Skip to content

[FIX] Remove hardcoded seq len #497

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 23, 2025
Merged

[FIX] Remove hardcoded seq len #497

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f6e0d94
enable larger seq len
Oct 23, 2025
4a4d1cd
update all configs
Oct 23, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion .meta/mast/qwen3_14b_mast.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,7 +61,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${local_batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down Expand Up @@ -106,6 +106,7 @@ ref_model:
flavor: 14B
hf_assets_path: /mnt/wsfuse/teamforge/hf/qwen3_14b
training:
seq_len: ${trainer.training.seq_len}
dtype: bfloat16
gc_freq: 1
compile:
Expand Down
3 changes: 2 additions & 1 deletion .meta/mast/qwen3_1_7b_mast.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,7 +62,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${local_batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down Expand Up @@ -108,6 +108,7 @@ ref_model:
hf_assets_path: /mnt/wsfuse/teamforge/hf/qwen3_1.7b
# hf_assets_path: hf://${model}
training:
seq_len: ${trainer.training.seq_len}
dtype: bfloat16
gc_freq: 1
compile:
Expand Down
3 changes: 2 additions & 1 deletion .meta/mast/qwen3_32b_mast.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,7 +61,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${local_batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down Expand Up @@ -106,6 +106,7 @@ ref_model:
flavor: 32B
hf_assets_path: /mnt/wsfuse/teamforge/hf/qwen3_32b
training:
seq_len: ${trainer.training.seq_len}
dtype: bfloat16
gc_freq: 1
compile:
Expand Down
3 changes: 2 additions & 1 deletion .meta/mast/qwen3_4b_mast.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -62,7 +62,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${local_batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down Expand Up @@ -108,6 +108,7 @@ ref_model:
hf_assets_path: /mnt/wsfuse/teamforge/hf/qwen3_4b
# hf_assets_path: hf://${model}
training:
seq_len: ${trainer.training.seq_len}
dtype: bfloat16
gc_freq: 1
compile:
Expand Down
3 changes: 2 additions & 1 deletion .meta/mast/qwen3_8b_mast.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,7 +61,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${local_batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down Expand Up @@ -106,6 +106,7 @@ ref_model:
flavor: 8B
hf_assets_path: /mnt/wsfuse/teamforge/hf/qwen3_8b
training:
seq_len: ${trainer.training.seq_len}
dtype: bfloat16
gc_freq: 1
compile:
Expand Down
6 changes: 3 additions & 3 deletions apps/grpo/qwen3_1_7b.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,8 +4,8 @@
# Global configuration
group_size: 8
local_batch_size: 16 # per-device batch size
max_req_tokens: 512
max_res_tokens: 512
max_req_tokens: 1024
max_res_tokens: 1024
model: "Qwen/Qwen3-1.7B"
off_by_n: 1 # Off by one by default

Expand Down Expand Up @@ -57,7 +57,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${local_batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down
2 changes: 1 addition & 1 deletion apps/grpo/qwen3_32b.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -60,7 +60,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${local_batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down
6 changes: 3 additions & 3 deletions apps/grpo/qwen3_8b.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,8 +4,8 @@
# Global configuration
group_size: 8
local_batch_size: 16 # per-device batch size
max_req_tokens: 512
max_res_tokens: 512
max_req_tokens: 1024
max_res_tokens: 1024
model: "Qwen/Qwen3-8B"
off_by_n: 1 # Off by one by default

Expand Down Expand Up @@ -53,7 +53,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${local_batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down
3 changes: 3 additions & 0 deletions src/forge/util/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,9 @@

from omegaconf import DictConfig, OmegaConf

# Add support for summing lists of numbers, e.g. ${sum:${max_req_tokens},${max_res_tokens}}
OmegaConf.register_new_resolver("sum", lambda *args: sum(args), replace=True)


def _has_component(node: Any) -> bool:
"""Check if a node has a _component_ field."""
Expand Down
2 changes: 1 addition & 1 deletion tests/integration_tests/fixtures/qwen3_1_7b_no_tp.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down
2 changes: 1 addition & 1 deletion tests/integration_tests/fixtures/qwen3_1_7b_tp.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ trainer:
warmup_steps: 1
training:
local_batch_size: ${batch_size}
seq_len: 2048
seq_len: ${sum:${max_req_tokens},${max_res_tokens}} # seq_len >= max_req_tokens + max_res_tokens
max_norm: 1.0
steps: 1000000
dtype: bfloat16
Expand Down
Loading