|
About the mixed-precision category
|
|
0
|
1523
|
August 24, 2020
|
|
FCN ResNet18 low precision on SUNRGBD dataset
|
|
0
|
5
|
November 20, 2024
|
|
Why tensor.to convert fp32 to fp8_e4m3=Nan if overflow
|
|
2
|
13
|
November 7, 2024
|
|
Any operator is supported on fp8 tensor?
|
|
7
|
56
|
November 5, 2024
|
|
Increased memory usage with AMP
|
|
6
|
4539
|
November 5, 2024
|
|
Can `autocast` handle networks with layers having different dtypes?
|
|
0
|
11
|
October 24, 2024
|
|
Reseting loss value
|
|
0
|
9
|
October 22, 2024
|
|
FSDP MixedPrecision vs AMP autocast?
|
|
0
|
14
|
October 11, 2024
|
|
Custom CUDA kernels with AMP
|
|
0
|
21
|
September 23, 2024
|
|
Half precision training time same as full precision
|
|
4
|
52
|
September 19, 2024
|
|
Why bf16 do not need loss scaling?
|
|
4
|
3341
|
September 5, 2024
|
|
What does the `use_fast_accum` option do in `torch._scaled_mm`
|
|
1
|
172
|
August 27, 2024
|
|
Weight parameters with 8, 14 bit precisions?
|
|
3
|
10
|
August 20, 2024
|
|
How to perform mixed precsion on single F.linear?
|
|
1
|
9
|
August 11, 2024
|
|
Model forward pass in AMP gives NaN
|
|
0
|
29
|
August 5, 2024
|
|
Fp16 inference time cost
|
|
2
|
50
|
August 1, 2024
|
|
Is it a good idea to use float16/bfloat16 for inference?
|
|
2
|
100
|
August 1, 2024
|
|
Why nn.LSTM still use float16 in hidden_state, even if set to bfloat16 or float32 already?
|
|
0
|
29
|
July 29, 2024
|
|
Prediction is different with or without padding: the model is sensiitive to floating point precision?
|
|
0
|
18
|
July 26, 2024
|
|
Question about bfloat16 operations in AMP and cuda
|
|
3
|
99
|
July 11, 2024
|
|
Torch.matmul launch different CUDA kernel from cublas
|
|
2
|
110
|
July 6, 2024
|
|
Fp16 overflow when computing matmul in autocast context
|
|
5
|
1294
|
July 5, 2024
|
|
Mixed precision training with transformer embeddings stored in fp16
|
|
0
|
91
|
June 13, 2024
|
|
Autocast keep cache across multiple forward pass
|
|
0
|
110
|
June 5, 2024
|
|
Precision 16 run problem
|
|
2
|
176
|
June 4, 2024
|
|
Torch.save numerical differences
|
|
6
|
1677
|
May 31, 2024
|
|
AMP during inference
|
|
1
|
311
|
May 31, 2024
|
|
GradScaler for CPU with AMP
|
|
8
|
910
|
May 28, 2024
|
|
BFloat16 training - explicit cast vs autocast
|
|
6
|
3419
|
May 15, 2024
|
|
Alternative to torch.inverse for 16 bit
|
|
2
|
1090
|
May 6, 2024
|