Salman-mohammadi's workspace
Runs
6
State
Notes
User
Tags
Created
Runtime
Sweep
batch_size
checkpointer._component_
checkpointer.checkpoint_dir
checkpointer.checkpoint_files
checkpointer.model_type
checkpointer.output_dir
compile
dataset._component_
dataset.add_eos
dataset.column
dataset.max_seq_len
dataset.source
dataset.split
dataset.train_on_input
device
dtype
enable_activation_checkpointing
forward_batch_size
gamma
gradient_accumulation_steps
init_kl_coef
initialise_value_head_from_reward_model
kl_coeff
kl_controller._component_
kl_controller.kl_coef
kl_horizon
kl_target
lmbda
log_every_n_steps
log_peak_memory_stats
loss._component_
loss.epsilon
loss.gamma
loss.lmbda
loss.value_clip_range
loss.value_coeff
max_generated_token
max_generated_tokens
metric_logger._component_
metric_logger.log_dir
metric_logger.project
min_response_length
model._component_
Killed
-
salman-mohammadi
35m 50s
-
256
torchtune.utils.FullModelHFCheckpointer
./target/7b/policy/
["pytorch_model-00001-of-00003.bin","pytorch_model-00002-of-00003.bin","pytorch_model-00003-of-00003.bin"]
MISTRAL
./target/7b/output/policy
true
torchtune.datasets.text_completion_dataset
false
prompt
-
trl-internal-testing/sentiment-trl-style
train
-
cuda
bf16
true
8
-
1
-
-
0.01
-
-
-
-
-
1
true
torchtune.modules.loss.PPOLoss
0.2
1
0.95
0.2
0.1
-
58
torchtune.utils.metric_logging.WandBLogger
./target/7b/output
torchtune_ppo
18
-
Crashed
-
salman-mohammadi
1h 38m 1s
-
64
torchtune.utils.FullModelHFCheckpointer
./target/7b/policy/
["pytorch_model-00001-of-00003.bin","pytorch_model-00002-of-00003.bin","pytorch_model-00003-of-00003.bin"]
MISTRAL
./target/7b/output/policy
true
torchtune.datasets.text_completion_dataset
false
prompt
-
trl-internal-testing/sentiment-trl-style
train
-
cuda
bf16
true
2
-
16
-
-
0.01
-
-
-
-
-
1
true
torchtune.modules.loss.PPOLoss
0.2
1
0.95
0.2
0.1
-
58
torchtune.utils.metric_logging.WandBLogger
./target/7b/output
torchtune_ppo
18
-
Crashed
-
salman-mohammadi
2h 14m 1s
-
256
torchtune.utils.FullModelHFCheckpointer
./target/dummy
["pytorch_model.bin"]
LLAMA2
./target/dummy/output
false
torchtune.datasets.text_completion_dataset
-
prompt
-
trl-internal-testing/sentiment-trl-style
train
-
cuda
bf16
true
128
-
4
-
true
-
torchtune.utils.ppo_utils.FixedKLController
0.01
-
-
-
1
false
torchtune.modules.loss.PPOLoss
0.2
1
0.95
0.2
0.1
-
58
torchtune.utils.metric_logging.WandBLogger
-
torchtune_ppo
-
-
Killed
-
salman-mohammadi
1h 51m 43s
-
256
torchtune.utils.FullModelHFCheckpointer
./target/dummy
["pytorch_model.bin"]
LLAMA2
./target/dummy/output
false
torchtune.datasets.text_completion_dataset
-
prompt
-
trl-internal-testing/sentiment-trl-style
train
-
cuda
bf16
true
128
-
4
-
true
-
torchtune.utils.ppo_utils.FixedKLController
0.01
-
-
-
1
false
torchtune.modules.loss.PPOLoss
0.2
1
0.95
0.2
0.1
-
53
torchtune.utils.metric_logging.WandBLogger
-
torchtune_ppo
-
-
Finished
-
salman-mohammadi
1h 11m 15s
-
512
torchtune.utils.FullModelHFCheckpointer
./target/dummy
["pytorch_model.bin"]
LLAMA2
./target/dummy/output
false
torchtune.datasets.text_completion_dataset
-
prompt
-
trl-internal-testing/sentiment-trl-style
train
-
cuda
bf16
true
128
-
8
-
true
-
torchtune.utils.ppo_utils.FixedKLController
0.01
-
-
-
1
false
torchtune.modules.loss.PPOLoss
0.2
1
0.95
0.2
0.1
-
53
torchtune.utils.metric_logging.WandBLogger
-
torchtune_ppo
-
-
Killed
-
salman-mohammadi
2h 3m 21s
-
32
torchtune.utils.FullModelHFCheckpointer
./target/dummy
["pytorch_model.bin"]
LLAMA2
./target/dummy/output
false
torchtune.datasets.text_completion_dataset
-
prompt
-
trl-internal-testing/sentiment-trl-style
train
-
mps
bf16
false
4
-
4
-
true
-
torchtune.utils.ppo_utils.FixedKLController
0.01
-
-
-
1
false
torchtune.modules.loss.PPOLoss
0.2
1
0.95
0.2
0.1
-
53
torchtune.utils.metric_logging.WandBLogger
-
torchtune_ppo
-
-
1-6
of 6