File size: 5,502 Bytes
101093d
 
 
 
 
 
 
 
 
 
 
3304383
101093d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3304383
101093d
3304383
101093d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3304383
 
101093d
 
 
 
 
3304383
101093d
3304383
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
101093d
3304383
dea3205
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
import sys
import argparse
from traceback import print_exc
import pickle
import tqdm
import pandas as pd
from concurrent.futures import ProcessPoolExecutor, ThreadPoolExecutor

import torch
import bittensor
from meta_utils import load_metagraphs
#TODO: make line charts and other cool stuff for each metagraph snapshot

def process(block, netuid=1, lite=True, difficulty=False, prune_weights=False, return_graph=False, half=True, subtensor=None):

    if subtensor is None:
        subtensor = bittensor.subtensor(network='finney')        
    
    try:        
        metagraph = subtensor.metagraph(block=block, netuid=netuid, lite=lite)
        if difficulty:
            metagraph.difficulty = subtensor.difficulty(block=block, netuid=netuid)
        
        if not lite:
            if half:
                metagraph.weights = torch.nn.Parameter(metagraph.weights.half(), requires_grad=False)
            if prune_weights:
                metagraph.weights = metagraph.weights[metagraph.weights.sum(axis=1) > 0]

        with open(f'data/metagraph/{netuid}/{block}.pkl', 'wb') as f:
            pickle.dump(metagraph, f)

        return metagraph if return_graph else True
    
    except Exception as e:
        print(f'Error processing block {block}: {e}')
        

def parse_arguments():
    parser = argparse.ArgumentParser(description='Process metagraphs for a given network.')
    parser.add_argument('--netuid', type=int, default=1, help='Network UID to use.')
    parser.add_argument('--difficulty', action='store_true', help='Include difficulty in metagraph.')
    parser.add_argument('--prune_weights', action='store_true', help='Prune weights in metagraph.')
    parser.add_argument('--return_graph', action='store_true', help='Return metagraph instead of True.')
    parser.add_argument('--no_dataframe', action='store_true', help='Do not create dataframe.')
    parser.add_argument('--max_workers', type=int, default=32, help='Max workers to use.')
    parser.add_argument('--start_block', type=int, default=1_500_000, help='Start block.')
    parser.add_argument('--end_block', type=int, default=600_000, help='End block.')
    parser.add_argument('--step_size', type=int, default=100, help='Step size.')
    return parser.parse_args()

if __name__ == '__main__':

    subtensor = bittensor.subtensor(network='finney')
    print(f'Current block: {subtensor.block}')
    
    args = parse_arguments()
       
    netuid=args.netuid
    difficulty=args.difficulty
    overwrite=False
    return_graph=args.return_graph       
     
    step_size = args.step_size
    start_block = args.start_block
    start_block = (min(subtensor.block, start_block)//step_size)*step_size # round to nearest step_size
    end_block = args.end_block
    blocks = range(start_block, end_block, -step_size)

    # only get weights for multiple of 500 blocks
    lite=lambda x: x%500!=0
    
    max_workers = min(args.max_workers, len(blocks))

    datadir = f'data/metagraph/{netuid}'
    os.makedirs(datadir, exist_ok=True)
    if not overwrite:
        blocks = [block for block in blocks if not os.path.exists(f'data/metagraph/{netuid}/{block}.pkl')]

    metagraphs = []
    
    if len(blocks)>0:
        
        print(f'Processing {len(blocks)} blocks from {blocks[0]}-{blocks[-1]} using {max_workers} workers.')
        
        with ProcessPoolExecutor(max_workers=max_workers) as executor:
            futures = [
                executor.submit(process, block, lite=lite(block), netuid=netuid, difficulty=difficulty) 
                for block in blocks
                ]

            success = 0        
            with tqdm.tqdm(total=len(futures)) as pbar:
                for block, future in zip(blocks,futures):
                    try:
                        metagraphs.append(future.result())
                        success += 1
                    except Exception as e:
                        print(f'generated an exception: {print_exc(e)}')
                    pbar.update(1)
                    pbar.set_description(f'Processed {success} blocks. Current block: {block}')

        if not success:
            raise ValueError('No blocks were successfully processed.')
        
        print(f'Processed {success} blocks.')
        if return_graph:
            for metagraph in metagraphs:
                print(f'{metagraph.block}: {metagraph.n.item()} nodes, difficulty={getattr(metagraph, "difficulty", None)}, weights={metagraph.weights.shape if hasattr(metagraph, "weights") else None}')

        print(metagraphs[-1])
    else:
        print(f'No blocks to process. Current block: {subtensor.block}')


    if not args.no_dataframe:
        save_path = f'data/metagraph/{netuid}/df.parquet'
        blocks = range(start_block, end_block, step_size)
        df_loaded = None
        if os.path.exists(save_path):
            df_loaded = pd.read_parquet(save_path)
            blocks = [block for block in blocks if block not in df_loaded.block.unique()]
            print(f'Loaded dataframe from {save_path!r}. {len(df_loaded)} rows. {len(blocks)} blocks to process.')
            if len(blocks)==0:
                print('No blocks to process.')
                sys.exit(0)
            
        df = load_metagraphs(blocks[0], blocks[-1], block_step=step_size, datadir=datadir)
        if df_loaded is not None:
            df = pd.concat([df, df_loaded], ignore_index=True)
        df.to_parquet(save_path)
        print(f'Saved dataframe to {save_path!r}')