From f9a863b574ff164adf09b5279236447fa83c2748 Mon Sep 17 00:00:00 2001
From: Eugene Vinitsky <eugenevinitsky@lab1034pc10.mtc.scps.nyu.edu>
Date: Sun, 31 May 2026 18:16:49 -0400
Subject: [PATCH 1/2] pufferl: add --run-name with {date}/{seed} template +
 wire into single_agent yaml
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

wandb.init() previously did not set name=, so wandb assigned random names
like "flowing-wind-34". Adds a --run-name CLI flag (top-level config key)
that pufferl's WandbLogger renders against {date} (launch-time YYYY-MM-DD)
and {seed} (args.train.seed) placeholders before passing to wandb.init.

Default is None — wandb auto-name preserved. The single-agent speed-run
yaml opts in by setting run_name: "{date}_seed{seed}", giving identifiable
runs like 2026-05-31_seed0 without per-seed launcher logic.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---
 pufferlib/pufferl.py           |  7 +++++++
 scripts/launch_single_agent.sh | 24 ++++++++++++++++--------
 2 files changed, 23 insertions(+), 8 deletions(-)

diff --git a/pufferlib/pufferl.py b/pufferlib/pufferl.py
index 809225b7dc..14a270defd 100644
--- a/pufferlib/pufferl.py
+++ b/pufferlib/pufferl.py
@@ -1265,6 +1265,7 @@ def __init__(self, args, load_id=None, resume="allow"):
 
         wandb.init(
             id=load_id or wandb.util.generate_id(),
+            name=args.get("run_name") or None,
             project=args["wandb_project"],
             group=args["wandb_group"],
             allow_val_change=True,
@@ -2117,6 +2118,12 @@ def load_config(env_name, config_dir=None):
     parser.add_argument("--wandb", action="store_true", help="Use wandb for logging")
     parser.add_argument("--wandb-project", type=str, default="pufferlib")
     parser.add_argument("--wandb-group", type=str, default="debug")
+    parser.add_argument(
+        "--run-name",
+        type=str,
+        default=None,
+        help="Wandb run display name. Unset → wandb auto-generates one.",
+    )
     parser.add_argument("--neptune", action="store_true", help="Use neptune for logging")
     parser.add_argument("--neptune-name", type=str, default="pufferai")
     parser.add_argument("--neptune-project", type=str, default="ablations")
diff --git a/scripts/launch_single_agent.sh b/scripts/launch_single_agent.sh
index 8963721b69..a4717f2c58 100755
--- a/scripts/launch_single_agent.sh
+++ b/scripts/launch_single_agent.sh
@@ -26,13 +26,21 @@ PARTITION="${PARTITION:-h200_tandon}"
 TIME="${TIME:-720}"
 SEEDS="${SEEDS:-0:1:2}"
 PREFIX="${PREFIX:-$(date +%Y-%m-%d)_single_agent}"
+DATE_STAMP="$(date +%Y-%m-%d)"
 
 source "/scratch/$USER/venvs/pufferdrive/bin/activate"
-python scripts/submit_cluster.py \
-    --save_dir "/scratch/$USER/runs" \
-    --prefix "$PREFIX" \
-    --compute_config "$COMPUTE_CONFIG" \
-    --program_config "$PROGRAM_CONFIG" \
-    --container --heartbeat \
-    --account "$ACCOUNT" --partition "$PARTITION" --time "$TIME" \
-    --args "train.seed=$SEEDS"
+
+# One submission per seed so we can pass a per-seed run_name (wandb display
+# name like 2026-05-31_seed0). submit_cluster.py would expand the colon sweep
+# into N jobs internally, but it wouldn't vary --args run_name across them.
+IFS=':' read -ra SEED_LIST <<< "$SEEDS"
+for SEED in "${SEED_LIST[@]}"; do
+    python scripts/submit_cluster.py \
+        --save_dir "/scratch/$USER/runs" \
+        --prefix "$PREFIX" \
+        --compute_config "$COMPUTE_CONFIG" \
+        --program_config "$PROGRAM_CONFIG" \
+        --container --heartbeat \
+        --account "$ACCOUNT" --partition "$PARTITION" --time "$TIME" \
+        --args "train.seed=$SEED" "run_name=${DATE_STAMP}_seed${SEED}"
+done

From f217eca7049e69a0261c4aacfab8dc46fcf4c8e8 Mon Sep 17 00:00:00 2001
From: Eugene Vinitsky <eugenevinitsky@users.noreply.github.com>
Date: Sun, 31 May 2026 18:20:46 -0400
Subject: [PATCH 2/2] Update launch_single_agent.sh

---
 scripts/launch_single_agent.sh | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/scripts/launch_single_agent.sh b/scripts/launch_single_agent.sh
index a4717f2c58..e0e1135ab2 100755
--- a/scripts/launch_single_agent.sh
+++ b/scripts/launch_single_agent.sh
@@ -31,8 +31,7 @@ DATE_STAMP="$(date +%Y-%m-%d)"
 source "/scratch/$USER/venvs/pufferdrive/bin/activate"
 
 # One submission per seed so we can pass a per-seed run_name (wandb display
-# name like 2026-05-31_seed0). submit_cluster.py would expand the colon sweep
-# into N jobs internally, but it wouldn't vary --args run_name across them.
+# name like 2026-05-31_seed0)
 IFS=':' read -ra SEED_LIST <<< "$SEEDS"
 for SEED in "${SEED_LIST[@]}"; do
     python scripts/submit_cluster.py \