_model_on_hub, error, _ = is_model_on_hub(
            model_name=base_model, revision=revision, token=TOKEN, test_tokenizer=True
        )
        if not base_model_on_hub:
            return styled_error(f'Base model "{base_model}" {error}')

    if not weight_type == "Adapter":
        model_on_hub, error, _ = is_model_on_hub(model_name=model, revision=revision, test_tokenizer=True)
        if not model_on_hub:
            return styled_error(f'Model "{model}" {error}')

    # Is the model info correctly filled?
    try:
        model_info = API.model_info(repo_id=model, revision=revision)
    except Exception:
        return styled_error("Could not get your model information. Please fill it up properly.")

    model_size = get_model_size(model_info=model_info, precision=precision)

    # Were the model card and license filled?
    try:
        license = model_info.cardData["license"]
    except Exception:
        return styled_error("Please select a license for your model")

    modelcard_OK, error_msg = check_model_card(model)
    if not modelcard_OK:
        return styled_error(error_msg)

    # Seems good, creating the eval
    print("Adding new eval")

    eval_entry = {
        "model": model,
        "base_model": base_model,
        "revision": revision,
        "precision": precision,
        "weight_type": weight_type,
        "status": "PENDING",
        "submitted_time": current_time,
        "model_type": model_type,
        "likes": model_info.likes,
        "params": model_size,
        "license": license,
    }

    # Check for duplicate submission
    if f"{model}_{revision}_{precision}" in REQUESTED_MODELS:
        return styled_warning("This model has been already submitted.")

    print("Creating eval file")
    OUT_DIR = f"{EVAL_REQUESTS_PATH}/{user_name}"
    os.makedirs(OUT_DIR, exist_ok=True)
    out_path = f"{OUT_DIR}/{model_path}_eval_request_False_{precision}_{weight_type}.json"

    with open(out_path, "w") as f:
        f.write(json.dumps(eval_entry))

    print("Uploading eval file")
    API.upload_file(
        path_or_fileobj=out_path,
        path_in_repo=out_path.split("eval-queue/")[1],
        repo_id=QUEUE_REPO,
        repo_type="dataset",
        commit_message=f"Add {model} to eval queue",
    )

    # Remove the local file
    os.remove(out_path)
    """
    return styled_message(
        "Your request has been submitted to the evaluation queue!\nPlease wait for up to an hour for the model to show in the PENDING list."
    )