training / flax /run_speed.sh
nullonesix's picture
Saving train state of step 1
a1be16b verified
raw
history blame
1.07 kB
#!/usr/bin/env bash
# --wandb_project "distil-whisper-speed-bench-1024-no-timestamps" \
batch_sizes=(1 16)
names=("openai/whisper-large-v2" "openai/whisper-medium.en" "openai/whisper-small.en" "openai/whisper-base.en" "openai/whisper-tiny.en" "patrickvonplaten/whisper-large-v2-32-2" "patrickvonplaten/whisper-medium-24-2")
# Double loop
for name in "${names[@]}"; do
for batch_size in "${batch_sizes[@]}"; do
CUDA_VISIBLE_DEVICES="1" python ./run_speed_pt.py \
--dataset_name "google/fleurs+distil-whisper/chime4+distil-whisper/earnings22+kensho/spgispeech" \
--wandb_name "T4-bsz${batch_size}-${name}" \
--model_name_or_path ${name} \
--wandb_project "beam-search-distil-whisper-speed-bench-256-no-timestamps" \
--dataset_config_name "en_us+1-channel+chunked+test" \
--dataset_split_name "test+test+test+test" \
--text_column_name "transcription+text+transcription+transcript" \
--samples_per_dataset "256" \
--attn_type "flash2" \
--num_beams 5 \
--batch_size ${batch_size}
done
done