DEV Community

#pytorch

Posts

đź‘‹ Sign in for the ability to sort posts by relevant, latest, or top.
SDXL images per dollar on four GPUs: RTX A6000, L40S, RTX 5090 and H100, eager vs torch.compile

SDXL images per dollar on four GPUs: RTX A6000, L40S, RTX 5090 and H100, eager vs torch.compile

Comments
6 min read
Batching is the feature: serving PyTorch at low traffic

Batching is the feature: serving PyTorch at low traffic

Comments
1 min read
4 GPUs, 2x speed: why PyTorch DDP stalled on a PCIe-only 4x RTX A6000, and two flags that took it to 3.6x

4 GPUs, 2x speed: why PyTorch DDP stalled on a PCIe-only 4x RTX A6000, and two flags that took it to 3.6x

Comments
4 min read
How long until a rented GPU runs CUDA? Launch to first matmul on six GPU types, timed

How long until a rented GPU runs CUDA? Launch to first matmul on six GPU types, timed

Comments
5 min read
My renderer's seed argument changes nothing — how do I prove the seed actually reaches the sampler?

My renderer's seed argument changes nothing — how do I prove the seed actually reaches the sampler?

Comments
1 min read
Beyond seeds: which determinism flags actually mattered for reproducible fine-tunes on a T4?

Beyond seeds: which determinism flags actually mattered for reproducible fine-tunes on a T4?

Comments
1 min read
Same fine-tune recipe, different seeds — one narrow behaviour swings wildly. What actually stabilises it?

Same fine-tune recipe, different seeds — one narrow behaviour swings wildly. What actually stabilises it?

Comments
1 min read
Google and Meta Open-Source TorchTPU for Running PyTorch on TPUs

Google and Meta Open-Source TorchTPU for Running PyTorch on TPUs

1
Comments
3 min read
Nodus vs Lambda: which is better for GPU training?

Nodus vs Lambda: which is better for GPU training?

Comments
8 min read
Four tools for seismic data

Four tools for seismic data

Comments
26 min read
Why a Keras model fails to convert to TFLite: 11 known cases, 5 traps

Why a Keras model fails to convert to TFLite: 11 known cases, 5 traps

Comments
6 min read
EMA's Decay Isn't a Dial. It's a Plateau, a Cliff, and a 10%-Accuracy Crater

EMA's Decay Isn't a Dial. It's a Plateau, a Cliff, and a 10%-Accuracy Crater

Comments
5 min read
TorchTPU vs CUDA: Can Native PyTorch Make Google TPUs Easier to Use?

TorchTPU vs CUDA: Can Native PyTorch Make Google TPUs Easier to Use?

1
Comments
8 min read
Why Long bf16 Training Runs Stall: The Residual Stream Has No Room for Small Updates

Why Long bf16 Training Runs Stall: The Residual Stream Has No Room for Small Updates

Comments
8 min read
Google Open-Sourced TorchTPU. PyTorch/XLA Is Deprecated and Gets Archived November 15

Google Open-Sourced TorchTPU. PyTorch/XLA Is Deprecated and Gets Archived November 15

Comments
7 min read
đź‘‹ Sign in for the ability to sort posts by relevant, latest, or top.