File size: 2,407 Bytes
31f6626 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 |
_allow_dotted_keys:
value: false
_convert_dict:
value: true
_fields:
value:
ckpt_steps: 1000
evals: |
val:
data:
name: imagenet2012
split: validation
log_steps: 2500
loss_name: softmax_xent
pp_fn: decode|resize_small(256)|central_crop(224)|value_range(-1, 1)|onehot(1000,
key="label", key_result="labels")|keep("image", "labels")
type: classification
grad_clip_norm: 1
input: |
accum_freq: 8
batch_size: 1024
cache_raw: false
data:
name: imagenet2012
split: train
pp: decode_jpeg_and_inception_crop(224)|flip_lr|randaug(2,10)|value_range(-1, 1)|onehot(1000,
key="label", key_result="labels")|keep("image", "labels")
shuffle_buffer_size: 150000
log_training_steps: 50
loss: softmax_xent
lr: 0.001
mixup: |
fold_in: null
p: 0.2
model: |
pool_type: gap
posemb: sincos2d
rep_size: false
variant: S/16
model_name: vit
num_classes: 1000
optax: |
mu_dtype: bfloat16
optax_name: scale_by_adam
pp_modules:
- ops_general
- ops_image
- ops_text
- archive.randaug
schedule: |
decay_type: cosine
warmup_steps: 10000
seed: 0
total_epochs: 90
wd: 0.0001
_locked:
value: true
_sort_keys:
value: true
_type_safe:
value: true
_wandb:
value:
cli_version: 0.18.7
m: []
python_version: 3.11.10
t:
"1":
- 1
- 2
- 3
- 12
- 41
- 45
- 55
"2":
- 1
- 2
- 3
- 12
- 41
- 45
- 55
"3":
- 5
- 13
- 14
- 16
- 23
- 55
- 62
"4": 3.11.10
"5": 0.18.7
"8":
- 5
"12": 0.18.7
"13": linux-x86_64
|