Initial K3 snapshot: 0.5B KDA/MLA/MoE train path
Standalone tree split from LLMRL/projects/kda. Includes Triton dt_bias backward fix, train_k3 --preset 0.5b, SFT, Docker runtime, and tests.
This commit is contained in:
@@ -0,0 +1,15 @@
|
||||
# Build context is projects/kda/. Keep the image a runtime, not a data dump.
|
||||
.venv/
|
||||
**/__pycache__/
|
||||
**/*.pyc
|
||||
**/*.pyo
|
||||
**/.pytest_cache/
|
||||
**/.ruff_cache/
|
||||
ckpts/
|
||||
*.pt
|
||||
notes/
|
||||
.git/
|
||||
.gitignore
|
||||
uv.lock
|
||||
pyrightconfig.json
|
||||
inspect_tensors.py
|
||||
Reference in New Issue
Block a user