Oliverstanley's workspace
Runs
4
State
Notes
User
Tags
Created
Runtime
Sweep
actor_rollout_ref.actor.clip_ratio
actor_rollout_ref.actor.entropy_coeff
actor_rollout_ref.actor.fsdp_config.fsdp_size
actor_rollout_ref.actor.fsdp_config.optimizer_offload
actor_rollout_ref.actor.fsdp_config.param_offload
actor_rollout_ref.actor.fsdp_config.wrap_policy.min_num_params
actor_rollout_ref.actor.grad_clip
actor_rollout_ref.actor.kl_loss_coef
actor_rollout_ref.actor.kl_loss_type
actor_rollout_ref.actor.optim.lr
actor_rollout_ref.actor.optim.lr_warmup_steps_ratio
actor_rollout_ref.actor.optim.total_training_steps
actor_rollout_ref.actor.optim.warmup_style
actor_rollout_ref.actor.ppo_epochs
actor_rollout_ref.actor.ppo_max_token_len_per_gpu
actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu
actor_rollout_ref.actor.ppo_mini_batch_size
actor_rollout_ref.actor.shuffle
actor_rollout_ref.actor.strategy
actor_rollout_ref.actor.ulysses_sequence_parallel_size
actor_rollout_ref.actor.use_dynamic_bsz
actor_rollout_ref.actor.use_kl_loss
actor_rollout_ref.hybrid_engine
actor_rollout_ref.model.enable_gradient_checkpointing
actor_rollout_ref.model.path
actor_rollout_ref.model.use_remove_padding
actor_rollout_ref.ref.fsdp_config.param_offload
actor_rollout_ref.ref.fsdp_config.wrap_policy.min_num_params
actor_rollout_ref.ref.log_prob_max_token_len_per_gpu
actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu
actor_rollout_ref.ref.log_prob_use_dynamic_bsz
actor_rollout_ref.ref.ulysses_sequence_parallel_size
actor_rollout_ref.rollout.disable_log_stats
actor_rollout_ref.rollout.do_sample
actor_rollout_ref.rollout.dtype
actor_rollout_ref.rollout.enable_chunked_prefill
actor_rollout_ref.rollout.enforce_eager
actor_rollout_ref.rollout.free_cache_engine
actor_rollout_ref.rollout.gpu_memory_utilization
actor_rollout_ref.rollout.ignore_eos
actor_rollout_ref.rollout.load_format
actor_rollout_ref.rollout.log_prob_max_token_len_per_gpu
actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu
actor_rollout_ref.rollout.log_prob_use_dynamic_bsz
Finished
starzmustdie
6h 45m 4s
-
0.2
0.001
-1
false
false
0
1
0.001
low_var_kl
0.000001
0
1500
constant
1
49152
8
32
false
fsdp
1
false
true
true
true
Qwen/Qwen2.5-3B-Instruct
true
true
0
49152
160
false
1
true
true
bfloat16
true
true
true
0.7
false
dummy_dtensor
49152
160
false
Crashed
starzmustdie
13h 32m
-
0.2
0.001
-1
false
false
0
1
0.001
low_var_kl
0.000001
0
1500
constant
1
49152
8
32
false
fsdp
1
false
true
true
true
Qwen/Qwen2.5-3B-Instruct
true
true
0
49152
160
false
1
true
true
bfloat16
true
true
true
0.7
false
dummy_dtensor
49152
160
false
Finished
starzmustdie
1m 55s
-
0.2
0.001
-1
false
false
0
1
0.001
low_var_kl
0.000001
0
1500
constant
1
49152
8
32
false
fsdp
1
false
true
true
true
Qwen/Qwen2.5-3B-Instruct
true
true
0
49152
160
false
1
true
true
bfloat16
true
true
true
0.7
false
dummy_dtensor
49152
160
false
Finished
starzmustdie
10m 46s
-
0.2
0.001
-1
false
false
0
1
0.001
low_var_kl
0.000001
0
500
constant
1
49152
8
32
false
fsdp
1
false
true
true
true
Qwen/Qwen2.5-3B-Instruct
true
true
0
49152
160
false
1
true
true
bfloat16
true
true
true
0.7
false
dummy_dtensor
49152
160
false
1-4
of 4