Skip to content

Commit 5f7f9ed

Browse files
farhadrghfarhadrgh
authored
Merge branch 'main' into farhadr/evo2_cleanup
2 parents a395a8b + 58b4456 commit 5f7f9ed

6 files changed

Lines changed: 87 additions & 133 deletions

File tree

‎ci/benchmarks/partial-conv/esm2_pretrain.yaml‎

Lines changed: 13 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -1,31 +1,22 @@
11
scope: partial-conv
22
time_limit: 14400
3+
key_segments:
4+
# Modify keys to be renamed (str) or excluded (False) from run identifier. By default, all args under script_args are included.
5+
data_path: False
36
script_args:
47
# All arguments referenced in the script string must be specified here.
58
# Arguments not referenced in the script string must have the 'arg' field specified.
69
# See jet/core/configs.py for the specification of the configuration class
7-
workspace:
8-
value: /workspace/bionemo2
9-
key_segment: False
10-
data_path:
11-
value: /data/20240809_uniref_2024_03/data
12-
key_segment: False
13-
model:
14-
value: esm2
15-
variant:
16-
value: train
17-
config_name:
18-
value: 650M
19-
precision:
20-
value: [bf16-mixed]
21-
nodes:
22-
value: [4]
23-
gpus:
24-
value: 8
25-
batch_size:
26-
value: 16
27-
max_steps:
28-
value: 500000
10+
workspace: /workspace/bionemo2
11+
data_path: /data/20240809_uniref_2024_03/data
12+
model: esm2
13+
variant: train
14+
config_name: 650M
15+
precision: [bf16-mixed]
16+
nodes: [4]
17+
gpus: 8
18+
batch_size: 16
19+
max_steps: 500000
2920
stop_steps: 26500
3021
script: |-
3122
WANDB_API_KEY=$BIONEMO_WANDB_API_KEY ${variant}_${model} \

‎ci/benchmarks/partial-conv/evo2_pretrain.yaml‎

Lines changed: 16 additions & 25 deletions
Original file line numberDiff line numberDiff line change
@@ -1,19 +1,20 @@
11
scope: partial-conv
22
time_limit: 14400
3+
key_segments:
4+
# Modify keys to be renamed (str) or excluded (False) from run identifier. By default, all args under script_args are included.
5+
data_path: False
6+
clip_grad: False
7+
lr: False
8+
min_lr: False
9+
wu_steps: False
310
script_args:
411
# All arguments referenced in the script string must be specified here.
512
# Arguments not referenced in the script string must have the 'arg' field specified.
613
# See jet/core/configs.py for the specification of the configuration class
7-
workspace:
8-
value: /workspace/bionemo2
9-
key_segment: False
10-
data_path:
11-
value: /data/evo2
12-
key_segment: False
13-
model:
14-
value: evo2
15-
variant:
16-
value: train
14+
workspace: /workspace/bionemo2
15+
data_path: /data/evo2
16+
model: evo2
17+
variant: train
1718
config_name: 1b
1819
precision: fp8
1920
gpus: 8
@@ -26,22 +27,12 @@ script_args:
2627
tp: 1
2728
seq_len: 8192
2829
acc_grad: 1
29-
clip_grad:
30-
value: 250
31-
key_segment: False
30+
clip_grad: 250
3231
seed: 3735928559
33-
lr:
34-
value: 0.00015
35-
key_segment: False
36-
min_lr:
37-
value: 0.000015
38-
key_segment: False
39-
wu_steps:
40-
value: 5000
41-
key_segment: False
42-
wd:
43-
value: 0.1
44-
key_segment: False
32+
lr: 0.00015
33+
min_lr: 0.000015
34+
wu_steps: 5000
35+
wd: 0.1
4536
script: |-
4637
WANDB_API_KEY=$BIONEMO_WANDB_API_KEY ${variant}_${model} \
4738
-d /workspace/bionemo2/sub-packages/bionemo-evo2/examples/configs/full_pretrain_shortphase_config.yaml \

‎ci/benchmarks/partial-conv/geneformer_pretrain.yaml‎

Lines changed: 19 additions & 30 deletions
Original file line numberDiff line numberDiff line change
@@ -1,38 +1,27 @@
11
scope: partial-conv
22
time_limit: 14400
3+
key_segments:
4+
# Modify keys to be renamed (str) or excluded (False) from run identifier. By default, all args under script_args are included.
5+
data_path: False
6+
val_check_interval: False
7+
lr: False
38
script_args:
49
# All arguments referenced in the script string must be specified here.
510
# Arguments not referenced in the script string must have the 'arg' field specified.
611
# See jet/core/configs.py for the specification of the configuration class
7-
workspace:
8-
value: /workspace/bionemo2
9-
key_segment: False
10-
data_path:
11-
value: /data/cellxgene_scdl
12-
key_segment: False
13-
model:
14-
value: geneformer
15-
variant:
16-
value: train
17-
config_name:
18-
value: geneformer_config
19-
precision:
20-
value: [bf16-mixed]
21-
nodes:
22-
value: [2]
23-
gpus:
24-
value: 8
25-
batch_size:
26-
value: 32
27-
max_steps:
28-
value: 37000
29-
lr:
30-
value: 0.001
31-
val_check_interval:
32-
value: 500
33-
acc:
34-
value: 1
35-
12+
workspace: /workspace/bionemo2
13+
data_path: /data/cellxgene_scdl
14+
model: geneformer
15+
variant: train
16+
config_name: geneformer_config
17+
precision: [bf16-mixed]
18+
nodes: [2]
19+
gpus: 8
20+
batch_size: 32
21+
max_steps: 37000
22+
lr: 0.001
23+
val_check_interval: 500
24+
acc_grad: 1
3625
script: |-
3726
WANDB_API_KEY=$BIONEMO_WANDB_API_KEY ${variant}_${model} \
3827
--data-dir ${data_path} \
@@ -55,6 +44,6 @@ script: |-
5544
--wandb-job-type=${pipeline_label} \
5645
--cosine-rampup-frac 0.004331629559040111 \
5746
--cosine-hold-frac 0.021658147795200554 \
58-
--accumulate-grad-batches ${acc} \
47+
--accumulate-grad-batches ${acc_grad} \
5948
--precision ${precision} \
6049
--disable-checkpointing;

‎ci/benchmarks/perf/esm2_pretrain.yaml‎

Lines changed: 5 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1,15 +1,14 @@
11
scope: perf
22
time_limit: 1800
3+
key_segments:
4+
# Modify keys to be renamed (str) or excluded (False) from run identifier. By default, all args under script_args are included.
5+
data_path: False
36
script_args:
47
# All arguments referenced in the script string must be specified here.
58
# Arguments not referenced in the script string must have the 'arg' field specified.
69
# See jet/core/configs.py for the specification of the configuration class
7-
workspace:
8-
value: /workspace/bionemo2
9-
key_segment: False
10-
data_path:
11-
value: /data/20240809_uniref_2024_03/data
12-
key_segment: False
10+
workspace: /workspace/bionemo2
11+
data_path: /data/20240809_uniref_2024_03/data
1312
model: esm2
1413
variant: train
1514
config_name: 650M

‎ci/benchmarks/perf/evo2_pretrain.yaml‎

Lines changed: 16 additions & 25 deletions
Original file line numberDiff line numberDiff line change
@@ -1,19 +1,20 @@
11
scope: perf
22
time_limit: 1800
3+
key_segments:
4+
# Modify keys to be renamed (str) or excluded (False) from run identifier. By default, all args under script_args are included.
5+
data_path: False
6+
clip_grad: False
7+
lr: False
8+
min_lr: False
9+
wu_steps: False
310
script_args:
411
# All arguments referenced in the script string must be specified here.
512
# Arguments not referenced in the script string must have the 'arg' field specified.
613
# See jet/core/configs.py for the specification of the configuration class
7-
workspace:
8-
value: /workspace/bionemo2
9-
key_segment: False
10-
data_path:
11-
value: /data/evo2
12-
key_segment: False
13-
model:
14-
value: evo2
15-
variant:
16-
value: train
14+
workspace: /workspace/bionemo2
15+
data_path: /data/evo2
16+
model: evo2
17+
variant: train
1718
config_name: 1b
1819
precision: fp8
1920
gpus: 8
@@ -23,22 +24,12 @@ script_args:
2324
cp: 1
2425
seq_len: 8192
2526
acc_grad: 1
26-
clip_grad:
27-
value: 250
28-
key_segment: False
27+
clip_grad: 250
2928
seed: 3735928559
30-
lr:
31-
value: 0.00015
32-
key_segment: False
33-
min_lr:
34-
value: 0.000015
35-
key_segment: False
36-
wu_steps:
37-
value: 5000
38-
key_segment: False
39-
wd:
40-
value: 0.1
41-
key_segment: False
29+
lr: 0.00015
30+
min_lr: 0.000015
31+
wu_steps: 5000
32+
wd: 0.1
4233
products:
4334
- nodes: 2
4435
pp: 1

‎ci/benchmarks/perf/geneformer_pretrain.yaml‎

Lines changed: 18 additions & 25 deletions
Original file line numberDiff line numberDiff line change
@@ -1,33 +1,25 @@
11
scope: perf
22
time_limit: 3600
3+
key_segments:
4+
# Modify keys to be renamed (str) or excluded (False) from run identifier. By default, all args under script_args are included.
5+
data_path: False
6+
val_check_interval: False
7+
lr: False
38
script_args:
49
# All arguments referenced in the script string must be specified here.
510
# Arguments not referenced in the script string must have the 'arg' field specified.
611
# See jet/core/configs.py for the specification of the configuration class
7-
workspace:
8-
value: /workspace/bionemo2
9-
key_segment: False
10-
data_path:
11-
value: /data/cellxgene_scdl
12-
key_segment: False
13-
model:
14-
value: geneformer
15-
variant:
16-
value: train
17-
config_name:
18-
value: geneformer_config
19-
precision:
20-
value: [bf16-mixed]
21-
gpus:
22-
value: 8
23-
max_steps:
24-
value: 1000
25-
lr:
26-
value: 0.001
27-
val_check_interval:
28-
value: 500
29-
acc:
30-
value: 1
12+
workspace: /workspace/bionemo2
13+
data_path: /data/cellxgene_scdl
14+
model: geneformer
15+
variant: train
16+
config_name: geneformer_config
17+
precision: [bf16-mixed]
18+
gpus: 8
19+
max_steps: 1000
20+
lr: 0.001
21+
val_check_interval: 500
22+
acc_grad: 1
3123
products:
3224
- nodes: 1
3325
batch_size: 32
@@ -52,10 +44,11 @@ script: |-
5244
--lr ${lr} \
5345
--create-tensorboard-logger \
5446
--result-dir=${tensorboard_dir} \
47+
--wandb-group=${model}_${variant}_${config_name}__${target} \
5548
--wandb-project ${wandb_project_name} \
5649
--wandb-job-type=${pipeline_label} \
5750
--cosine-rampup-frac 0.004331629559040111 \
5851
--cosine-hold-frac 0.021658147795200554 \
59-
--accumulate-grad-batches ${acc} \
52+
--accumulate-grad-batches ${acc_grad} \
6053
--precision ${precision} \
6154
--disable-checkpointing;

0 commit comments

Comments
 (0)