Jiayipan's workspace
Runs
89
Name
3 visualized
State
Notes
User
Tags
Created
Runtime
Sweep
data.batch_size
data.eval_batch_size
data.train_batch_size
data.train_data_path
data.val_data_path
logging.log_interval
logging.use_current_time
logging.verbose
logging.wandb_name
logging.wandb_project
model.gradient_checkpointing
model.master_port
model.mem_fraction_static
model.name
model.server_url
rollout.eval_temperature
rollout.group_size
rollout.mode
rollout.sample_temperature
rollout.temperature
training.grad_accum_chunk_size
training.grad_clip
training.inner_steps
training.kl_beta
training.learning_rate
training.log_probs_chunk_size
training.num_steps
training.output_dir
training.ppo_clip_ratio
training.save_steps
training.val_samples
training.validate_per_steps
training.length_penalty
training.length_penalty_power
training.length_penalty_type
training.target_length
training.max_allowed_threads
training.thread_penalty
avg_reward
loss_step_0
loss_step_1
max_advantage
mean_advantage
min_advantage
Finished
jiayipan
10h 18m 10s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
42609
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
1
0.1
0.60313
0.0044109
0.004411
2
0.020189
-2
Finished
jiayipan
23h 58m 2s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
43409
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
1
0.2
0.57188
0.0087423
0.0087416
2
-0.024583
-2
Killed
jiayipan
1d 1m
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
5840
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
4
0.2
0.6125
0.0000025261
0.0000018247
2
-0.087585
-2
Failed
jiayipan
19m 16s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
5840
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
4
0.2
-
-
-
-
-
-
Failed
jiayipan
20m 28s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
42609
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
1
0.1
-
-
-
-
-
-
Failed
jiayipan
20m 48s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
43409
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
1
0.2
-
-
-
-
-
-
Crashed
jiayipan
1d 40m 16s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
45419
0.9
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
1
1
one-side
2048
3
0.1
0.4375
0.001165
0.0011647
2
-0.066629
-2
Failed
jiayipan
17m 26s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
5840
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
4
0.2
-
-
-
-
-
-
Failed
jiayipan
19m 13s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
43409
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
1
0.2
-
-
-
-
-
-
Failed
jiayipan
20m 36s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
42609
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
1
0.1
-
-
-
-
-
-
Failed
jiayipan
7s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
43409
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
1
0.2
-
-
-
-
-
-
Finished
jiayipan
2m 26s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
43409
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
1
0.2
-
-
-
-
-
-
Finished
jiayipan
12m 45s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
45409
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
1
1
one-side
2048
1
0.1
-
-
-
-
-
-
Crashed
jiayipan
5m 1s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
45409
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
1
1
one-side
2048
1
0.1
-
-
-
-
-
-
Crashed
jiayipan
46s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
45840
0.8
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
-
-
-
-
4
0.2
-
-
-
-
-
-
Crashed
jiayipan
11h 44m
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
45419
0.9
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
1
1
one-side
2048
3
0.1
0.59062
0.001256
0.0012555
2
-0.059637
-2
Failed
jiayipan
18h 41m 27s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
45419
0.9
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
1
1
one-side
2048
3
0.1
0
0.0056644
0.0056637
0
0
0
Failed
jiayipan
18h 51m 1s
-
-
128
64
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
45848
0.9
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
-
25
1
1
one-side
4096
3
0.05
0.29375
0.0007413
0.00074082
2
-0.015967
-2
Failed
jiayipan
1m 3s
-
-
128
4
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
44519
0.9
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
16
25
1
1
one-side
2048
3
0.1
-
-
-
-
-
-
Failed
jiayipan
1m 3s
-
-
128
4
data/hsp_train_prefix.json
data/hsp_val_prefix.json
1
true
false
grpo_llama-hsp-v3_bs64_rollout5-lr1e-5-seq-weighted-kl0.01
tinyrl
true
44519
0.9
LM-Parallel/llama-hsp-v3
1
5
hsp
1
-
4
1
2
0.01
0.00001
16
400
checkpoints/hsp
0.2
50
16
25
1
1
one-side
2048
3
0.1
-
-
-
-
-
-
1-20
of 89