Hugging Face's logo

impossibleexchange
/
smoothtest

Any-to-Any
omega
omegalabs
bittensor
agi
Model card Files
xet
Community
impossibleexchange commited on
Commit
a4f270f
·
verified ·
1 Parent(s): 9ed0c41

Update training_config.yml

Browse files
Files changed (1) hide show
  1. training_config.yml +6 -6
training_config.yml CHANGED
@@ -29,7 +29,7 @@ temperature: 0.6
29
  top_k: 231
30
  dataset:
31
  _component_: ds.EvenBatcher
32
- buffer_size: 64
33
  dataset:
34
  _component_: ds.RoundRobinDataset
35
  datasets:
@@ -49,17 +49,17 @@ shuffle: true
49
  batch_size: 8
50
  optimizer:
51
  _component_: torch.optim.AdamW
52
- weight_decay: 1.0e-05
53
- lr: 0.001
54
  lr_scheduler:
55
  _component_: torchtune.modules.get_cosine_schedule_with_warmup
56
- num_warmup_steps: 100
57
  loss:
58
  _component_: torch.nn.CrossEntropyLoss
59
- label_smoothing: 0.1
60
  epochs: 10
61
  max_steps_per_epoch: null
62
- gradient_accumulation_steps: 128
63
  compile: false
64
  output_dir: /tmp/lora_finetune_output
65
  metric_logger:
 
29
  top_k: 231
30
  dataset:
31
  _component_: ds.EvenBatcher
32
+ buffer_size: 648
33
  dataset:
34
  _component_: ds.RoundRobinDataset
35
  datasets:
 
49
  batch_size: 8
50
  optimizer:
51
  _component_: torch.optim.AdamW
52
+ weight_decay: 1.1
53
+ lr: 0.1
54
  lr_scheduler:
55
  _component_: torchtune.modules.get_cosine_schedule_with_warmup
56
+ num_warmup_steps: 1
57
  loss:
58
  _component_: torch.nn.CrossEntropyLoss
59
+
60
  epochs: 10
61
  max_steps_per_epoch: null
62
+ gradient_accumulation_steps: 512
63
  compile: false
64
  output_dir: /tmp/lora_finetune_output
65
  metric_logger: