Andrewzhao112's workspace
Runs
5
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.grad_offload
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.pretrained_tokenizer
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.include_ref
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
Crashed
-
andrewzhao112
all
code_f
code_i
code_o
codeio
executor-qwq
max_new
uniform
4d 14h 7m 30s
-
0.2
0.001
-1
false
false
false
0
1
0
low_var_kl
0.000001
0
640
constant
1
16384
16
384
false
fsdp
8
false
false
true
true
/home/fit/huangg/WORK/zqc/models/Qwen2.5-Coder-14B
true
true
true
0
false
16384
64
false
8
true
true
bfloat16
true
false
false
0.4
false
dummy_dtensor
Finished
-
andrewzhao112
all
code_f
code_i
code_o
codeio
executor-qwq
max_new
uniform
4d 3h 36m 52s
-
0.2
0.001
-1
false
false
false
0
1
0
low_var_kl
0.000001
0
384
constant
1
16384
8
384
false
fsdp
4
false
false
true
true
meta-llama/Llama-3.1-8B
true
true
true
0
false
16384
64
false
4
true
true
bfloat16
true
false
false
0.4
false
dummy_dtensor
Finished
-
andrewzhao112
all
code_f
code_i
code_o
codeio
executor-qwq
max_new
uniform
4d 19h 38m 28s
-
0.2
0.001
-1
false
false
false
0
1
0
low_var_kl
0.000001
0
384
constant
1
16384
16
384
false
fsdp
8
false
false
true
true
/home/fit/huangg/WORK/zqc/models/Qwen2.5-14B
true
true
true
0
false
16384
64
false
8
true
true
bfloat16
true
false
false
0.4
false
dummy_dtensor
Crashed
-
andrewzhao112
all
code_f
code_i
code_o
codeio
executor-qwq
max_new
uniform
4d 11h 53m 50s
-
0.2
0.001
-1
false
false
false
0
1
0
low_var_kl
0.000001
0
640
constant
1
16384
16
384
false
fsdp
4
false
false
true
true
/home/fit/huangg/WORK/zqc/models/Qwen2.5-Coder-7B
true
true
true
0
false
16384
64
false
4
true
true
bfloat16
true
false
false
0.4
false
dummy_dtensor
Crashed
-
andrewzhao112
all
code_f
code_i
code_o
codeio
executor-qwq
max_new
uniform
2d 17h 32m 19s
-
0.2
0.001
-1
false
false
false
0
1
0
low_var_kl
0.000001
0
960
constant
1
16384
16
384
false
fsdp
4
false
false
true
true
/home/fit/huangg/WORK/zqc/models/Qwen2.5-7B
true
true
true
0
false
16384
64
false
4
true
true
bfloat16
true
false
false
0.4
false
dummy_dtensor
1-5
of 5