nroggendorff commited on
Commit
9a5c3c2
·
verified ·
1 Parent(s): d1b6413

Update train.py

Browse files
Files changed (1) hide show
  1. train.py +7 -7
train.py CHANGED
@@ -17,17 +17,17 @@ EPOCHS = 1
17
  LEARNING_RATE = 2e-4
18
  FACTOR = 12 ** 3 // 3
19
  MAX_SEQ_LENGTH = 128
20
- VOCAB_SIZE = 32000
21
  INPUT_DATASET = "HuggingFaceTB/smollm-corpus"
22
  INSTRUCT_DATASET = "nroggendorff/elephant"
23
  OUTPUT_REPO = "nroggendorff/smallama"
24
- INSTRUCT_FINETUNE_BOOL = True
25
- INIT = 6
26
- SHARD_SIZE = int(15e+5)
27
  FP16 = True
28
- WARMUP_STEPS = 0
29
- WEIGHT_DECAY = 0.
30
- GRADIENT_ACCUMULATION_STEPS = 1
31
  PUSH_TO_HUB = True
32
 
33
  class Space:
 
17
  LEARNING_RATE = 2e-4
18
  FACTOR = 12 ** 3 // 3
19
  MAX_SEQ_LENGTH = 128
20
+ VOCAB_SIZE = 52000
21
  INPUT_DATASET = "HuggingFaceTB/smollm-corpus"
22
  INSTRUCT_DATASET = "nroggendorff/elephant"
23
  OUTPUT_REPO = "nroggendorff/smallama"
24
+ INSTRUCT_FINETUNE_BOOL = False
25
+ INIT = 0
26
+ SHARD_SIZE = int(2e+6)
27
  FP16 = True
28
+ WARMUP_STEPS = 50
29
+ WEIGHT_DECAY = 1e-3
30
+ GRADIENT_ACCUMULATION_STEPS = 2
31
  PUSH_TO_HUB = True
32
 
33
  class Space: