repo_name stringlengths 8 38 | pr_number int64 3 47.1k | pr_title stringlengths 8 175 | pr_description stringlengths 2 19.8k ⌀ | author null | date_created stringlengths 25 25 | date_merged stringlengths 25 25 | filepath stringlengths 6 136 | before_content stringlengths 54 884k ⌀ | after_content stringlengths 56 884k | pr_author stringlengths 3 21 | previous_commit stringlengths 40 40 | pr_commit stringlengths 40 40 | comment stringlengths 2 25.4k | comment_author stringlengths 3 29 | __index_level_0__ int64 0 5.1k |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
py-why/dowhy | 478 | Adding Non Linear Sensitivity Analysis | This PR implements the non-parametric sensitivity analysis from Chernozhukov et al. https://arxiv.org/abs/2112.13398
It implements two sensitivity analyzers:
1. For Partial Linear DGPs and estimators like LinearDML
2. For general non-parametric DGPs and estimators like KernelDML.
The notebook in this PR provid... | null | 2022-06-20 14:37:11+00:00 | 2022-09-16 03:57:26+00:00 | dowhy/causal_refuters/linear_sensitivity_analyzer.py | import logging
import sys
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import statsmodels.api as sm
from scipy.stats import t
from dowhy.utils.api import parse_state
class LinearSensitivityAnalyzer:
"""
Class to perform sensitivity analysis
See: https://carloscinelli.com/files/... | import logging
import sys
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import statsmodels.api as sm
from scipy.stats import t
from dowhy.utils.api import parse_state
class LinearSensitivityAnalyzer:
"""
Class to perform sensitivity analysis
See: https://carloscinelli.com/files/... | anusha0409 | 81841c697bd5e80ecf9e731432305f6186666f1f | bb446c333f2256074304b0dec9cb5628d284b542 | yes, this is to fix a bug in the prior code. | amit-sharma | 384 |
mdn/kuma | 8,071 | user with is_staff=True should be subscribers | null | null | 2022-04-06 11:04:23+00:00 | 2022-04-07 13:25:10+00:00 | kuma/users/tasks.py | import json
from celery import task
from django.contrib.auth import get_user_model
from kuma.users.auth import KumaOIDCAuthenticationBackend
from kuma.users.models import AccountEvent, UserProfile
from kuma.users.utils import get_valid_subscription_type_or_none
@task
def process_event_delete_user(event_id):
eve... | import json
from celery import task
from django.contrib.auth import get_user_model
from kuma.users.auth import KumaOIDCAuthenticationBackend
from kuma.users.models import AccountEvent, UserProfile
from kuma.users.utils import get_valid_subscription_type_or_none
@task
def process_event_delete_user(event_id):
eve... | fiji-flo | 57285dcf43694852d10e68159136c39a3e63cb29 | e1cef9d866060531069a0a73405f42b012c61627 | Don't update capabilities for `is_staff`. (We need to clean this manually). | fiji-flo | 0 |
mdn/kuma | 8,071 | user with is_staff=True should be subscribers | null | null | 2022-04-06 11:04:23+00:00 | 2022-04-07 13:25:10+00:00 | kuma/users/tasks.py | import json
from celery import task
from django.contrib.auth import get_user_model
from kuma.users.auth import KumaOIDCAuthenticationBackend
from kuma.users.models import AccountEvent, UserProfile
from kuma.users.utils import get_valid_subscription_type_or_none
@task
def process_event_delete_user(event_id):
eve... | import json
from celery import task
from django.contrib.auth import get_user_model
from kuma.users.auth import KumaOIDCAuthenticationBackend
from kuma.users.models import AccountEvent, UserProfile
from kuma.users.utils import get_valid_subscription_type_or_none
@task
def process_event_delete_user(event_id):
eve... | fiji-flo | 57285dcf43694852d10e68159136c39a3e63cb29 | e1cef9d866060531069a0a73405f42b012c61627 | Don't update capabilities for `is_staff`. (We need to clean this manually). | fiji-flo | 1 |
mdn/kuma | 8,070 | feat(notifications): process Content Updates via /update/ route | Part of https://github.com/mdn/yari-private/issues/981.
## Before
1. `/update/` created only BCD Update Notifications.
2. `/create/pr/` created a single Content Update Notification independently.
## After
1. `/update/` creates both BCD **and Content** Update Notifications.
2. `/create/pr/` creates a singl... | null | 2022-04-05 17:20:14+00:00 | 2022-04-07 13:35:40+00:00 | kuma/api/v1/plus/notifications.py | from __future__ import annotations
import datetime
import json
from typing import Optional
# import requests
import requests
from django.conf import settings
from django.db.models import Q
from django.middleware.csrf import get_token
from ninja import Field, Router
from ninja.pagination import paginate
from kuma.doc... | from __future__ import annotations
import datetime
import json
from typing import Optional
# import requests
import requests
from django.conf import settings
from django.db.models import Q
from django.middleware.csrf import get_token
from ninja import Field, Router
from ninja.pagination import paginate
from kuma.doc... | caugner | e1cef9d866060531069a0a73405f42b012c61627 | a5247595160b84987621c7ef899b454e370db188 | Nice change. | Guyzeroth | 2 |
mdn/kuma | 8,070 | feat(notifications): process Content Updates via /update/ route | Part of https://github.com/mdn/yari-private/issues/981.
## Before
1. `/update/` created only BCD Update Notifications.
2. `/create/pr/` created a single Content Update Notification independently.
## After
1. `/update/` creates both BCD **and Content** Update Notifications.
2. `/create/pr/` creates a singl... | null | 2022-04-05 17:20:14+00:00 | 2022-04-07 13:35:40+00:00 | kuma/api/v1/plus/notifications.py | from __future__ import annotations
import datetime
import json
from typing import Optional
# import requests
import requests
from django.conf import settings
from django.db.models import Q
from django.middleware.csrf import get_token
from ninja import Field, Router
from ninja.pagination import paginate
from kuma.doc... | from __future__ import annotations
import datetime
import json
from typing import Optional
# import requests
import requests
from django.conf import settings
from django.db.models import Q
from django.middleware.csrf import get_token
from ninja import Field, Router
from ninja.pagination import paginate
from kuma.doc... | caugner | e1cef9d866060531069a0a73405f42b012c61627 | a5247595160b84987621c7ef899b454e370db188 | Could this get a test? | Guyzeroth | 3 |
mdn/kuma | 8,070 | feat(notifications): process Content Updates via /update/ route | Part of https://github.com/mdn/yari-private/issues/981.
## Before
1. `/update/` created only BCD Update Notifications.
2. `/create/pr/` created a single Content Update Notification independently.
## After
1. `/update/` creates both BCD **and Content** Update Notifications.
2. `/create/pr/` creates a singl... | null | 2022-04-05 17:20:14+00:00 | 2022-04-07 13:35:40+00:00 | kuma/api/v1/plus/notifications.py | from __future__ import annotations
import datetime
import json
from typing import Optional
# import requests
import requests
from django.conf import settings
from django.db.models import Q
from django.middleware.csrf import get_token
from ninja import Field, Router
from ninja.pagination import paginate
from kuma.doc... | from __future__ import annotations
import datetime
import json
from typing import Optional
# import requests
import requests
from django.conf import settings
from django.db.models import Q
from django.middleware.csrf import get_token
from ninja import Field, Router
from ninja.pagination import paginate
from kuma.doc... | caugner | e1cef9d866060531069a0a73405f42b012c61627 | a5247595160b84987621c7ef899b454e370db188 | Added in 8f35a2f40f3942051bccdc6180eac06b6bed6dcf, thanks for helping me out with this. | caugner | 4 |
mdn/kuma | 8,070 | feat(notifications): process Content Updates via /update/ route | Part of https://github.com/mdn/yari-private/issues/981.
## Before
1. `/update/` created only BCD Update Notifications.
2. `/create/pr/` created a single Content Update Notification independently.
## After
1. `/update/` creates both BCD **and Content** Update Notifications.
2. `/create/pr/` creates a singl... | null | 2022-04-05 17:20:14+00:00 | 2022-04-07 13:35:40+00:00 | kuma/api/v1/plus/notifications.py | from __future__ import annotations
import datetime
import json
from typing import Optional
# import requests
import requests
from django.conf import settings
from django.db.models import Q
from django.middleware.csrf import get_token
from ninja import Field, Router
from ninja.pagination import paginate
from kuma.doc... | from __future__ import annotations
import datetime
import json
from typing import Optional
# import requests
import requests
from django.conf import settings
from django.db.models import Q
from django.middleware.csrf import get_token
from ninja import Field, Router
from ninja.pagination import paginate
from kuma.doc... | caugner | e1cef9d866060531069a0a73405f42b012c61627 | a5247595160b84987621c7ef899b454e370db188 | ```suggestion
def update_content(request, body: ContentUpdateNotificationSchema):
``` | caugner | 5 |
mdn/kuma | 8,070 | feat(notifications): process Content Updates via /update/ route | Part of https://github.com/mdn/yari-private/issues/981.
## Before
1. `/update/` created only BCD Update Notifications.
2. `/create/pr/` created a single Content Update Notification independently.
## After
1. `/update/` creates both BCD **and Content** Update Notifications.
2. `/create/pr/` creates a singl... | null | 2022-04-05 17:20:14+00:00 | 2022-04-07 13:35:40+00:00 | kuma/notifications/utils.py | from collections import defaultdict
from kuma.notifications.browsers import browsers
from kuma.notifications.models import Notification, NotificationData, Watch
def publish_notification(path, text, dry_run=False, data=None):
# This traverses down the path to see if there's top level watchers
parts = path.spl... | import re
from collections import defaultdict
from kuma.documenturls.models import DocumentURL
from kuma.notifications.browsers import browsers
from kuma.notifications.models import Notification, NotificationData, Watch
def publish_bcd_notification(path, text, data=None):
# This traverses down the path to see if... | caugner | e1cef9d866060531069a0a73405f42b012c61627 | a5247595160b84987621c7ef899b454e370db188 | Is dry_run something we are interested in maintaining? | Guyzeroth | 6 |
mdn/kuma | 8,070 | feat(notifications): process Content Updates via /update/ route | Part of https://github.com/mdn/yari-private/issues/981.
## Before
1. `/update/` created only BCD Update Notifications.
2. `/create/pr/` created a single Content Update Notification independently.
## After
1. `/update/` creates both BCD **and Content** Update Notifications.
2. `/create/pr/` creates a singl... | null | 2022-04-05 17:20:14+00:00 | 2022-04-07 13:35:40+00:00 | kuma/notifications/utils.py | from collections import defaultdict
from kuma.notifications.browsers import browsers
from kuma.notifications.models import Notification, NotificationData, Watch
def publish_notification(path, text, dry_run=False, data=None):
# This traverses down the path to see if there's top level watchers
parts = path.spl... | import re
from collections import defaultdict
from kuma.documenturls.models import DocumentURL
from kuma.notifications.browsers import browsers
from kuma.notifications.models import Notification, NotificationData, Watch
def publish_bcd_notification(path, text, data=None):
# This traverses down the path to see if... | caugner | e1cef9d866060531069a0a73405f42b012c61627 | a5247595160b84987621c7ef899b454e370db188 | +1 for removing that. | fiji-flo | 7 |
mdn/kuma | 8,070 | feat(notifications): process Content Updates via /update/ route | Part of https://github.com/mdn/yari-private/issues/981.
## Before
1. `/update/` created only BCD Update Notifications.
2. `/create/pr/` created a single Content Update Notification independently.
## After
1. `/update/` creates both BCD **and Content** Update Notifications.
2. `/create/pr/` creates a singl... | null | 2022-04-05 17:20:14+00:00 | 2022-04-07 13:35:40+00:00 | kuma/notifications/utils.py | from collections import defaultdict
from kuma.notifications.browsers import browsers
from kuma.notifications.models import Notification, NotificationData, Watch
def publish_notification(path, text, dry_run=False, data=None):
# This traverses down the path to see if there's top level watchers
parts = path.spl... | import re
from collections import defaultdict
from kuma.documenturls.models import DocumentURL
from kuma.notifications.browsers import browsers
from kuma.notifications.models import Notification, NotificationData, Watch
def publish_bcd_notification(path, text, data=None):
# This traverses down the path to see if... | caugner | e1cef9d866060531069a0a73405f42b012c61627 | a5247595160b84987621c7ef899b454e370db188 | Fixed in 049a72725. | caugner | 8 |
mdn/kuma | 8,026 | Renew access token every 12 hours. | null | null | 2022-01-12 11:37:13+00:00 | 2022-01-12 12:50:08+00:00 | kuma/users/auth.py | import time
from django.conf import settings
from django.contrib.auth import get_user_model
from mozilla_django_oidc.auth import OIDCAuthenticationBackend
from .models import UserProfile
class KumaOIDCAuthenticationBackend(OIDCAuthenticationBackend):
"""Extend mozilla-django-oidc authbackend."""
def __init... | import time
import requests
from django.conf import settings
from django.contrib.auth import get_user_model
from mozilla_django_oidc.auth import OIDCAuthenticationBackend
from .models import UserProfile
class KumaOIDCAuthenticationBackend(OIDCAuthenticationBackend):
"""Extend mozilla-django-oidc authbackend."""... | akatsoulas | 20857c3ed1eef5bcbe2774da660c4b93ca9d9d42 | 0ab25fbff2725d12c9fe2e8c83089236639193d6 | ```suggestion
import time
import requests
from django.conf import settings
``` | fiji-flo | 9 |
mdn/kuma | 8,026 | Renew access token every 12 hours. | null | null | 2022-01-12 11:37:13+00:00 | 2022-01-12 12:50:08+00:00 | kuma/users/middleware.py | import time
import requests
from django.conf import settings
from django.contrib.auth import logout
from django.core.exceptions import MiddlewareNotUsed
from mozilla_django_oidc.middleware import SessionRefresh
class ValidateAccessTokenMiddleware(SessionRefresh):
"""Validate the access token every hour.
Ver... | import time
from django.conf import settings
from django.contrib.auth import logout
from django.core.exceptions import MiddlewareNotUsed
from mozilla_django_oidc.middleware import SessionRefresh
from kuma.users.auth import KumaOIDCAuthenticationBackend
class ValidateAccessTokenMiddleware(SessionRefresh):
"""Val... | akatsoulas | 20857c3ed1eef5bcbe2774da660c4b93ca9d9d42 | 0ab25fbff2725d12c9fe2e8c83089236639193d6 | ```suggestion
token_info = KumaOIDCAuthenticationBackend.refresh_access_token(
profile.fxa_refresh_token
)
new_access_token = token_info.get("access_token")
``` | fiji-flo | 10 |
mdn/kuma | 8,020 | support core users | null | null | 2021-12-15 19:51:40+00:00 | 2021-12-16 17:31:33+00:00 | kuma/api/v1/views.py | from django.conf import settings
from django.http import HttpResponseForbidden, JsonResponse
from django.middleware.csrf import get_token
from django.views.decorators.cache import never_cache
from django.views.decorators.http import require_GET
from kuma.api.v1.forms import AccountSettingsForm
from kuma.users.models i... | from django.conf import settings
from django.http import HttpResponseForbidden, JsonResponse
from django.middleware.csrf import get_token
from django.views.decorators.cache import never_cache
from django.views.decorators.http import require_GET
from kuma.api.v1.forms import AccountSettingsForm
from kuma.users.models i... | fiji-flo | cbda569e9a2cd16b8a6d16bafb721e6dd2769467 | 053c45db24a549999707650a93e3808238521b40 | nit: Since this is getting a value of `True` you could just assign `data['is_subscriber'] = profile.is_subscriber` to remove the `if` clause | akatsoulas | 11 |
mdn/kuma | 8,020 | support core users | null | null | 2021-12-15 19:51:40+00:00 | 2021-12-16 17:31:33+00:00 | kuma/users/auth.py | import time
from django.conf import settings
from django.contrib.auth import get_user_model
from mozilla_django_oidc.auth import OIDCAuthenticationBackend
from .models import UserProfile
class KumaOIDCAuthenticationBackend(OIDCAuthenticationBackend):
"""Extend mozilla-django-oidc authbackend."""
def __init... | import time
from django.conf import settings
from django.contrib.auth import get_user_model
from mozilla_django_oidc.auth import OIDCAuthenticationBackend
from .models import UserProfile
class KumaOIDCAuthenticationBackend(OIDCAuthenticationBackend):
"""Extend mozilla-django-oidc authbackend."""
def __init... | fiji-flo | cbda569e9a2cd16b8a6d16bafb721e6dd2769467 | 053c45db24a549999707650a93e3808238521b40 | super nit: You can omit completely and change the check in the CallbackView to
`if self.request.get("created") and not is_subscriber` | akatsoulas | 12 |
mdn/kuma | 8,020 | support core users | null | null | 2021-12-15 19:51:40+00:00 | 2021-12-16 17:31:33+00:00 | kuma/users/models.py | import json
from django.contrib.auth import get_user_model
from django.db import models
class UserProfile(models.Model):
user = models.OneToOneField(get_user_model(), on_delete=models.CASCADE)
locale = models.CharField(max_length=6, null=True)
created = models.DateTimeField(auto_now_add=True)
modifie... | import json
from django.contrib.auth import get_user_model
from django.db import models
class UserProfile(models.Model):
user = models.OneToOneField(get_user_model(), on_delete=models.CASCADE)
locale = models.CharField(max_length=6, null=True)
created = models.DateTimeField(auto_now_add=True)
modifie... | fiji-flo | cbda569e9a2cd16b8a6d16bafb721e6dd2769467 | 053c45db24a549999707650a93e3808238521b40 | Shouldn't this be `self.is_subscriber`? | akatsoulas | 13 |
mdn/kuma | 8,020 | support core users | null | null | 2021-12-15 19:51:40+00:00 | 2021-12-16 17:31:33+00:00 | kuma/users/models.py | import json
from django.contrib.auth import get_user_model
from django.db import models
class UserProfile(models.Model):
user = models.OneToOneField(get_user_model(), on_delete=models.CASCADE)
locale = models.CharField(max_length=6, null=True)
created = models.DateTimeField(auto_now_add=True)
modifie... | import json
from django.contrib.auth import get_user_model
from django.db import models
class UserProfile(models.Model):
user = models.OneToOneField(get_user_model(), on_delete=models.CASCADE)
locale = models.CharField(max_length=6, null=True)
created = models.DateTimeField(auto_now_add=True)
modifie... | fiji-flo | cbda569e9a2cd16b8a6d16bafb721e6dd2769467 | 053c45db24a549999707650a93e3808238521b40 | good catch but I just pushed it 👍 | fiji-flo | 14 |
dbcli/litecli | 165 | Add key binding to accept completion with the right-arrow key | ## Description
zsh-autosuggestions uses right-arrow to select a suggestion. I found that I end up hitting right-arrow often when using litecli. This change adds a key binding for right-arrow to accept a completion.
## Checklist
<!--- We appreciate your help and want to give you credit. Please take a moment ... | null | 2023-09-27 15:26:47+00:00 | 2023-09-27 15:33:32+00:00 | tests/test_smart_completion_public_schema_only.py | # coding: utf-8
from __future__ import unicode_literals
import pytest
from mock import patch
from prompt_toolkit.completion import Completion
from prompt_toolkit.document import Document
metadata = {
"users": ["id", "email", "first_name", "last_name"],
"orders": ["id", "ordered_date", "status"],
"select": ... | # coding: utf-8
from __future__ import unicode_literals
import pytest
from mock import patch
from prompt_toolkit.completion import Completion
from prompt_toolkit.document import Document
metadata = {
"users": ["id", "email", "first_name", "last_name"],
"orders": ["id", "ordered_date", "status"],
"select": ... | liamhennebury | 5975d2010278fda42aa224be5770113fc15ee28f | fbaf48d9f6d9c33794ca25876f851b6318c402a6 | Pre-existing unit test failure, unrelated to change. | liamhennebury | 0 |
dbcli/litecli | 160 | Fixing startupcommands and successful perpetually set as True | ## Description
### Startup commands
As seen in https://github.com/dbcli/litecli/issues/56 there is a wish for having the option to define a set of commands that are to be executed on startup of litecli. For my case this would for instance be `.tables` - as I always seem to forget their names.
Startupcommands a... | null | 2023-05-03 21:01:28+00:00 | 2023-05-12 01:02:31+00:00 | tests/liteclirc | [main]
# Multi-line mode allows breaking up the sql statements into multiple lines. If
# this is set to True, then the end of the statements must have a semi-colon.
# If this is set to False then sql statements can't be split into multiple
# lines. End of line (return) is considered as the end of the statement.
multi_... | [main]
# Multi-line mode allows breaking up the sql statements into multiple lines. If
# this is set to True, then the end of the statements must have a semi-colon.
# If this is set to False then sql statements can't be split into multiple
# lines. End of line (return) is considered as the end of the statement.
multi_... | bjornasm | e5dacd9f0861d1c3a45e8f339ca3a71f5dee2359 | e95c17f435ccc16f57d3f7dba52d546676690e0c | Doesn't this need a `[startup_commands]` section? | amjith | 1 |
dbcli/litecli | 160 | Fixing startupcommands and successful perpetually set as True | ## Description
### Startup commands
As seen in https://github.com/dbcli/litecli/issues/56 there is a wish for having the option to define a set of commands that are to be executed on startup of litecli. For my case this would for instance be `.tables` - as I always seem to forget their names.
Startupcommands a... | null | 2023-05-03 21:01:28+00:00 | 2023-05-12 01:02:31+00:00 | tests/liteclirc | [main]
# Multi-line mode allows breaking up the sql statements into multiple lines. If
# this is set to True, then the end of the statements must have a semi-colon.
# If this is set to False then sql statements can't be split into multiple
# lines. End of line (return) is considered as the end of the statement.
multi_... | [main]
# Multi-line mode allows breaking up the sql statements into multiple lines. If
# this is set to True, then the end of the statements must have a semi-colon.
# If this is set to False then sql statements can't be split into multiple
# lines. End of line (return) is considered as the end of the statement.
multi_... | bjornasm | e5dacd9f0861d1c3a45e8f339ca3a71f5dee2359 | e95c17f435ccc16f57d3f7dba52d546676690e0c | Oh yes, sorry. Fixed that and added a test to check for the startup commands. Now only miss a test for the execution of the startup commands which I am not sure on how I should do - the check itself is fine but not sure how I invoke the cli without the cli taking over. | bjornasm | 2 |
lucidrains/DALLE-pytorch | 327 | Generate text with DALLE | Since DALLE trains a multimodal language model, the text part of the sequence can also be generated from scratch.
I added a new method to generate text in the DALLE class and also an argument in generate.py so that the generated image can be conditioned on a generated text instead of an input text.
To make this work ... | null | 2021-06-30 09:36:00+00:00 | 2021-07-08 18:57:49+00:00 | dalle_pytorch/dalle_pytorch.py | from math import log2, sqrt
import torch
from torch import nn, einsum
import torch.nn.functional as F
from axial_positional_embedding import AxialPositionalEmbedding
from einops import rearrange
from dalle_pytorch import distributed_utils
from dalle_pytorch.vae import OpenAIDiscreteVAE, VQGanVAE
from dalle_pytorch.tr... | from math import log2, sqrt
import torch
from torch import nn, einsum
import torch.nn.functional as F
import numpy as np
from axial_positional_embedding import AxialPositionalEmbedding
from einops import rearrange
from dalle_pytorch import distributed_utils, tokenizer
from dalle_pytorch.vae import OpenAIDiscreteVAE, ... | jules-samaran | 01e402e4001d8075004c85b07b12429b8a01e822 | fd931e16925bc1844277be83b96c19d13ab6f196 | being able to provide a text here could be interesting.
Instead of starting from scratch, that could make it possible to complete an initial text | rom1504 | 0 |
lucidrains/DALLE-pytorch | 327 | Generate text with DALLE | Since DALLE trains a multimodal language model, the text part of the sequence can also be generated from scratch.
I added a new method to generate text in the DALLE class and also an argument in generate.py so that the generated image can be conditioned on a generated text instead of an input text.
To make this work ... | null | 2021-06-30 09:36:00+00:00 | 2021-07-08 18:57:49+00:00 | dalle_pytorch/tokenizer.py | # take from https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
# to give users a quick easy start to training DALL-E without doing BPE
import torch
import youtokentome as yttm
from tokenizers import Tokenizer
from tokenizers.processors import ByteLevel
from transformers import BertTokenizer
import htm... | # take from https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
# to give users a quick easy start to training DALL-E without doing BPE
import torch
import youtokentome as yttm
from tokenizers import Tokenizer
from tokenizers.processors import ByteLevel
from transformers import BertTokenizer
import htm... | jules-samaran | 01e402e4001d8075004c85b07b12429b8a01e822 | fd931e16925bc1844277be83b96c19d13ab6f196 | the variable used below seems to be pad_tokens not ignore_pad_tokens | rom1504 | 1 |
lucidrains/DALLE-pytorch | 327 | Generate text with DALLE | Since DALLE trains a multimodal language model, the text part of the sequence can also be generated from scratch.
I added a new method to generate text in the DALLE class and also an argument in generate.py so that the generated image can be conditioned on a generated text instead of an input text.
To make this work ... | null | 2021-06-30 09:36:00+00:00 | 2021-07-08 18:57:49+00:00 | dalle_pytorch/tokenizer.py | # take from https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
# to give users a quick easy start to training DALL-E without doing BPE
import torch
import youtokentome as yttm
from tokenizers import Tokenizer
from tokenizers.processors import ByteLevel
from transformers import BertTokenizer
import htm... | # take from https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
# to give users a quick easy start to training DALL-E without doing BPE
import torch
import youtokentome as yttm
from tokenizers import Tokenizer
from tokenizers.processors import ByteLevel
from transformers import BertTokenizer
import htm... | jules-samaran | 01e402e4001d8075004c85b07b12429b8a01e822 | fd931e16925bc1844277be83b96c19d13ab6f196 | what do you think about making this a set instead of a list ? the `in` operator of a set is O(1) instead of O(n) for a list. It will be a little bit faster | rom1504 | 2 |
lucidrains/DALLE-pytorch | 327 | Generate text with DALLE | Since DALLE trains a multimodal language model, the text part of the sequence can also be generated from scratch.
I added a new method to generate text in the DALLE class and also an argument in generate.py so that the generated image can be conditioned on a generated text instead of an input text.
To make this work ... | null | 2021-06-30 09:36:00+00:00 | 2021-07-08 18:57:49+00:00 | dalle_pytorch/tokenizer.py | # take from https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
# to give users a quick easy start to training DALL-E without doing BPE
import torch
import youtokentome as yttm
from tokenizers import Tokenizer
from tokenizers.processors import ByteLevel
from transformers import BertTokenizer
import htm... | # take from https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
# to give users a quick easy start to training DALL-E without doing BPE
import torch
import youtokentome as yttm
from tokenizers import Tokenizer
from tokenizers.processors import ByteLevel
from transformers import BertTokenizer
import htm... | jules-samaran | 01e402e4001d8075004c85b07b12429b8a01e822 | fd931e16925bc1844277be83b96c19d13ab6f196 | Nice catch | jules-samaran | 3 |
lucidrains/DALLE-pytorch | 327 | Generate text with DALLE | Since DALLE trains a multimodal language model, the text part of the sequence can also be generated from scratch.
I added a new method to generate text in the DALLE class and also an argument in generate.py so that the generated image can be conditioned on a generated text instead of an input text.
To make this work ... | null | 2021-06-30 09:36:00+00:00 | 2021-07-08 18:57:49+00:00 | dalle_pytorch/tokenizer.py | # take from https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
# to give users a quick easy start to training DALL-E without doing BPE
import torch
import youtokentome as yttm
from tokenizers import Tokenizer
from tokenizers.processors import ByteLevel
from transformers import BertTokenizer
import htm... | # take from https://github.com/openai/CLIP/blob/main/clip/simple_tokenizer.py
# to give users a quick easy start to training DALL-E without doing BPE
import torch
import youtokentome as yttm
from tokenizers import Tokenizer
from tokenizers.processors import ByteLevel
from transformers import BertTokenizer
import htm... | jules-samaran | 01e402e4001d8075004c85b07b12429b8a01e822 | fd931e16925bc1844277be83b96c19d13ab6f196 | Good idea, I've just fixed that | jules-samaran | 4 |
lucidrains/DALLE-pytorch | 320 | stable_softmax, wanb_entity, visible discord, replace buggy colab | edit: alright rom1504 is being awesome and implementing things the proper modular way for us. I'm gonna focus this PR on a few outstanding issues
> Seems the CompVis team hasn't updated their PyPi because their latest `pip` wheel still doesn't contain the necessary `GumbelVQ` class. I've had to install this as a s... | null | 2021-06-26 19:35:03+00:00 | 2021-06-30 17:15:24+00:00 | README.md | <img src="./images/birds.png" width="500px"></img>
** current best, trained by <a href="https://github.com/kobiso">Kobiso</a> **
## DALL-E in Pytorch
Implementation / replication of <a href="https://openai.com/blog/dall-e/">DALL-E</a> (<a href="https://arxiv.org/abs/2102.12092">paper</a>), OpenAI's Text to Image Tra... | # DALL-E in Pytorch
<p align='center'>
<a href="https://colab.research.google.com/gist/afiaka87/b29213684a1dd633df20cab49d05209d/train_dalle_pytorch.ipynb">
<img alt="Train DALL-E w/ DeepSpeed" src="https://colab.research.google.com/assets/colab-badge.svg">
</a>
<a href="https://discord.gg/dall-e"><img ... | afiaka87 | 7eb2e34ac07076a5bc99808b38795bb12e285f26 | 5a255eab032bcd821c2038c808b9682e485b3f1a | This shouldn't be all in bold imo (I mean all the text below, not the title) | rom1504 | 5 |
lucidrains/DALLE-pytorch | 320 | stable_softmax, wanb_entity, visible discord, replace buggy colab | edit: alright rom1504 is being awesome and implementing things the proper modular way for us. I'm gonna focus this PR on a few outstanding issues
> Seems the CompVis team hasn't updated their PyPi because their latest `pip` wheel still doesn't contain the necessary `GumbelVQ` class. I've had to install this as a s... | null | 2021-06-26 19:35:03+00:00 | 2021-06-30 17:15:24+00:00 | dalle_pytorch/distributed_backends/distributed_backend.py | """
An abstract backend for distributed deep learning.
Provides several standard utility methods under a common API.
Please check the documentation of the class `DistributedBackend` for
details to implement a new backend.
"""
from importlib import import_module
class DistributedBackend:
"""An abstract backend c... | """
An abstract backend for distributed deep learning.
Provides several standard utility methods under a common API.
Please check the documentation of the class `DistributedBackend` for
details to implement a new backend.
"""
from importlib import import_module
class DistributedBackend:
"""An abstract backend c... | afiaka87 | 7eb2e34ac07076a5bc99808b38795bb12e285f26 | 5a255eab032bcd821c2038c808b9682e485b3f1a | any reason for this change? | rom1504 | 6 |
lucidrains/DALLE-pytorch | 320 | stable_softmax, wanb_entity, visible discord, replace buggy colab | edit: alright rom1504 is being awesome and implementing things the proper modular way for us. I'm gonna focus this PR on a few outstanding issues
> Seems the CompVis team hasn't updated their PyPi because their latest `pip` wheel still doesn't contain the necessary `GumbelVQ` class. I've had to install this as a s... | null | 2021-06-26 19:35:03+00:00 | 2021-06-30 17:15:24+00:00 | dalle_pytorch/distributed_backends/distributed_backend.py | """
An abstract backend for distributed deep learning.
Provides several standard utility methods under a common API.
Please check the documentation of the class `DistributedBackend` for
details to implement a new backend.
"""
from importlib import import_module
class DistributedBackend:
"""An abstract backend c... | """
An abstract backend for distributed deep learning.
Provides several standard utility methods under a common API.
Please check the documentation of the class `DistributedBackend` for
details to implement a new backend.
"""
from importlib import import_module
class DistributedBackend:
"""An abstract backend c... | afiaka87 | 7eb2e34ac07076a5bc99808b38795bb12e285f26 | 5a255eab032bcd821c2038c808b9682e485b3f1a | got it, this is just the abstract class and this is done in all inherited classes | rom1504 | 7 |
lucidrains/DALLE-pytorch | 302 | Expose flops_profiler, attn_dropout, ff_dropout | null | null | 2021-06-13 12:41:35+00:00 | 2021-06-15 15:53:18+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | afiaka87 | dc147ca0acaa487950d58935aaf157d1763b1d7d | cec0797e8c114f9c8947b4bb4de710720bbc8359 | Woops. ;) | janEbert | 8 |
lucidrains/DALLE-pytorch | 302 | Expose flops_profiler, attn_dropout, ff_dropout | null | null | 2021-06-13 12:41:35+00:00 | 2021-06-15 15:53:18+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | afiaka87 | dc147ca0acaa487950d58935aaf157d1763b1d7d | cec0797e8c114f9c8947b4bb4de710720bbc8359 | Could it be that this does not work due to the exception you raise? Just an intuitive guess; maybe `print([...]); return` helps? | janEbert | 9 |
lucidrains/DALLE-pytorch | 302 | Expose flops_profiler, attn_dropout, ff_dropout | null | null | 2021-06-13 12:41:35+00:00 | 2021-06-15 15:53:18+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | afiaka87 | dc147ca0acaa487950d58935aaf157d1763b1d7d | cec0797e8c114f9c8947b4bb4de710720bbc8359 | Good idea! Forgot about this thanks; Definitely the better method compared to parsing the scroll back. | afiaka87 | 10 |
lucidrains/DALLE-pytorch | 296 | Save/Resume optimizer state, scheduler state, and epoch | Save/Resume optimizer state, scheduler state, and epoch
Previously only the weights were saved, but for resuming we also need optimizer state, scheduler state, and epoch.
| null | 2021-06-12 08:23:32+00:00 | 2021-06-16 01:13:10+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | mehdidc | 50fb9711cdbf0af0aac823ff9770f86937bdff9c | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | Nitpicking - could we get a space before that `=` sign there? | afiaka87 | 11 |
lucidrains/DALLE-pytorch | 296 | Save/Resume optimizer state, scheduler state, and epoch | Save/Resume optimizer state, scheduler state, and epoch
Previously only the weights were saved, but for resuming we also need optimizer state, scheduler state, and epoch.
| null | 2021-06-12 08:23:32+00:00 | 2021-06-16 01:13:10+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | mehdidc | 50fb9711cdbf0af0aac823ff9770f86937bdff9c | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | Oh wow this has bit me so many times on resume! Finally. | afiaka87 | 12 |
lucidrains/DALLE-pytorch | 296 | Save/Resume optimizer state, scheduler state, and epoch | Save/Resume optimizer state, scheduler state, and epoch
Previously only the weights were saved, but for resuming we also need optimizer state, scheduler state, and epoch.
| null | 2021-06-12 08:23:32+00:00 | 2021-06-16 01:13:10+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | mehdidc | 50fb9711cdbf0af0aac823ff9770f86937bdff9c | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | Till now I've had to use the DeepSpeed native optimizers to resume state; awesome. | afiaka87 | 13 |
lucidrains/DALLE-pytorch | 296 | Save/Resume optimizer state, scheduler state, and epoch | Save/Resume optimizer state, scheduler state, and epoch
Previously only the weights were saved, but for resuming we also need optimizer state, scheduler state, and epoch.
| null | 2021-06-12 08:23:32+00:00 | 2021-06-16 01:13:10+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | mehdidc | 50fb9711cdbf0af0aac823ff9770f86937bdff9c | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | Thanks, fixed | mehdidc | 14 |
lucidrains/DALLE-pytorch | 293 | add vqgan_model_path and vqgan_config_path parameters for custom vqgan support | load the vqgan model from the provided path and config when not None
This is still a draft:
* I need to run it to check it works
* This feels kind of awkward with the vae_params support directly stored in dalle.pt: where should we go ?
Should we move to storing all kind of vae weights in dalle.pt ? or should we... | null | 2021-06-11 09:53:45+00:00 | 2021-06-15 20:44:54+00:00 | dalle_pytorch/vae.py | import io
import sys
import os, sys
import requests
import PIL
import warnings
import os
import hashlib
import urllib
import yaml
from pathlib import Path
from tqdm import tqdm
from math import sqrt
from omegaconf import OmegaConf
from taming.models.vqgan import VQModel
import torch
from torch import nn
import torch.n... | import io
import sys
import os, sys
import requests
import PIL
import warnings
import os
import hashlib
import urllib
import yaml
from pathlib import Path
from tqdm import tqdm
from math import sqrt, log
from omegaconf import OmegaConf
from taming.models.vqgan import VQModel
import torch
from torch import nn
import to... | rom1504 | cec0797e8c114f9c8947b4bb4de710720bbc8359 | 50fb9711cdbf0af0aac823ff9770f86937bdff9c | I chose to kept image size hardcoded because adding image_size parameter would be a larger change. could be done in another pr | rom1504 | 15 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | this is a small unrelated change, which I found useful as if you start multiple runs from the same folder wandb gets confused and "resume" from any of the currently running runs
I can revert the change if you think that's not good | rom1504 | 16 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | I've had mixed success with this; sometimes it is in fact able to resume but even in that case, it has lost the state regarding the current epoch (which should perhaps be saved in the dalle.pt checkpoint?
At any rate - I don't mind this change but I think the proper way to go about things would be to retrieve the ge... | afiaka87 | 17 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | Could we put something like "destructive" or "warning" in the help for this? | afiaka87 | 18 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | "(Careful) Deletes old deepspeed checkpoints if there are more than n" perhaps? | afiaka87 | 19 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | Hmm, do we know if the `convert_to_fp32.py` file changes the date modified of the checkpoint? Because that might be an edge case where files are accidentally deleted.
Aside from that; I think as long as users know this a destructive process this should be fine. | afiaka87 | 20 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | not sure what this script does (does it even work?), but if it changes the files yeah for sure it will change the modified time. The impact would be these recent files would be kept, that seems reasonable to me. | rom1504 | 21 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | yeah that kind of "fixing the wandb resuming issue properly" fixes should be done in another PR.
Do you want me to revert this change in this PR ?
With the current code (resume set at true in wandb when resuming), bugs are occuring, so that's why I changed it here. | rom1504 | 22 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | added it | rom1504 | 23 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | Ah yeah - I forgot how time worked apparently. Not sure what I was thinking.
Anyway looks ready to go in my opinion.
@lucidrains are you available to merge this? | afiaka87 | 24 |
lucidrains/DALLE-pytorch | 285 | Add an option to keep only N deepspeed checkpoints | Very useful to avoid filling up the disk with hundred of GBs of checkpoints | null | 2021-06-05 20:11:06+00:00 | 2021-06-13 03:09:59+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | rom1504 | 77da3edf4e76ba614fdf8bb57c7455ede104858c | 653b5f9edca3131ec13d9148104fa899c35795ed | (lets just revert; it'll be easier to get it merged that way) | afiaka87 | 25 |
lucidrains/DALLE-pytorch | 280 | Added support for webdataset | null | null | 2021-06-01 20:58:02+00:00 | 2021-06-16 22:13:04+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | robvanvolt | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | 2eceb841b4a5795a56165a941b69a30da4cad3e6 | awesome imports great job | chesse20 | 26 |
lucidrains/DALLE-pytorch | 280 | Added support for webdataset | null | null | 2021-06-01 20:58:02+00:00 | 2021-06-16 22:13:04+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | robvanvolt | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | 2eceb841b4a5795a56165a941b69a30da4cad3e6 | Adjusted the imports so they have only the necessary functions imported, thank you for your feedback!:) | robvanvolt | 27 |
lucidrains/DALLE-pytorch | 280 | Added support for webdataset | null | null | 2021-06-01 20:58:02+00:00 | 2021-06-16 22:13:04+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | robvanvolt | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | 2eceb841b4a5795a56165a941b69a30da4cad3e6 | Looks like you changed the defaults @robvanvolt
Did you mean to ?
Otherwise let's revert this part | rom1504 | 28 |
lucidrains/DALLE-pytorch | 280 | Added support for webdataset | null | null | 2021-06-01 20:58:02+00:00 | 2021-06-16 22:13:04+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | robvanvolt | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | 2eceb841b4a5795a56165a941b69a30da4cad3e6 | #311 | rom1504 | 29 |
lucidrains/DALLE-pytorch | 280 | Added support for webdataset | null | null | 2021-06-01 20:58:02+00:00 | 2021-06-16 22:13:04+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | import argparse
from pathlib import Path
import time
from glob import glob
import os
import shutil
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.util... | robvanvolt | d6107ccc24f2fbbc72e1fabbd97f01e6b143606d | 2eceb841b4a5795a56165a941b69a30da4cad3e6 | I did it for training on my older computers, as I always ran out of RAM, so it was much easier for me with the default being 128 - I don't mind, we can revert it again!:) | robvanvolt | 30 |
lucidrains/DALLE-pytorch | 256 | Deepspeed fix : save the normal model too | useful to be able to get the model for generation even when using deepspeed | null | 2021-05-26 09:26:12+00:00 | 2021-06-05 20:24:24+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | rom1504 | 1335a1b383f5d2b34f1fc95d45f6fc30ad0376d4 | 80996978cbb7390f981a0832d972e4d8f5bae945 | Can we get a `time.sleep(2)` after this line? In particular, when using DeepSpeed there is a lot of scrollback and this message gets lost almost immediately. | afiaka87 | 31 |
lucidrains/DALLE-pytorch | 256 | Deepspeed fix : save the normal model too | useful to be able to get the model for generation even when using deepspeed | null | 2021-05-26 09:26:12+00:00 | 2021-06-05 20:24:24+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | rom1504 | 1335a1b383f5d2b34f1fc95d45f6fc30ad0376d4 | 80996978cbb7390f981a0832d972e4d8f5bae945 | One nice thing about python's tracebacks is that you get to see the comment if an error occurs directly on a given line. I've made an wiki page you can link to instead here - which will be clickable in many terminals as well.
https://github.com/lucidrains/DALLE-pytorch/wiki/DeepSpeed-Checkpoints | afiaka87 | 32 |
lucidrains/DALLE-pytorch | 256 | Deepspeed fix : save the normal model too | useful to be able to get the model for generation even when using deepspeed | null | 2021-05-26 09:26:12+00:00 | 2021-06-05 20:24:24+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | rom1504 | 1335a1b383f5d2b34f1fc95d45f6fc30ad0376d4 | 80996978cbb7390f981a0832d972e4d8f5bae945 | Nitpicking at this point -
How about this instead?
```python
# Saves a checkpoint before training begins to fail early when mis-configured.
# See https://github.com/lucidrains/DALLE-pytorch/wiki/DeepSpeed-Checkpoints
``` | afiaka87 | 33 |
lucidrains/DALLE-pytorch | 256 | Deepspeed fix : save the normal model too | useful to be able to get the model for generation even when using deepspeed | null | 2021-05-26 09:26:12+00:00 | 2021-06-05 20:24:24+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | rom1504 | 1335a1b383f5d2b34f1fc95d45f6fc30ad0376d4 | 80996978cbb7390f981a0832d972e4d8f5bae945 | agreed, done | rom1504 | 34 |
lucidrains/DALLE-pytorch | 256 | Deepspeed fix : save the normal model too | useful to be able to get the model for generation even when using deepspeed | null | 2021-05-26 09:26:12+00:00 | 2021-06-05 20:24:24+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | rom1504 | 1335a1b383f5d2b34f1fc95d45f6fc30ad0376d4 | 80996978cbb7390f981a0832d972e4d8f5bae945 | do you mean putting that in the comment ? if so, done | rom1504 | 35 |
lucidrains/DALLE-pytorch | 256 | Deepspeed fix : save the normal model too | useful to be able to get the model for generation even when using deepspeed | null | 2021-05-26 09:26:12+00:00 | 2021-06-05 20:24:24+00:00 | train_dalle.py | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | import argparse
from pathlib import Path
import time
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch i... | rom1504 | 1335a1b383f5d2b34f1fc95d45f6fc30ad0376d4 | 80996978cbb7390f981a0832d972e4d8f5bae945 | done | rom1504 | 36 |
lucidrains/DALLE-pytorch | 244 | save and report dalle.pt at the end of each epoch | also add a parameter to allow specifying a different name to avoid overwriting
if running 2 dalle on the same folder
this solves:
* avoid using a lot of disk with one model every 100 steps under wandb folder
* avoid confusion and errors while running 2 models in the same folder | null | 2021-05-11 21:38:16+00:00 | 2021-05-25 15:44:18+00:00 | train_dalle.py | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | rom1504 | bdb04280c9ab55eb20f86b375dc1aad20fbd5315 | e7d3d2792c8a1747c2efcef03003ac538522b400 | Please move this out of the `if` block so `save_model` is called on every worker. | janEbert | 37 |
lucidrains/DALLE-pytorch | 244 | save and report dalle.pt at the end of each epoch | also add a parameter to allow specifying a different name to avoid overwriting
if running 2 dalle on the same folder
this solves:
* avoid using a lot of disk with one model every 100 steps under wandb folder
* avoid confusion and errors while running 2 models in the same folder | null | 2021-05-11 21:38:16+00:00 | 2021-05-25 15:44:18+00:00 | train_dalle.py | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | rom1504 | bdb04280c9ab55eb20f86b375dc1aad20fbd5315 | e7d3d2792c8a1747c2efcef03003ac538522b400 | why do we want to save the model in every worker ?
for example that wouldn't work in my case where the file system is shared between workers | rom1504 | 38 |
lucidrains/DALLE-pytorch | 244 | save and report dalle.pt at the end of each epoch | also add a parameter to allow specifying a different name to avoid overwriting
if running 2 dalle on the same folder
this solves:
* avoid using a lot of disk with one model every 100 steps under wandb folder
* avoid confusion and errors while running 2 models in the same folder | null | 2021-05-11 21:38:16+00:00 | 2021-05-25 15:44:18+00:00 | train_dalle.py | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | rom1504 | bdb04280c9ab55eb20f86b375dc1aad20fbd5315 | e7d3d2792c8a1747c2efcef03003ac538522b400 | Please take a look at the definition of `save_model`. The method has another guard inside. It's important to call this on all workers so DeepSpeed checkpoints are correctly handled. | janEbert | 39 |
lucidrains/DALLE-pytorch | 244 | save and report dalle.pt at the end of each epoch | also add a parameter to allow specifying a different name to avoid overwriting
if running 2 dalle on the same folder
this solves:
* avoid using a lot of disk with one model every 100 steps under wandb folder
* avoid confusion and errors while running 2 models in the same folder | null | 2021-05-11 21:38:16+00:00 | 2021-05-25 15:44:18+00:00 | train_dalle.py | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | rom1504 | bdb04280c9ab55eb20f86b375dc1aad20fbd5315 | e7d3d2792c8a1747c2efcef03003ac538522b400 | got it, I made the change | rom1504 | 40 |
lucidrains/DALLE-pytorch | 207 | Fix various DeepSpeed issues | Revert #204 which disabled GPU usage for VAE training.
Fix #161, fix #185.
- We now let DeepSpeed handle converting the model to FP16 and moving it to GPU(s).
- Remove hacks regarding DeepSpeed and GPU memory usage.
- Register external parameters (could probably be detected automatically with DeepSpeed >= 0.3.15 ... | null | 2021-04-20 14:25:47+00:00 | 2021-04-20 15:43:24+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | janEbert | c9d371281e7d6f7e9fdde7cf0248a64e10dc74c0 | 9cce36d4bd9fb1ff590209a843ffe048d69ee59c | ah - finally :) | afiaka87 | 41 |
lucidrains/DALLE-pytorch | 207 | Fix various DeepSpeed issues | Revert #204 which disabled GPU usage for VAE training.
Fix #161, fix #185.
- We now let DeepSpeed handle converting the model to FP16 and moving it to GPU(s).
- Remove hacks regarding DeepSpeed and GPU memory usage.
- Register external parameters (could probably be detected automatically with DeepSpeed >= 0.3.15 ... | null | 2021-04-20 14:25:47+00:00 | 2021-04-20 15:43:24+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | janEbert | c9d371281e7d6f7e9fdde7cf0248a64e10dc74c0 | 9cce36d4bd9fb1ff590209a843ffe048d69ee59c | Let's hope that's the last of it. ;) | janEbert | 42 |
lucidrains/DALLE-pytorch | 207 | Fix various DeepSpeed issues | Revert #204 which disabled GPU usage for VAE training.
Fix #161, fix #185.
- We now let DeepSpeed handle converting the model to FP16 and moving it to GPU(s).
- Remove hacks regarding DeepSpeed and GPU memory usage.
- Register external parameters (could probably be detected automatically with DeepSpeed >= 0.3.15 ... | null | 2021-04-20 14:25:47+00:00 | 2021-04-20 15:43:24+00:00 | train_vae.py | import math
from math import sqrt
import argparse
# torch
import torch
from torch.optim import Adam
from torch.optim.lr_scheduler import ExponentialLR
# vision imports
from torchvision import transforms as T
from torch.utils.data import DataLoader
from torchvision.datasets import ImageFolder
from torchvision.utils ... | import math
from math import sqrt
import argparse
# torch
import torch
from torch.optim import Adam
from torch.optim.lr_scheduler import ExponentialLR
# vision imports
from torchvision import transforms as T
from torch.utils.data import DataLoader
from torchvision.datasets import ImageFolder
from torchvision.utils ... | janEbert | c9d371281e7d6f7e9fdde7cf0248a64e10dc74c0 | 9cce36d4bd9fb1ff590209a843ffe048d69ee59c | oh wow - forgout about the cuda call to the vae. | afiaka87 | 43 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | .gitignore | # dall-e generation outputs
outputs/
# Byte-compiled / optimized / DLL files
__pycache__/
*.py[cod]
*$py.class
# C extensions
*.so
# Distribution / packaging
.Python
build/
develop-eggs/
dist/
downloads/
eggs/
.eggs/
lib/
lib64/
parts/
sdist/
var/
wheels/
pip-wheel-metadata/
share/python-wheels/
*.egg-info/
.install... | # dall-e generation outputs
outputs/
*.pt
taming/
wandb/
# Byte-compiled / optimized / DLL files
__pycache__/
*.py[cod]
*$py.class
# C extensions
*.so
# Distribution / packaging
.Python
build/
develop-eggs/
dist/
downloads/
eggs/
.eggs/
lib/
lib64/
parts/
sdist/
var/
wheels/
pip-wheel-metadata/
share/python-wheels/
... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | hmm, perhaps this should be *.pt ? | lucidrains | 44 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | Please negate this, we want to _avoid_ shuffling only if using Horovod. :) | janEbert | 45 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | You can remove this `argparse` import; it's already imported (and the PEP 8 import order would be violated here as `argparse` is in the stdlib). | janEbert | 46 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | I thought you wanted those kwargs without spaces. ;) | janEbert | 47 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | Kwarg with inconsistent spaces here. | janEbert | 48 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | well since you're in support of it ha - i went ahead and ran autopep8 on train_dalle.py. I don't really mind one way or the other so long as we're all on the same page. | afiaka87 | 49 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | It would appear that I do, but only on code I touch ha. I just ran autopep8 on it. | afiaka87 | 50 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | i thought @janEbert removed this in an earlier pull request | lucidrains | 51 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | Yeah, you can remove this line. :) | janEbert | 52 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | Here, please keep moving the model to the GPU (`dalle = dalle.cuda()`) in the `if not using_deepspeed` block. We want to let DeepSpeed handle everything after model creation, both FP16 conversion and moving to GPU. That's why I put it into the `if`-block as well. | janEbert | 53 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | I like consistency, but it's not my codebase, so my opinion really doesn't matter. :p | janEbert | 54 |
lucidrains/DALLE-pytorch | 205 | Refactor ImageTextDataset to its own file. Implement error handling a… | …nd index skipping in ImageTextDataset. Refactor args handling in train_dalle.py. | null | 2021-04-19 21:22:09+00:00 | 2021-04-21 23:23:14+00:00 | train_dalle.py | import argparse
from random import choice
from pathlib import Path
# torch
import torch
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
# vision imports
from PIL import Image
from torchvision import transforms as T
from torch.utils.data ... | import argparse
from pathlib import Path
import torch
import wandb # Quit early if user doesn't have wandb installed.
from torch.nn.utils import clip_grad_norm_
from torch.optim import Adam
from torch.optim.lr_scheduler import ReduceLROnPlateau
from torch.utils.data import DataLoader
from dalle_pytorch import OpenAI... | afiaka87 | 130da7f21767c3c0cebb1e3622b2c68abc270d76 | 2d314aaed157ce5d734561dc064f2854ebe36866 | Gotcha that was a mis-merge. Good catch. | afiaka87 | 55 |
posativ/isso | 952 | Allow umlaut domains for website addresses | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [x] (If adding features:) I have added tests to cover my changes
- [x] I ha... | null | 2023-04-18 06:57:32+00:00 | 2023-08-04 13:01:56+00:00 | isso/views/comments.py | # -*- encoding: utf-8 -*-
import collections
import re
import time
import functools
import json # json.dumps to put URL in <script>
import pkg_resources
from configparser import NoOptionError
from datetime import datetime, timedelta
from html import escape
from io import BytesIO as StringIO
from os import path as os... | # -*- encoding: utf-8 -*-
import collections
import re
import time
import functools
import json # json.dumps to put URL in <script>
import pkg_resources
from configparser import NoOptionError
from datetime import datetime, timedelta
from html import escape
from io import BytesIO as StringIO
from os import path as os... | schneidr | 9a0e1867e4ebe7e1ee7106584adb29b16880a955 | 73d9886100fd56cbceb38e2e00b84f52f0328a8c | This comment seems odd to me - urlparse handles port numbers in URLs fine, so there must be something else going on? | jelmer | 0 |
posativ/isso | 952 | Allow umlaut domains for website addresses | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [x] (If adding features:) I have added tests to cover my changes
- [x] I ha... | null | 2023-04-18 06:57:32+00:00 | 2023-08-04 13:01:56+00:00 | isso/views/comments.py | # -*- encoding: utf-8 -*-
import collections
import re
import time
import functools
import json # json.dumps to put URL in <script>
import pkg_resources
from configparser import NoOptionError
from datetime import datetime, timedelta
from html import escape
from io import BytesIO as StringIO
from os import path as os... | # -*- encoding: utf-8 -*-
import collections
import re
import time
import functools
import json # json.dumps to put URL in <script>
import pkg_resources
from configparser import NoOptionError
from datetime import datetime, timedelta
from html import escape
from io import BytesIO as StringIO
from os import path as os... | schneidr | 9a0e1867e4ebe7e1ee7106584adb29b16880a955 | 73d9886100fd56cbceb38e2e00b84f52f0328a8c | My bad, the reason for removing the port was not urlparse, it is `validators.domain()` which does not accept `domain:port`. I guess I could clean this up by using `hostname` instead of `netloc`, but if the complete URL is supposed to be validated these lines are most probably not staying anyway. | schneidr | 1 |
posativ/isso | 952 | Allow umlaut domains for website addresses | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [x] (If adding features:) I have added tests to cover my changes
- [x] I ha... | null | 2023-04-18 06:57:32+00:00 | 2023-08-04 13:01:56+00:00 | isso/views/comments.py | # -*- encoding: utf-8 -*-
import collections
import re
import time
import functools
import json # json.dumps to put URL in <script>
import pkg_resources
from configparser import NoOptionError
from datetime import datetime, timedelta
from html import escape
from io import BytesIO as StringIO
from os import path as os... | # -*- encoding: utf-8 -*-
import collections
import re
import time
import functools
import json # json.dumps to put URL in <script>
import pkg_resources
from configparser import NoOptionError
from datetime import datetime, timedelta
from html import escape
from io import BytesIO as StringIO
from os import path as os... | schneidr | 9a0e1867e4ebe7e1ee7106584adb29b16880a955 | 73d9886100fd56cbceb38e2e00b84f52f0328a8c | I did add my test case in [test_comments.py](https://github.com/posativ/isso/blob/90019450483c601c1f3dee3de1e973a41679e4d9/isso/tests/test_comments.py#L172). | schneidr | 2 |
posativ/isso | 903 | migrate: Handle single newlines in WordPress comments as line breaks | WordPress renders a single newline in a comment as a <br> tag, but Isso renders a single newline in the comment as a single newline in the HTML. This is rendered the same as if it was a space, all text on one line.
To fix, detect single newlines when importing WordPress comments and convert to a line break in Markdo... | null | 2022-06-06 23:04:49+00:00 | 2022-06-12 10:46:19+00:00 | isso/migrate.py | # -*- encoding: utf-8 -*-
import functools
import io
import json
import logging
import os
import re
import sys
import textwrap
from collections import defaultdict
from time import mktime, strptime, time
from urllib.parse import urlparse
from xml.etree import ElementTree
from isso.utils import anonymize
logger = log... | # -*- encoding: utf-8 -*-
import functools
import io
import json
import logging
import os
import re
import sys
import textwrap
from collections import defaultdict
from time import mktime, strptime, time
from urllib.parse import urlparse
from xml.etree import ElementTree
from isso.utils import anonymize
logger = log... | projectgus | b1021d1e57dd595a581a614ecd26edea4ae69557 | 30ef2180f70ebd58b929673e27d73d889442eb99 | ```suggestion
text = re.sub(r'(?!^\n)\n(?!^\n)', ' \n', text, 0)
``` | ix5 | 3 |
posativ/isso | 903 | migrate: Handle single newlines in WordPress comments as line breaks | WordPress renders a single newline in a comment as a <br> tag, but Isso renders a single newline in the comment as a single newline in the HTML. This is rendered the same as if it was a space, all text on one line.
To fix, detect single newlines when importing WordPress comments and convert to a line break in Markdo... | null | 2022-06-06 23:04:49+00:00 | 2022-06-12 10:46:19+00:00 | isso/migrate.py | # -*- encoding: utf-8 -*-
import functools
import io
import json
import logging
import os
import re
import sys
import textwrap
from collections import defaultdict
from time import mktime, strptime, time
from urllib.parse import urlparse
from xml.etree import ElementTree
from isso.utils import anonymize
logger = log... | # -*- encoding: utf-8 -*-
import functools
import io
import json
import logging
import os
import re
import sys
import textwrap
from collections import defaultdict
from time import mktime, strptime, time
from urllib.parse import urlparse
from xml.etree import ElementTree
from isso.utils import anonymize
logger = log... | projectgus | b1021d1e57dd595a581a614ecd26edea4ae69557 | 30ef2180f70ebd58b929673e27d73d889442eb99 | I meant literally that you should insert two spaces at the end of the line. E.g. this:
```markdown
This is a text with 2 trailing spaces
Next line
Two newlines should not be affected.
```
will render as
```html
<p>This is a text with 2 trailing spaces<br>
Next Line</p>
<p>Two newlines should not be affect... | ix5 | 4 |
posativ/isso | 903 | migrate: Handle single newlines in WordPress comments as line breaks | WordPress renders a single newline in a comment as a <br> tag, but Isso renders a single newline in the comment as a single newline in the HTML. This is rendered the same as if it was a space, all text on one line.
To fix, detect single newlines when importing WordPress comments and convert to a line break in Markdo... | null | 2022-06-06 23:04:49+00:00 | 2022-06-12 10:46:19+00:00 | isso/migrate.py | # -*- encoding: utf-8 -*-
import functools
import io
import json
import logging
import os
import re
import sys
import textwrap
from collections import defaultdict
from time import mktime, strptime, time
from urllib.parse import urlparse
from xml.etree import ElementTree
from isso.utils import anonymize
logger = log... | # -*- encoding: utf-8 -*-
import functools
import io
import json
import logging
import os
import re
import sys
import textwrap
from collections import defaultdict
from time import mktime, strptime, time
from urllib.parse import urlparse
from xml.etree import ElementTree
from isso.utils import anonymize
logger = log... | projectgus | b1021d1e57dd595a581a614ecd26edea4ae69557 | 30ef2180f70ebd58b929673e27d73d889442eb99 | @ix5 oh snap, sorry I totally misread your first suggestion. And I clearly need to read the Markdown spec more often!
Updated, thanks for re-explaining. | projectgus | 5 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/css/isso.css | #isso-thread * {
-webkit-box-sizing: border-box;
-moz-box-sizing: border-box;
box-sizing: border-box;
}
#isso-thread .isso-comment-header a {
text-decoration: none;
}
#isso-thread {
padding: 0;
margin: 0;
}
#isso-thread > h4 {
color: #555;
font-weight: bold;
}
#isso-thread > .isso-feedl... | #isso-thread * {
-webkit-box-sizing: border-box;
-moz-box-sizing: border-box;
box-sizing: border-box;
}
#isso-thread .isso-comment-header a {
text-decoration: none;
}
#isso-thread {
padding: 0;
margin: 0;
}
#isso-thread > h4 {
color: #555;
font-weight: bold;
}
#isso-thread > .isso-feedl... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | The `outline: 0` should stay, no? Not that I'm opposed to changing that, but any styling changes should go into another PR. | ix5 | 6 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/css/isso.css | #isso-thread * {
-webkit-box-sizing: border-box;
-moz-box-sizing: border-box;
box-sizing: border-box;
}
#isso-thread .isso-comment-header a {
text-decoration: none;
}
#isso-thread {
padding: 0;
margin: 0;
}
#isso-thread > h4 {
color: #555;
font-weight: bold;
}
#isso-thread > .isso-feedl... | #isso-thread * {
-webkit-box-sizing: border-box;
-moz-box-sizing: border-box;
box-sizing: border-box;
}
#isso-thread .isso-comment-header a {
text-decoration: none;
}
#isso-thread {
padding: 0;
margin: 0;
}
#isso-thread > h4 {
color: #555;
font-weight: bold;
}
#isso-thread > .isso-feedl... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | I'll add it back, but I can't really tell the difference. But you're right, it can be looked at later in a CSS cleanup. | BBaoVanC | 7 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/js/app/isso.js | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | Was the focus loss (`Node.blur()` taken out intentionally? I'm guessing because we don't need to re-add the placeholder text on `blur` event, no sense in triggering it? | ix5 | 8 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/js/app/isso.js | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | this is not "on initial resize" but rather when returning from preview to editing again, no? | ix5 | 9 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/js/app/isso.js | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | Yeah, I'm using the `placeholder` attribute on the textarea. And I feel like it would already trigger the event anyways, since clicking the submit button would make the textarea lose focus. Regardless I can't tell any difference when adding/removing that line. | BBaoVanC | 10 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/js/app/isso.js | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | This function triggers once when you press the Edit link on an existing comment. I originally added this line when I was using `autosize` to resize the textbox, and when it did that, then the first resize of the textarea would make it scroll off screen if the comment was longer than a few lines.
It actually looks li... | BBaoVanC | 11 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/js/app/isso.js | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | rows still 10 here, please set to 5 | ix5 | 12 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/js/app/isso.js | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | /* Isso – Ich schrei sonst!
*/
var $ = require("app/dom");
var utils = require("app/utils");
var config = require("app/config");
var api = require("app/api");
var template = require("app/template");
var i18n = require("app/i18n");
var identicons = require("app/lib/identicons");
var globals = require("app/globals");
"u... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | Forgot about that, fixed now | BBaoVanC | 13 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/js/app/templates/postbox.js | var html = function (globals) {
var i18n = globals.i18n;
var conf = globals.conf;
var author = globals.author;
var email = globals.email;
var website = globals.website;
var notify = conf["reply-notifications-default-enabled"] ? " checked" : '';
return "" +
"<div class='isso-postbox'>"
+ "<div class='iss... | var html = function (globals) {
var i18n = globals.i18n;
var conf = globals.conf;
var author = globals.author;
var email = globals.email;
var website = globals.website;
var notify = conf["reply-notifications-default-enabled"] ? " checked" : '';
return "" +
"<div class='isso-postbox'>"
+ "<div class='iss... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | a) I can understand why you set **rows to 10** here - since we don't have auto-resizing (yet), you want the textarea to be big enough for people to enter medium-length text. But that makes the area quite long; especially on mobile already half a page length is taken up only by the postbox and it's not immediately clear... | ix5 | 14 |
posativ/isso | 887 | js, templates: Replace `contenteditable` `div` with `textarea` | <!-- Just like NASA going to the moon, it's always good to have a checklist when
creating changes.
The following items are listed to help you create a great Pull Request: -->
## Checklist
- [x] All new and existing **tests are passing**
- [ ] (If adding features:) I have added tests to cover my changes
- [ ] ~~(I... | null | 2022-05-26 01:26:57+00:00 | 2022-05-30 17:04:00+00:00 | isso/js/tests/unit/isso.test.js | /**
* @jest-environment jsdom
*/
/* Keep the above exactly as-is!
* https://jestjs.io/docs/configuration#testenvironment-string
* https://jestjs.io/docs/configuration#testenvironmentoptions-object
*/
"use strict";
/*
* Test goals:
* - Test editorify()
* - Test insert()
* - Test insert_loader()
* - Test Pos... | /**
* @jest-environment jsdom
*/
/* Keep the above exactly as-is!
* https://jestjs.io/docs/configuration#testenvironment-string
* https://jestjs.io/docs/configuration#testenvironmentoptions-object
*/
"use strict";
/*
* Test goals:
* - Test editorify()
* - Test insert()
* - Test insert_loader()
* - Test Pos... | BBaoVanC | b2a1c611461b73f3b67c892e04377536b9a2ce4c | 10e5a90df94d8f1b710f5b1d9a22172fb84e8978 | Please instead `test.skip()` this for now. We'll want to test the whole widget at some point in the future again. | ix5 | 15 |
posativ/isso | 867 | README: Include more information, new screenshot | Add more sections, with more links, small install guide, add new screenshot.
This should make the README more inviting to people browsing GH - and also PyPI, as the README is also automatically uploaded there.
Live preview @ [TestPyPI](https://test.pypi.org/project/isso-testpypi/)
Screenshot used:
.
See **[posativ.org/isso](http://posativ.org/isso/)** for more details and
documentation.
.
## Features
- **Comments written in Markdown**
Users can edit or delete own comments (within 15 minu... | ix5 | 25724bc8fdcd71d57dfd983f45e38f56116e76ed | 05ba6023da7337340208c16e1108318195ceb5d4 | Why do we need devel headers or a C compiler? @ix5 | jelmer | 16 |
posativ/isso | 867 | README: Include more information, new screenshot | Add more sections, with more links, small install guide, add new screenshot.
This should make the README more inviting to people browsing GH - and also PyPI, as the README is also automatically uploaded there.
Live preview @ [TestPyPI](https://test.pypi.org/project/isso-testpypi/)
Screenshot used:
.
See **[posativ.org/isso](http://posativ.org/isso/)** for more details and
documentation.
.
## Features
- **Comments written in Markdown**
Users can edit or delete own comments (within 15 minu... | ix5 | 25724bc8fdcd71d57dfd983f45e38f56116e76ed | 05ba6023da7337340208c16e1108318195ceb5d4 | This was a copy-paste from the docs. Probably necessary for misaka and CFFI things? | ix5 | 17 |
posativ/isso | 867 | README: Include more information, new screenshot | Add more sections, with more links, small install guide, add new screenshot.
This should make the README more inviting to people browsing GH - and also PyPI, as the README is also automatically uploaded there.
Live preview @ [TestPyPI](https://test.pypi.org/project/isso-testpypi/)
Screenshot used:
.
See **[posativ.org/isso](http://posativ.org/isso/)** for more details and
documentation.
.
## Features
- **Comments written in Markdown**
Users can edit or delete own comments (within 15 minu... | ix5 | 25724bc8fdcd71d57dfd983f45e38f56116e76ed | 05ba6023da7337340208c16e1108318195ceb5d4 | It looks like you were also the one who added it to the docs in 93d05c46fcb946a5bc2de5808158015c03fce988
I don't think we need C headers for something that's CFFI. | jelmer | 18 |
posativ/isso | 867 | README: Include more information, new screenshot | Add more sections, with more links, small install guide, add new screenshot.
This should make the README more inviting to people browsing GH - and also PyPI, as the README is also automatically uploaded there.
Live preview @ [TestPyPI](https://test.pypi.org/project/isso-testpypi/)
Screenshot used:
.
See **[posativ.org/isso](http://posativ.org/isso/)** for more details and
documentation.
.
## Features
- **Comments written in Markdown**
Users can edit or delete own comments (within 15 minu... | ix5 | 25724bc8fdcd71d57dfd983f45e38f56116e76ed | 05ba6023da7337340208c16e1108318195ceb5d4 | SQLite3 is bundled with python, isn't it? You don't really need the command-line tool. | jelmer | 19 |
posativ/isso | 867 | README: Include more information, new screenshot | Add more sections, with more links, small install guide, add new screenshot.
This should make the README more inviting to people browsing GH - and also PyPI, as the README is also automatically uploaded there.
Live preview @ [TestPyPI](https://test.pypi.org/project/isso-testpypi/)
Screenshot used:
.
See **[posativ.org/isso](http://posativ.org/isso/)** for more details and
documentation.
.
## Features
- **Comments written in Markdown**
Users can edit or delete own comments (within 15 minu... | ix5 | 25724bc8fdcd71d57dfd983f45e38f56116e76ed | 05ba6023da7337340208c16e1108318195ceb5d4 | Devel headers were added way back in [d1a0b3f6](https://github.com/posativ/isso/commit/d1a0b3f6f9d8904b1772d82733608e7fa98105de#diff-9fac36599c82a968157f5bf7ece3fc8c3176fa46bd69a1fc6bd295178b850c1a), I just happen to show up in the blame logs because I bumped versions a couple of times and moved files ;)
As for the ... | ix5 | 20 |
posativ/isso | 867 | README: Include more information, new screenshot | Add more sections, with more links, small install guide, add new screenshot.
This should make the README more inviting to people browsing GH - and also PyPI, as the README is also automatically uploaded there.
Live preview @ [TestPyPI](https://test.pypi.org/project/isso-testpypi/)
Screenshot used:
.
See **[posativ.org/isso](http://posativ.org/isso/)** for more details and
documentation.
.
## Features
- **Comments written in Markdown**
Users can edit or delete own comments (within 15 minu... | ix5 | 25724bc8fdcd71d57dfd983f45e38f56116e76ed | 05ba6023da7337340208c16e1108318195ceb5d4 | Same as for the devel headers, let's verify and then I'm not at all opposed to removing those docs lines. | ix5 | 21 |
posativ/isso | 846 | [client] js: Support enabling reply notifications checkbox by default | Fixes #837
Configure by setting `data-isso-reply-notifications-default-enabled` to `"true"` in the client settings. If set to true, then the checkbox for enabling reply notifications will be checked by default:
 {
var i18n = globals.i18n;
var author = globals.author;
var email = globals.email;
var website = globals.website;
return "" +
"<div class='isso-postbox'>"
+ "<div class='isso-form-wrapper'>"
+ "<div class='isso-textarea-wrapper'>"
+ "<div class='isso-textarea isso-placeh... | var html = function (globals) {
var i18n = globals.i18n;
var conf = globals.conf;
var author = globals.author;
var email = globals.email;
var website = globals.website;
var notify = conf["reply-notifications-default-enabled"] ? " checked" : '';
return "" +
"<div class='isso-postbox'>"
+ "<div class='iss... | BBaoVanC | 0b33a29fe58ccc128f460df26f9e489625ecb235 | 795aed68ea5c39ccb4d96b3a5e1a660d2ea4174b | Maybe someone has a nicer suggestion than this ternary. But not a big deal. | ix5 | 22 |
posativ/isso | 846 | [client] js: Support enabling reply notifications checkbox by default | Fixes #837
Configure by setting `data-isso-reply-notifications-default-enabled` to `"true"` in the client settings. If set to true, then the checkbox for enabling reply notifications will be checked by default:
 {
var i18n = globals.i18n;
var author = globals.author;
var email = globals.email;
var website = globals.website;
return "" +
"<div class='isso-postbox'>"
+ "<div class='isso-form-wrapper'>"
+ "<div class='isso-textarea-wrapper'>"
+ "<div class='isso-textarea isso-placeh... | var html = function (globals) {
var i18n = globals.i18n;
var conf = globals.conf;
var author = globals.author;
var email = globals.email;
var website = globals.website;
var notify = conf["reply-notifications-default-enabled"] ? " checked" : '';
return "" +
"<div class='isso-postbox'>"
+ "<div class='iss... | BBaoVanC | 0b33a29fe58ccc128f460df26f9e489625ecb235 | 795aed68ea5c39ccb4d96b3a5e1a660d2ea4174b | Would it be good to change it to be a boolean (just `conf["reply-notifications-default-enabled"]`), and then put the ternary inside with the other template code? | BBaoVanC | 23 |
posativ/isso | 846 | [client] js: Support enabling reply notifications checkbox by default | Fixes #837
Configure by setting `data-isso-reply-notifications-default-enabled` to `"true"` in the client settings. If set to true, then the checkbox for enabling reply notifications will be checked by default:
 {
var i18n = globals.i18n;
var author = globals.author;
var email = globals.email;
var website = globals.website;
return "" +
"<div class='isso-postbox'>"
+ "<div class='isso-form-wrapper'>"
+ "<div class='isso-textarea-wrapper'>"
+ "<div class='isso-textarea isso-placeh... | var html = function (globals) {
var i18n = globals.i18n;
var conf = globals.conf;
var author = globals.author;
var email = globals.email;
var website = globals.website;
var notify = conf["reply-notifications-default-enabled"] ? " checked" : '';
return "" +
"<div class='isso-postbox'>"
+ "<div class='iss... | BBaoVanC | 0b33a29fe58ccc128f460df26f9e489625ecb235 | 795aed68ea5c39ccb4d96b3a5e1a660d2ea4174b | Eh, it's nitpicking and I'm not even qualified to comment on any Javascript styles. I was hoping for someone to have a better idea, but it's not important. Leave it as-is. | ix5 | 24 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.