Kuis Pelatihan Presisi Campuran

Jawaban dan penjelasan muncul saat Anda mengikuti kuis.

1. In an FP16 training recipe that keeps an FP32 master weight, why retain it?
   • Untuk menjaga presisi dalam pembaruan bobot kecil yang dapat diselesaikan oleh 16-bit
   • Untuk membuat file model lebih kecil pada disk
   • Untuk mengizinkan pelatihan tanpa GPU
   • Untuk mengenkripsi parameter model
2. Masalah apa yang dipecahkan oleh penskalaan kerugian dalam pelatihan FP16?
   • Model melakukan overfitting pada data
   • Gradien berkurang hingga nol karena rentang FP16 sempit
   • GPU kehabisan ruang disk
   • Pemuatan data lambat dari disk
3. Apa keunggulan bfloat16 dibandingkan FP16?
   • Ini menggunakan total bit yang lebih sedikit
   • Ini menyimpan teks, bukan angka
   • Ini memiliki rentang eksponen yang sama luasnya dengan FP32, sehingga mengurangi kebutuhan penskalaan kerugian
   • Ini hanya berjalan di CPU
4. For many supported Tensor Core matrix operations, how are partial sums commonly accumulated?
   • In integer arithmetic with no rounding
   • In text before converting to FP16
   • In FP16 only, which prevents every numerical error
   • In FP32, which reduces rounding error compared with low-precision accumulation
5. What is a common benefit of using 16-bit values for supported training operations?
   • They can reduce tensor memory and improve throughput, with gains depending on hardware and workload
   • They guarantee higher model accuracy
   • They remove the need to compute gradients
   • They always double end-to-end training speed
6. Which current PyTorch APIs are commonly used for automatic mixed-precision training?
   • torch.cuda.amp only, which is the recommended new namespace
   • torch.amp.autocast and, for FP16 gradient scaling, torch.amp.GradScaler
   • torch.load and torch.save
   • torch.nn.DataParallel and torch.manual_seed
7. Why does bfloat16 often need less loss scaling than FP16?
   • It has more mantissa bits than FP32
   • It stores gradients as integers
   • It has an exponent range like FP32, so very small gradients are less likely to underflow
   • It never rounds values
8. What does autocast do in automatic mixed precision?
   • Converts every tensor and operation to FP16 unconditionally
   • Replaces the optimizer with a different algorithm
   • Guarantees identical bitwise results to FP32
   • Selects operation dtypes so supported operations can use lower precision while others retain safer precision