NousResearch/atropos
1
0
Fork 0
mirror of https://github.com/NousResearch/atropos.git synced 2026-04-19 12:57:58 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Fix typos in SLURM.md

Browse source
This commit is contained in:
HusseinAdeiza 2026-01-28 16:48:39 +01:00
parent e8fd85429f
commit 91cf15b933
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
SLURM.md
View file

@ -166,7 +166,7 @@ done
* `--api_addr`: This is the address of the main Atropos API server (usually started with `run-api`). The script needs this to fetch W&B project/group info and the current training step. In the example, we construct it assuming the API runs on the `training_node` (first node in the allocation) at port `8000` (the default for `run-api`). **Ensure this port is correct and accessible from the inference nodes.** * `--api_addr`: This is the address of the main Atropos API server (usually started with `run-api`). The script needs this to fetch W&B project/group info and the current training step. In the example, we construct it assuming the API runs on the `training_node` (first node in the allocation) at port `8000` (the default for `run-api`). **Ensure this port is correct and accessible from the inference nodes.**
* `--tp`: This should be the same tensor parallelism size (`TP_SIZE`) used when launching the inference servers. It tells the watcher how many server instances (ports 9000 to 9000 + `8 // TP_SIZE` - 1) to monitor on the local node. * `--tp`: This should be the same tensor parallelism size (`TP_SIZE`) used when launching the inference servers. It tells the watcher how many server instances (ports 9000 to 9000 + `8 // TP_SIZE` - 1) to monitor on the local node.
* `--node_num`: A unique integer identifying this specific inference node within the Slurm job. This helps distinguish the metrics from different nodes in W&B (e.g., `server/server_heath_0_0`, `server/server_heath_1_0`). The example script assigns sequential indices starting from 0. * `--node_num`: A unique integer identifying this specific inference node within the Slurm job. This helps distinguish the metrics from different nodes in W&B (e.g., `server/server_health_0_0`, `server/server_health_1_0`). The example script assigns sequential indices starting from 0.
**Important Notes:** **Important Notes:**