Skip to content

fix small bugs #1128

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

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

fix small bugs #1128

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
b127f5d
update
Sep 30, 2025
04b3242
update warning
Oct 10, 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
4 changes: 2 additions & 2 deletions xtuner/v1/datasets/sft_tokenize_fn/openai.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,8 +37,8 @@ def __init__(
self.max_length = max_length

def __call__(self, item: dict | list, **kwargs) -> DataItem | CacheItem:
if isinstance(item, dict) and "messages" in item:
item = item["messages"]
if isinstance(item, dict) and ("messages" in item or "dialogs" in item):
item = item.get("messages", item.get("dialogs"))
messages = ChatMessages(messages=item)
tokenized = messages.tokenize(self.tokenizer, self.chat_template)

Expand Down
2 changes: 1 addition & 1 deletion xtuner/v1/ops/attn_imp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -213,7 +213,7 @@ def flash_attention(q, k, v, window_size=(-1, -1), s_aux=None, **kwargs) -> torc
if flash_attn_exception is not None:
traceback.print_exception(flash_attn_exception)
raise flash_attn_exception
attention_output = flash_attn_varlen_func(q, k, v, **kwargs)
attention_output = flash_attn_varlen_func(q, k, v, window_size=window_size, **kwargs)
else:
if flash_sink_attn_exception is not None:
traceback.print_exception(flash_sink_attn_exception)
Expand Down
2 changes: 1 addition & 1 deletion xtuner/v1/rl/base/worker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -187,7 +187,7 @@ def _build_ref_model(
if ref_model_fsdp_cfg is None:
ref_model_fsdp_cfg = FSDPConfig(recompute_ratio=0, cpu_offload=False, requires_grad=False)
model = model.fully_shard(ref_model_fsdp_cfg, float8_handler)
model.from_hf(hf_path=load_from)
model.from_hf(hf_path=load_from, strict=False)
Copy link
Collaborator

@hhaAndroid hhaAndroid Oct 15, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

这能改吗

Copy link
Collaborator Author

@LKJacky LKJacky Oct 16, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

不改的话，无法处理lm head 和 embedding 共享参数的情况

model.eval()
if float8_handler is not None:
# As the ref model is not updated, we only compute params' scales once
Expand Down
6 changes: 4 additions & 2 deletions xtuner/v1/train/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -986,8 +986,10 @@ def _log_step(
grad_norm: float,
):
"""Log the training step information."""
tgs = step_consumed_tokens / step_time
e2e_tgs = total_consumed_tokens / train_time
if step_consumed_tokens == 0:
logger.warning("step_consumed_tokens is 0 due to all tokens being padding tokens")
tgs = max(1, step_consumed_tokens / step_time)
e2e_tgs = max(1, total_consumed_tokens / train_time)
lr = self._lr_scheduler.get_last_lr()[0]

remaining_steps = self.total_step - self.cur_step
Expand Down
Loading