 Alexander Korolev
|
be65b44ddb
fix: stop containers in single rm -f step
|
il y a 4 mois |
|
Alexander Korolev
|
6ff1e710aa
feat: record cluster label on containers for auto-detection
|
il y a 4 mois |
|
Alexander Korolev
|
ef03b8c784
feat: auto-scale PP when pipeline_parallel_size == required_nodes
|
il y a 4 mois |
|
Alexander Korolev
|
04413a8991
feat: auto-scale TP when recipe runs on larger cluster
|
il y a 4 mois |
|
Alexander Korolev
|
25540b4f48
fix: allow recipe with fewer required_nodes to run on larger cluster
|
il y a 4 mois |
|
Alexander Korolev
|
9d54d3f94c
gemma4-12b: working recipe with distributed_backend=none
|
il y a 4 mois |
|
Alexander Korolev
|
03c293c4ca
refactor(docker): move health_check to submodule
|
il y a 4 mois |
|
Alexander Korolev
|
71855e9f45
feat(recipe): add Gemma 4 12B IT (MTP) recipe + --node flag for single-node serve
|
il y a 4 mois |
|
Alexander Korolev
|
17a4f0c39a
feat(i18n): localize all remaining module output strings
|
il y a 4 mois |
|
Alexander Korolev
|
54de1eb75d
fix: restore VLLM_LOGGING_COLOR=1 alongside TERM=xterm-256color
|
il y a 4 mois |
|
Alexander Korolev
|
679c256a60
refactor: resolved_hf_cache_dir returns PathBuf instead of String
|
il y a 4 mois |
|
Alexander Korolev
|
cd8d9d66d5
refactor: replace wildcard imports with explicit imports (docker/mod.rs)
|
il y a 4 mois |
|
Alexander Korolev
|
68132ce91a
feat: set TERM=xterm-256color on head node for colored logs
|
il y a 4 mois |
|
Alexander Korolev
|
227eb18915
feat: auto-set VLLM_LOGGING_COLOR=1 on head node (override via recipe [env])
|
il y a 4 mois |
|
Alexander Korolev
|
7b01e19ec2
fix: HF_HOME -> /root/.cache/huggingface (fixes stale cache)
|
il y a 4 mois |
|
Alexander Korolev
|
9241bdc682
fix: don't pass --distributed-executor-backend mp (auto-detected)
|
il y a 4 mois |
|
Alexander Korolev
|
6f2c33a82f
fix: only add --shm-size when recipe explicitly sets it
|
il y a 4 mois |
|
Alexander Korolev
|
d9c3f845c2
fix(deepseek): align parameters with working system
|
il y a 4 mois |
|
Alexander Korolev
|
2af1104e7f
feat: launcher auto-adds --host/--port; recipes no longer hardcode
|
il y a 4 mois |
|
Alexander Korolev
|
7ba894be67
feat: per-recipe shm_size + DeepSeek V4 Flash recipe
|
il y a 4 mois |
|
Alexander Korolev
|
9c0e7680a4
feat: set safe NCCL defaults (IGNORE_CPU_AFFINITY, NET, CROSS_NIC, DEBUG)
|
il y a 4 mois |
|
Alexander Korolev
|
1b02f53b07
feat: set GLOO_SOCKET_IFNAME and TP_SOCKET_IFNAME for vLLM
|
il y a 4 mois |
|
Alexander Korolev
|
bfe94fa39e
feat: pre-serve hooks + exec subcommand for cluster mass-execution
|
il y a 4 mois |
|
Alexander Korolev
|
5097d55c77
fix: redirect stderr in `cargo run logs --follow` docker pipeline
|
il y a 4 mois |
|
Alexander Korolev
|
d92da29eb5
fix: shell-quote entrypoint script so 'sh -c SCRIPT' survives bash word-split
|
il y a 4 mois |
|
Alexander Korolev
|
75860da26d
feat: test-sleep recipe streams logs; add --timestamps to docker logs
|
il y a 4 mois |
|
Alexander Korolev
|
2b0be29648
feat: cluster defaults, configurable init_port, --headless for workers
|
il y a 4 mois |
|
Alexander Korolev
|
afd2fc35bf
feat: switch SGLang to --dist-init-addr, add memlock/stack ulimits
|
il y a 4 mois |
|
Alexander Korolev
|
8194da83d4
fix: stop/logs use CLI recipe name, health uses correct label filter
|
il y a 4 mois |
|
Alexander Korolev
|
4d3c297c50
feat: add --rm to containers by default, --keep flag to persist
|
il y a 4 mois |
