Upload folder using huggingface_hub

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/best_000039484_20215808_reward_1972.608.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a2b005e94ef0e9a3d6a8478a244162df970a96c08deeb4a89a59da4b80dcb16
+size 36514668

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000018292_9365504.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48ea5c56551bb4680383813944d30a8df82ac87c67a807010a2ab5c55cf4056c
+size 36515106

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000025552_13082624.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3694b1b3a0026ff4a69d7273e2f9ef4d03f2ccfc9c2f3f20cb2a1323357f1655
+size 36515106

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000032804_16795648.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:906c6e3dd39cdfcc7f37b02bc34fe5c2f2cf0dc8e8d5f96044593a6a9f82e701
+size 36515106

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000040808_20893696.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b7e1a896d591f6e7d32697f905945f605b2d842a1d588bd0162d7be4b5b902d
+size 36515106

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f2e5826dae30a804db695a2389f57f1ff0faf662bd3d16643cf180923f3f3051
+oid sha256:f431bf5c6b32a54d1c43cd0cbd6a5d912c046d1e4b6426d1c07b4862524eec90
 size 36515106

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/config.json

@@ -10,13 +10,13 @@
   "num_policies": 1,
   "async_rl": true,
   "serial_mode": false,
-  "batched_sampling": true,
+  "batched_sampling": false,
   "num_batches_to_accumulate": 2,
   "worker_num_splits": 2,
   "policy_workers_per_policy": 1,
   "max_policy_lag": 1000,
   "num_workers": 32,
-  "num_envs_per_worker": 2,
+  "num_envs_per_worker": 4,
   "batch_size": 1024,
   "num_batches_per_epoch": 2,
   "num_epochs": 2,
@@ -97,7 +97,7 @@
   "initial_stddev": 1.0,
   "use_env_info_cache": false,
   "env_gpu_actions": false,
-  "env_gpu_observations": false,
+  "env_gpu_observations": true,
   "env_frameskip": 1,
   "env_framestack": 1,
   "pixel_format": "CHW",
@@ -173,14 +173,14 @@
   "disable_reward_scaling": true,
   "frame_stack": 4,
   "max_episode_steps": 0,
-  "simulator": "envpool",
-  "envpool_batch_size": 2,
+  "simulator": "cpu",
+  "envpool_batch_size": 1,
   "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14/episode_metrics.jsonl",
   "vizdoom_init_max_parallel": 10,
   "vizdoom_init_lock_timeout_sec": 5.0,
   "vizdoom_init_lock_max_wait_sec": 60.0,
   "vizdoom_init_retry_jitter_sec": 0.25,
-  "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --env_gpu_observations False --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 2 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride6 seed14 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator envpool --envpool-batch-size 2 --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready",
+  "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride6 seed14 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready",
   "cli_args": {
     "algo": "APPO",
     "env": "doom_deadly_corridor",
@@ -191,10 +191,9 @@
     "seed": 14,
     "num_policies": 1,
     "async_rl": true,
-    "batched_sampling": true,
     "worker_num_splits": 2,
     "num_workers": 32,
-    "num_envs_per_worker": 2,
+    "num_envs_per_worker": 4,
     "batch_size": 1024,
     "num_batches_per_epoch": 2,
     "num_epochs": 2,
@@ -219,7 +218,6 @@
     "use_rnn": true,
     "rnn_size": 512,
     "rnn_type": "gru",
-    "env_gpu_observations": false,
     "with_wandb": true,
     "wandb_project": "latency-sensitive-bench",
     "wandb_group": "deadly-corridor-stride6-uniform_u2_6",
@@ -253,13 +251,12 @@
     "eval_raw_reward": true,
     "disable_reward_scaling": true,
     "frame_stack": 4,
-    "simulator": "envpool",
-    "envpool_batch_size": 2,
+    "simulator": "cpu",
     "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14/episode_metrics.jsonl",
     "env_frameskip": 1,
     "eval_env_frameskip": 1
   },
-  "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859",
+  "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d",
   "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git",
   "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride6/seed_14",
   "wandb_unique_id": "deadly-corridor-stride6-uniform_u2_6-s14"

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/episode_metrics.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c2d8de15177570f007667d873d229097a41793e30f3ee0f862e12ea2848e795f
-size 267655919
+oid sha256:86433fad5ed9b8dfe3bff215d1c95a3d8142a4a3e6ed16e268cb8bc51e5d7344
+size 262279469

factor_sweeps/deadly_corridor/observation_stride/train/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/git.diff

@@ -1,7 +1,26 @@
-diff --git a/starVLA b/starVLA
-index 538fd52..9d8c567 160000
---- a/starVLA
-+++ b/starVLA
-@@ -1 +1 @@
--Subproject commit 538fd523fe4fa4e7abce3201611919314bc6b664
-+Subproject commit 9d8c567188a3aa2a825296016cf17f3977101d8f
+diff --git a/latency_bench/run.py b/latency_bench/run.py
+index 163ca1b..e9016f3 100644
+--- a/latency_bench/run.py
++++ b/latency_bench/run.py
+@@ -152,6 +152,8 @@ def _record_stratified_replay_videos(
+     video_cfg = config["logging"]["video"]
+     if not video_cfg["enabled"]:
+         return
++    if not config["logging"].get("save_step_records", False):
++        return
+     if ExecutorMode(config["executor"]["mode"]) == ExecutorMode.REALTIME:
+         return
+     selections = select_episode_return_stratified(
+diff --git a/scripts/experiment_runner/run_manifest_jobs.py b/scripts/experiment_runner/run_manifest_jobs.py
+index 18376d9..646fe8f 100644
+--- a/scripts/experiment_runner/run_manifest_jobs.py
++++ b/scripts/experiment_runner/run_manifest_jobs.py
+@@ -192,6 +192,8 @@ def _resume_stage(job: dict[str, str], states: dict[str, str]) -> str:
+         return "upload"
+     if stage == "train_succeeded":
+         return "eval"
++    if stage == "eval_failed":
++        return "eval"
+     return "train"
+ 
+