nroggendorff commited on
Commit
12f0b8b
·
verified ·
1 Parent(s): a41bf75

Update train.py

Browse files
Files changed (1) hide show
  1. train.py +2 -2
train.py CHANGED
@@ -11,7 +11,7 @@ MAX_SEQ_LENGTH = 2048
11
  BATCH_SIZE = 4
12
  EPOCHS = 2
13
  LEARNING_RATE = 5e-5
14
- FACTOR = 1024
15
  VOCAB_SIZE = 32000
16
  INPUT_DATASET = "HuggingFaceTB/smollm-corpus"
17
  INSTRUCT_DATASET = "nroggendorff/openhermes"
@@ -19,7 +19,7 @@ OUTPUT_REPO = "smallama"
19
  FP16 = True
20
  WARMUP_STEPS = 200
21
  DECAY = 1e-4
22
- GRADIENT_ACCUMULATION_STEPS = 4
23
  CLIPPING = 1.0
24
  PUSH_TO_HUB = True
25
 
 
11
  BATCH_SIZE = 4
12
  EPOCHS = 2
13
  LEARNING_RATE = 5e-5
14
+ FACTOR = 2048
15
  VOCAB_SIZE = 32000
16
  INPUT_DATASET = "HuggingFaceTB/smollm-corpus"
17
  INSTRUCT_DATASET = "nroggendorff/openhermes"
 
19
  FP16 = True
20
  WARMUP_STEPS = 200
21
  DECAY = 1e-4
22
+ GRADIENT_ACCUMULATION_STEPS = 16
23
  CLIPPING = 1.0
24
  PUSH_TO_HUB = True
25