[docker] upgrade fla to 0.4.1 (#1452)

Author: zhuzilin

docker/Dockerfile

@@ -34,7 +34,7 @@ RUN git clone https://github.com/Dao-AILab/flash-attention.git && \
 
 RUN pip install git+https://github.com/ISEEKYAN/mbridge.git@89eb10887887bc74853f89a4de258c0702932a1c --no-deps
 
-RUN pip install flash-linear-attention==0.4.0
+RUN pip install flash-linear-attention==0.4.1
 RUN pip install tilelang -f https://tile-ai.github.io/whl/nightly/cu128/
 
 # TE does not have wheel on cuda 13 yet, thus need to install from source

docker/version.txt

@@ -1 +1 @@
-nightly-dev-20260118a
+nightly-dev-20260119a

slime/backends/megatron_utils/arguments.py

@@ -14,7 +14,8 @@ def set_default_megatron_args(args):
     # TODO: maybe change this after megatron has good fp8 support
     args.bf16 = not args.fp16
     # placeholders
-    args.seq_length = 4096
+    if args.seq_length is None:
+        args.seq_length = 4096
     args.max_position_embeddings = args.seq_length
     # TODO: revisit this when megatron(dev) have solved the optimizer-cpu-offload ckpt saving bug
     args.dist_ckpt_save_pre_mcore_014 = True