Kastan's group workspace
Group: Aug-05__13:14
Name
64 visualized
State
Notes
User
Tags
Created
Runtime
Sweep
BATCH_SIZE
LEARNING_RATE
LOG_PATH
NUM_EPOCHS
SEQ_LENGTH
TOTAL_BATCH_SIZE
VOCAB_SIZE
WARMUP_EPOCHS
WEIGHT_DECAY
clip_grad_norm
conda_env_name
data_dir
fp16.mode
gpt2_8B
gpt2_large
gpt2_medium
gpt2_xl
gradient_accumulation
model.checkpoint
model.decoder_dtype
model.embed_dtype
model.head_dtype
model.layernorm_dtype
model.max_position_embeddings
model.vocab_size
model_dtypes.decoder_dtype
model_dtypes.embed_dtype
model_dtypes.head_dtype
model_dtypes.layernorm_dtype
num_gpus_per_node
optimizer.lr
optimizer.weight_decay
parallel.pipeline
quant_gpt2_8B
quant_gpt2_micro
quant_gpt2_small
quant_gpt2_xl
total_gpus
MICRO_BATCH_SIZE
NUM_MICRO_BATCHES
PIPELINE_SIZE
TENSOR_PARALLEL_MODE
TENSOR_PARALLEL_SIZE
loss.type
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 11s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 2s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 11s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 11s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 3s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 12s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Failed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
28s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Failed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
28s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Failed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
46s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 3s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 12s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 3s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 4s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Failed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
51s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 9s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 4s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 11s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 4s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 3s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
Crashed
kastan
Aug-05__13:14
BATCH_SIZE32
MICRO_BATCH_SIZE=4
NUM_EPOCHS=4
NUM_MICRO_BATCHES=8
PP=2
SLURM=513867
TP=4
WORLD_SIZE=64
3m 1s
-
32
0.00015
./gpt2_2.5d_tp4_bs32_lr0.00015_accum2_clip_grad1.0/
4
1024
64
50304
1
0.01
1
col_ai_quant
/u/kastanday/LLM-Distributed-Quantization/datasets/small-gpt-dataset.json
AMP_TYPE.NAIVE
titans.model.gpt.gpt.gpt2_8B
-
-
titans.model.gpt.gpt.gpt2_xl
2
false
-
-
-
-
1024
50304
-
-
-
-
4
0.00015
0.01
4
-
-
-
-
64
4
8
2
2.5d
4
titans.loss.lm_loss.gpt_lmloss.GPTLMLoss
1-20
of 64