Levmckinney's group workspace
Group: llama-7b
Name
2 visualized
State
Notes
User
Tags
Created
Runtime
Sweep
constant
data
data.max_length
data.name
data.split
data.text_column
dist
dist.cpu_offload
dist.fsdp
loss
model
model.name
model.revision
model.slow_tokenizer
num_steps
opt
opt.lr_scale
opt.momentum
opt.optimizer
opt.weight_decay
opt.zero
output
per_gpu_batch_size
pre_ln
seed
separate_unembeddings
tokens_per_step
wandb
wandb_upload_checkpoints
bias_only
checkpoint_dir
checkpoint_freq
data.shuffle_seed
dist.nccl_timeout
dist.per_gpu_batch_size
model.precision
opt.warmup_steps
data.dataset_shuffle
data.dataset_shuffle_seed
dist.dataloader_shuffle
w_ce
w_kl
data.max_seq_len
bias_norm/0.ffn
Finished
-
levmckinney
1d 1h 48m 55s
-
-
-
-
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
KL
-
meta-llama/Meta-Llama-3-8B
main
false
250
-
1
0.9
SGD
0.001
false
/home/lev/tuned_lenses/meta-llama/Meta-Llama-3-8B
-
-
42
-
262144
tuned-lens-llama3-8b-redpj
-
false
/home/lev/tuned_lenses/meta-llama/Meta-Llama-3-8B/checkpoints
30
-
7200
1
float32
0
false
42
true
-
-
2048
1.25328
Finished
-
levmckinney
1d 1h 55m 7s
-
-
-
-
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
KL
-
meta-llama/Meta-Llama-3-8B-Instruct
main
false
250
-
1
0.9
SGD
0.001
false
/home/lev/tuned_lenses/meta-llama/Meta-Llama-3-8B-Instruct
-
-
42
-
262144
tuned-lens-llama3-8b-redpj
-
false
/home/lev/tuned_lenses/meta-llama/Meta-Llama-3-8B-Instruct/checkpoints
30
-
7200
1
float32
0
false
42
true
-
-
2048
1.21811
Finished
-
levmckinney
12h 22m 48s
-
-
-
2048
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
huggyllama/llama-7b
main
true
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-ce-0.95-blcxk-f6qmk
-
-
42
-
262144
tuned-lens-llama-7b-redpj-contrastive-ce-0.95-blcxk-f6qmk
-
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-ce-0.95-blcxk-f6qmk/checkpoints
30
-
7200
1
auto
0
false
42
true
-0.95
1
-
0.85205
Crashed
-
levmckinney
4h 33m 45s
-
-
-
2048
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
huggyllama/llama-7b
main
true
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-kl-0.95-r8rgp-p6jlg
-
-
42
-
262144
tuned-lens-llama-7b-redpj-contrastive-kl-0.95-r8rgp-p6jlg
-
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-kl-0.95-r8rgp-p6jlg/checkpoints
30
-
7200
1
auto
0
false
42
true
1
-0.95
-
0.45435
Finished
-
levmckinney
12h 14m 15s
-
-
-
2048
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
huggyllama/llama-7b
main
true
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-ce-0.8-tzq5l-rhjx4
-
-
42
-
262144
tuned-lens-llama-7b-redpj-contrastive-ce-0.8-tzq5l-rhjx4
-
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-ce-0.8-tzq5l-rhjx4/checkpoints
30
-
7200
1
auto
0
false
42
true
-0.8
1
-
0.60791
Finished
-
levmckinney
12h 13m 7s
-
-
-
2048
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
huggyllama/llama-7b
main
true
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-kl-0.8-9c7hl-vq22q
-
-
42
-
262144
tuned-lens-llama-7b-redpj-contrastive-kl-0.8-9c7hl-vq22q
-
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-kl-0.8-9c7hl-vq22q/checkpoints
30
-
7200
1
auto
0
false
42
true
1
-0.8
-
0.54248
Finished
-
levmckinney
12h 15m 3s
-
-
-
2048
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
huggyllama/llama-7b
main
true
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-kl-0.5-g8wk2-rj7b4
-
-
42
-
262144
tuned-lens-llama-7b-redpj-contrastive-kl-0.5-g8wk2-rj7b4
-
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-kl-0.5-g8wk2-rj7b4/checkpoints
30
-
7200
1
auto
0
false
42
true
1
-0.5
-
0.55273
Finished
-
levmckinney
12h 15m 48s
-
-
-
2048
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
huggyllama/llama-7b
main
true
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-ce-0.5-72vbr-btw7d
-
-
42
-
262144
tuned-lens-llama-7b-redpj-contrastive-ce-0.5-72vbr-btw7d
-
false
/root/checkpoints/tuned-lens-llama-7b-redpj-contrastive-ce-0.5-72vbr-btw7d/checkpoints
30
-
7200
1
auto
0
false
42
true
-0.5
1
-
0.55225
Finished
-
levmckinney
6h 50m 18s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-ce-0.8-gqnrh-rrkwj
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-ce-0.8-gqnrh-rrkwj
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-ce-0.8-gqnrh-rrkwj/checkpoints
30
-
7200
4
auto
0
false
42
true
-0.8
1
-
0.37061
Finished
-
levmckinney
6h 50m 2s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-kl-0.8-8ffl6-2pknm
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-kl-0.8-8ffl6-2pknm
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-kl-0.8-8ffl6-2pknm/checkpoints
30
-
7200
4
auto
0
false
42
true
1
-0.8
-
0.35669
Finished
-
levmckinney
6h 51m 11s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-kl-1-fdlmz-9c9cz
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-kl-1-fdlmz-9c9cz
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-kl-1-fdlmz-9c9cz/checkpoints
30
-
7200
4
auto
0
false
42
true
1
-1
-
0.55859
Finished
-
levmckinney
6h 49m 24s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-ce-1-7hc7m-mcb6h
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-ce-1-7hc7m-mcb6h
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-ce-1-7hc7m-mcb6h/checkpoints
30
-
7200
4
auto
0
false
42
true
-1
1
-
0.44531
Finished
-
levmckinney
6h 49m 44s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-ce-0.1-6ndq8-fw6vk
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-ce-0.1-6ndq8-fw6vk
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-ce-0.1-6ndq8-fw6vk/checkpoints
30
-
7200
4
auto
0
false
42
true
-0.1
1
-
0.36646
Finished
-
levmckinney
6h 50m 52s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-ce-0.5-67nzv-7z9b4
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-ce-0.5-67nzv-7z9b4
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-ce-0.5-67nzv-7z9b4/checkpoints
30
-
7200
4
auto
0
false
42
true
-0.5
1
-
0.36865
Finished
-
levmckinney
6h 50m 4s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-0.5-jp9sx-8dft6
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-0.5-jp9sx-8dft6
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-0.5-jp9sx-8dft6/checkpoints
30
-
7200
4
auto
0
false
42
true
1
-0.5
-
0.36353
Finished
-
levmckinney
6h 49m 45s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-0.01-ww76d-dmb64
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-0.01-ww76d-dmb64
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-0.01-ww76d-dmb64/checkpoints
30
-
7200
4
auto
0
false
42
true
1
-0.01
-
0.36499
Finished
-
levmckinney
6h 50m 16s
-
-
-
2048
["/root/datasets/pile/val.jsonl"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
EleutherAI/pythia-410m-deduped
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-0.1-j728w-gdhjj
-
-
42
-
262144
tuned-lens-pythia-410m-pile-contrastive-0.1-j728w-gdhjj
-
false
/root/checkpoints/tuned-lens-pythia-410m-pile-contrastive-0.1-j728w-gdhjj/checkpoints
30
-
7200
4
auto
0
false
42
true
1
-0.1
-
0.36499
Finished
-
levmckinney
12h 18m 12s
-
-
-
2048
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
LossChoice.CONTRASTIVE
-
huggyllama/llama-7b
main
true
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama-7b-redpj-wm5g5-kw5gj
-
-
42
-
262144
tuned-lens-llama-7b-redpj-wm5g5-kw5gj
-
false
/root/checkpoints/tuned-lens-llama-7b-redpj-wm5g5-kw5gj/checkpoints
30
-
7200
1
auto
0
false
42
true
1
-1
-
1.27637
Finished
-
levmckinney
23h 18m 26s
-
-
-
2048
["togethercomputer/RedPajama-Data-1T-Sample"]
train
text
-
false
false
LossChoice.KL
-
meta-llama/Llama-2-7b-chat-hf
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama2-7b-chat-redpj-jkhtr-d75vm
-
-
42
-
262144
tuned-lens-llama2-7b-chat-redpj-jkhtr-d75vm
-
false
/root/checkpoints/tuned-lens-llama2-7b-chat-redpj-jkhtr-d75vm/checkpoints
30
42
7200
1
float32
0
-
-
-
-
-
-
0.59027
Finished
-
levmckinney
14h 27m 4s
-
-
-
2048
["Anthropic/hh-rlhf"]
train
chosen
-
false
false
LossChoice.KL
-
meta-llama/Llama-2-7b-chat-hf
main
false
250
-
1
0.9
OptimizerOption.SGD
0.001
false
/root/checkpoints/tuned-lens-llama2-7b-chat-hh-rlhf-nwfg2-l7r6j
-
-
42
-
262144
tuned-lens-llama2-7b-chat-hh-rlhf-nwfg2-l7r6j
-
false
/root/checkpoints/tuned-lens-llama2-7b-chat-hh-rlhf-nwfg2-l7r6j/checkpoints
30
42
7200
1
float32
0
-
-
-
-
-
-
0.5746
1-20
of 46