nroggendorff commited on
Commit
cecdc0b
·
verified ·
1 Parent(s): eb3036f

Update train.py

Browse files
Files changed (1) hide show
  1. train.py +3 -3
train.py CHANGED
@@ -17,8 +17,8 @@ INPUT_DATASET = "HuggingFaceTB/smollm-corpus"
17
  INSTRUCT_DATASET = "nroggendorff/elephant"
18
  OUTPUT_REPO = "nroggendorff/smallama"
19
  INSTRUCT_FINETUNE_BOOL = False
20
- INIT = 0#/4
21
- SHARD_SIZE = int(3e+6)
22
  FP16 = True
23
  WARMUP_STEPS = 0
24
  DECAY = 1e-3
@@ -28,7 +28,7 @@ PUSH_TO_HUB = True
28
  def load_data():
29
  if not INSTRUCT_FINETUNE_BOOL:
30
  dataset = load_dataset(INPUT_DATASET, "cosmopedia-v2", split="train", streaming=True)
31
- dataset = Dataset.from_generator(lambda: dataset.take(int(15e+6)))
32
  # dataset = dataset.shard(num_shards=len(dataset) // SHARD_SIZE, index=INIT)
33
  else:
34
  dataset = load_dataset(INSTRUCT_DATASET, split="train")#, streaming=True)
 
17
  INSTRUCT_DATASET = "nroggendorff/elephant"
18
  OUTPUT_REPO = "nroggendorff/smallama"
19
  INSTRUCT_FINETUNE_BOOL = False
20
+ INIT = 0#/15
21
+ SHARD_SIZE = int(5e+5)
22
  FP16 = True
23
  WARMUP_STEPS = 0
24
  DECAY = 1e-3
 
28
  def load_data():
29
  if not INSTRUCT_FINETUNE_BOOL:
30
  dataset = load_dataset(INPUT_DATASET, "cosmopedia-v2", split="train", streaming=True)
31
+ dataset = Dataset.from_generator(lambda: dataset.take(int(8e+6)))
32
  # dataset = dataset.shard(num_shards=len(dataset) // SHARD_SIZE, index=INIT)
33
  else:
34
  dataset = load_dataset(INSTRUCT_DATASET, split="train")#, streaming=True)