a D-type, better known as a dtype in the numpy community,
March 12, 2026 at 9:51 PM
a D-type, better known as a dtype in the numpy community,
Thank you for making nse dtype like a true loser i Sm
February 6, 2025 at 2:19 AM
Thank you for making nse dtype like a true loser i Sm
July 5, 2025 at 12:27 AM
Did you know that on Argilla, we’re adding a new feature to export labeled datasets directly to the Hugging Face Hub? 🤔
We’re leveraging the Hugging Face datasets library for seamless integration, including defining span labeling
Stay tuned for the release!🧠✨
#MachineLearning #NLP #DataLabeling
We’re leveraging the Hugging Face datasets library for seamless integration, including defining span labeling
Stay tuned for the release!🧠✨
#MachineLearning #NLP #DataLabeling
November 26, 2024 at 12:52 PM
Did you know that on Argilla, we’re adding a new feature to export labeled datasets directly to the Hugging Face Hub? 🤔
We’re leveraging the Hugging Face datasets library for seamless integration, including defining span labeling
Stay tuned for the release!🧠✨
#MachineLearning #NLP #DataLabeling
We’re leveraging the Hugging Face datasets library for seamless integration, including defining span labeling
Stay tuned for the release!🧠✨
#MachineLearning #NLP #DataLabeling
nice attention :) is it causal? and multiheaded, i assume?quantized? what dtype? btw is it sparse? latent? is it fast, memory-efficient, and exact? :3 oh with io-awareness? are yr queries grouped? ^^ w/ a sliding window? oh oh is it gated? is it linear? is ur positional encoding rotary? :D
June 5, 2026 at 3:17 AM
nice attention :) is it causal? and multiheaded, i assume?quantized? what dtype? btw is it sparse? latent? is it fast, memory-efficient, and exact? :3 oh with io-awareness? are yr queries grouped? ^^ w/ a sliding window? oh oh is it gated? is it linear? is ur positional encoding rotary? :D
"write this same function (dtype x arch) more times and test it" is almost perfectly suited for the intern bot
April 20, 2026 at 3:55 PM
"write this same function (dtype x arch) more times and test it" is almost perfectly suited for the intern bot
Jaguar D Type 1954 to 1957 UK 🇬🇧
#jaguar #dtype #jaguardtype #realjaguarcars #racingcar #britishcars #classiccars #carphotography #carspotting #jaguarracing
#jaguar #dtype #jaguardtype #realjaguarcars #racingcar #britishcars #classiccars #carphotography #carspotting #jaguarracing
January 1, 2025 at 8:46 AM
Jaguar D Type 1954 to 1957 UK 🇬🇧
#jaguar #dtype #jaguardtype #realjaguarcars #racingcar #britishcars #classiccars #carphotography #carspotting #jaguarracing
#jaguar #dtype #jaguardtype #realjaguarcars #racingcar #britishcars #classiccars #carphotography #carspotting #jaguarracing
March 26, 2025 at 7:31 AM
i am mostly confident i can do this because there is already a good cuda kernel for fwht it just only works on one dtype on one gpu arch
April 20, 2026 at 3:51 PM
i am mostly confident i can do this because there is already a good cuda kernel for fwht it just only works on one dtype on one gpu arch
bizarre; what's the dtype? bf16?
August 8, 2025 at 11:42 PM
bizarre; what's the dtype? bf16?
`torch_dtype` is deprecated! Use `dtype` instead!
October 30, 2025 at 8:40 PM
`torch_dtype` is deprecated! Use `dtype` instead!
Would recommend github.com/patrick-kidg... as an updated+better version!
(Even for pytorch, despite the name.)
(Even for pytorch, despite the name.)
GitHub - patrick-kidger/jaxtyping: Type annotations and runtime checking for shape and dtype of JAX/NumPy/PyTorch/etc. arrays. https://docs.kidger.site/jaxtyping/
Type annotations and runtime checking for shape and dtype of JAX/NumPy/PyTorch/etc. arrays. https://docs.kidger.site/jaxtyping/ - patrick-kidger/jaxtyping
github.com
December 19, 2024 at 12:37 PM
Would recommend github.com/patrick-kidg... as an updated+better version!
(Even for pytorch, despite the name.)
(Even for pytorch, despite the name.)
massive shoutout to pandas for changing the dtype of an int column to a float when accessing a row, which makes a series (because a series can't have mixed types)... hence breaking a filename I needed to get out of the int 😑
February 10, 2025 at 9:42 AM
massive shoutout to pandas for changing the dtype of an int column to a float when accessing a row, which makes a series (because a series can't have mixed types)... hence breaking a filename I needed to get out of the int 😑
Jaguar D Type 1954 to 1957 UK 🇬🇧
#jaguar #dtype #jaguardtype #realjaguarcars #racingcar #britishcars #classiccars #carphotography #carspotting #jaguarracing
#jaguar #dtype #jaguardtype #realjaguarcars #racingcar #britishcars #classiccars #carphotography #carspotting #jaguarracing
December 24, 2024 at 12:59 PM
Jaguar D Type 1954 to 1957 UK 🇬🇧
#jaguar #dtype #jaguardtype #realjaguarcars #racingcar #britishcars #classiccars #carphotography #carspotting #jaguarracing
#jaguar #dtype #jaguardtype #realjaguarcars #racingcar #britishcars #classiccars #carphotography #carspotting #jaguarracing
Published arro3-core v0.6 with wider dtype support for converting numpy ndarrays to Arrow. It now additionally supports:
- datetime64
- timedelta64
- fixed-width strings/bytes
- variable-width strings
- object arrays of str/bytes
And the Python wheel is still just 2.5MB!
github.com/kylebarron/a...
- datetime64
- timedelta64
- fixed-width strings/bytes
- variable-width strings
- object arrays of str/bytes
And the Python wheel is still just 2.5MB!
github.com/kylebarron/a...
Release py-v0.6.0 · kylebarron/arro3
arro3-core
Breaking Changes 🔧
No breaking changes.
New features ✨
Expanded dtype support in Array.from_numpy. It now additionally supports datetime64, timedelta64, fixed-width strings and bytes, v...
github.com
August 22, 2025 at 5:23 PM
Published arro3-core v0.6 with wider dtype support for converting numpy ndarrays to Arrow. It now additionally supports:
- datetime64
- timedelta64
- fixed-width strings/bytes
- variable-width strings
- object arrays of str/bytes
And the Python wheel is still just 2.5MB!
github.com/kylebarron/a...
- datetime64
- timedelta64
- fixed-width strings/bytes
- variable-width strings
- object arrays of str/bytes
And the Python wheel is still just 2.5MB!
github.com/kylebarron/a...
First modification : Add an RS-232 transceiver and 9 way Dtype for the second SIO channel.
January 29, 2025 at 3:48 PM
First modification : Add an RS-232 transceiver and 9 way Dtype for the second SIO channel.
ok so matching the dtype (fp64), resolution (4096^2), network size (272 params), & training iterations (1000) of the original paper gives me... 7 and a half hours to render a single frame on my 3090 🙃
February 22, 2026 at 9:10 PM
ok so matching the dtype (fp64), resolution (4096^2), network size (272 params), & training iterations (1000) of the original paper gives me... 7 and a half hours to render a single frame on my 3090 🙃
![In [36]: import pandas as pd
In [37]: from pandas import Series, DataFrame
In [38]: s = pd.Series([10, 20, 30], index=list('abc'))
In [39]: s.index
Out[39]: Index(['a', 'b', 'c'], dtype='str')
In [40]:
...: (
...: s
...: .set_axis(list('qrs'))
...: .loc[['q', 's']]
...: )
Out[40]:
q 10
s 30
dtype: int64
In [41]: s.index
Out[41]: Index(['a', 'b', 'c'], dtype='str')](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:6edw5l7icvzqoofcpfx6c64j/bafkreiei3nmfj3koyau35o2nldwtxz5vf7gir26wkqoakizvkdqf43a3cm)
Polars 2.0 RC2 is out. Breaking changes first: a new Map dtype, cut/qcut deprecated, Parquet ENUM now reads as String. Then the optimizer work: dynamic predicates for hash joins, predicates derived from join conditions. Run your pipelines against it now.
github.com/pola-rs/pola...
github.com/pola-rs/pola...
Release Python Polars 2.0.0-rc.2 · pola-rs/polars
💥 Breaking changes Read Parquet ENUM type as pl.String (#29331) More map operations (#29296) Deprecate cut/qcut (#29329) Remove legacy POLARS_STREAMING_CHUNK_SIZE (#29046) Introduce the Map dtype ...
github.com
September 25, 2026 at 10:00 PM
Polars 2.0 RC2 is out. Breaking changes first: a new Map dtype, cut/qcut deprecated, Parquet ENUM now reads as String. Then the optimizer work: dynamic predicates for hash joins, predicates derived from join conditions. Run your pipelines against it now.
github.com/pola-rs/pola...
github.com/pola-rs/pola...
random_vector = torch.randn(HIDDEN_DIM, dtype=torch.float32, device=device)
noise = torch.randn(VOCAB_SIZE, HIDDEN_DIM, dtype=torch.float32, device=device) * INIT_SIGMA
init = random_vector + noise
model.transformer.wte.weight[:] = init.to(torch.bfloat16)
↑ I think this does it. Maybe. Not sure.
noise = torch.randn(VOCAB_SIZE, HIDDEN_DIM, dtype=torch.float32, device=device) * INIT_SIGMA
init = random_vector + noise
model.transformer.wte.weight[:] = init.to(torch.bfloat16)
↑ I think this does it. Maybe. Not sure.
November 10, 2025 at 2:42 AM
random_vector = torch.randn(HIDDEN_DIM, dtype=torch.float32, device=device)
noise = torch.randn(VOCAB_SIZE, HIDDEN_DIM, dtype=torch.float32, device=device) * INIT_SIGMA
init = random_vector + noise
model.transformer.wte.weight[:] = init.to(torch.bfloat16)
↑ I think this does it. Maybe. Not sure.
noise = torch.randn(VOCAB_SIZE, HIDDEN_DIM, dtype=torch.float32, device=device) * INIT_SIGMA
init = random_vector + noise
model.transformer.wte.weight[:] = init.to(torch.bfloat16)
↑ I think this does it. Maybe. Not sure.
