Kastan's group workspace
Group: gpt2_vanilla
Name
3 visualized
Name: gpt2_vanilla
Name: gpt2_vanilla
3
State
Notes
User
Tags
Created
Runtime
Sweep
BATCH_SIZE
HIDDEN_SIZE
LEARNING_RATE
MODE
NUM_EPOCHS
NUM_MICRO_BATCHES
PIPELINE
SEQ_LEN
SEQ_LENGTH
TENSOR_PARALLEL
TENSOR_PARALLEL_MODE
TENSOR_PARALLEL_SIZE
TENSOR_SHAPE
TOTAL_BATCH_SIZE
VOCAB_SIZE
WARMUP_EPOCHS
WEIGHT_DECAY
backend
clip_grad_norm
colossal_config_file
config
data_dir
fp16.mode
from_torch
gpt2_8B
gpt2_small
gpt2_xl
gradient_accumulation
loss.type
loss_fn.type
model.checkpoint
model.dtype
model.fuse_scale_mask_softmax
model.max_position_embeddings
model.type
model.vocab_size
optimizer.lr
optimizer.type
optimizer.weight_decay
parallel.pipeline
parallel.tensor.mode
parallel.tensor.size
zero.model_config.reuse_fp16_shard
zero.model_config.shard_strategy
Finished
kastan
7m 28s
-
1
-
-
-
60
-
-
1024
-
-
-
-
-
-
-
-
-
nccl
-
gpt2_configs/gpt2_vanilla.py
gpt2_configs/gpt2_vanilla.py
/u/kastan/colossal/data/train_data_FINAL.json
AMP_TYPE.NAIVE
true
-
titans.model.gpt.gpt.gpt2_small
-
-
-
-
true
-
-
-
titans.model.gpt.gpt.gpt2_small
-
0.00015
torch.optim.adam.Adam
0.01
1
-
1
-
-
Finished
kastan
only_3_gpu
7m 28s
-
1
-
-
-
60
-
-
1024
-
-
-
-
-
-
-
-
-
nccl
-
gpt2_configs/gpt2_vanilla.py
gpt2_configs/gpt2_vanilla.py
/u/kastan/colossal/data/train_data_FINAL.json
AMP_TYPE.NAIVE
true
-
titans.model.gpt.gpt.gpt2_small
-
-
-
-
true
-
-
-
titans.model.gpt.gpt.gpt2_small
-
0.00015
torch.optim.adam.Adam
0.01
1
-
1
-
-
Finished
kastan
7m 28s
-
1
-
-
-
60
-
-
1024
-
-
-
-
-
-
-
-
-
nccl
-
gpt2_configs/gpt2_vanilla.py
gpt2_configs/gpt2_vanilla.py
/u/kastan/colossal/data/train_data_FINAL.json
AMP_TYPE.NAIVE
true
-
titans.model.gpt.gpt.gpt2_small
-
-
-
-
true
-
-
-
titans.model.gpt.gpt.gpt2_small
-
0.00015
torch.optim.adam.Adam
0.01
1
-
1
-
-
Finished
kastan
7m 28s
-
1
-
-
-
60
-
-
1024
-
-
-
-
-
-
-
-
-
nccl
-
gpt2_configs/gpt2_vanilla.py
gpt2_configs/gpt2_vanilla.py
/u/kastan/colossal/data/train_data_FINAL.json
AMP_TYPE.NAIVE
true
-
titans.model.gpt.gpt.gpt2_small
-
-
-
-
true
-
-
-
titans.model.gpt.gpt.gpt2_small
-
0.00015
torch.optim.adam.Adam
0.01
1
-
1
-
-
1-1
of 1