wav2vec2-large-xlsr-arabic / train_results.json
kmfoda's picture
First version of the wav2vec2-large-xlsr-arabice model and tokenizer.
30e4171
raw
history blame
437 Bytes
{
"epoch": 30.0,
"init_mem_cpu_alloc_delta": 620838,
"init_mem_cpu_peaked_delta": 18306,
"init_mem_gpu_alloc_delta": 1262070784,
"init_mem_gpu_peaked_delta": 0,
"train_mem_cpu_alloc_delta": 0,
"train_mem_cpu_peaked_delta": 0,
"train_mem_gpu_alloc_delta": 3788204544,
"train_mem_gpu_peaked_delta": 0,
"train_runtime": 115603.8298,
"train_samples": 21744,
"train_samples_per_second": 0.176
}