Add files using upload-large-folder tool

.gitattributes

@@ -58,3 +58,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+metadata/train_manifest.jsonl filter=lfs diff=lfs merge=lfs -text

metadata/chunk_summary.json

@@ -0,0 +1,15 @@
+{
+  "study": "maze2d_action_chunking_NONCOT_actiononly_GT_ENVFEEDBACK",
+  "chunk_size": 1,
+  "difficulty": "easy",
+  "date_tag": "20260707_perseg",
+  "source_root": "/data/home/raychai/hf_datasets/maze2d_easy_native256_stopreq_stepmsg_fixedstart_ordered_20260606_stepmsg_fixedstart_v2",
+  "total_rows": 100000,
+  "prompt_contract": "SAME instruction as the CoT arm (imported cot_instruction: rules + total-budget sentence + env-feedback cadence + 'Starting maze:')",
+  "format": "MULTI-TURN: user(rules+budget+cadence, frame0 ctx); per chunk assistant({K actions}, NO think/no imagined frames); BETWEEN chunks user(ENV_FEEDBACK_LABEL + REAL chunk-end frame, loss=0 INPUT); K='inf' -> ONE assistant turn (full plan), ZERO feedback",
+  "env_feedback_label": "Action executed.",
+  "imagination": "NONE -- pure action policy; assistant generates no frames (no MSE targets)",
+  "loss": "CE on assistant action text only; frame0 + Env-Feedback frames = loss=0 INPUT context",
+  "loader": "VLMGymDatasetSokobanInterleavedCoTEnvFB (loss type 'interleaved_cot_envfb'; assistant has no images -> CE-only turn)",
+  "jpeg_quality": 95
+}

metadata/train_manifest.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac273374862a47aa7ee9ceac293090273d491a835b0532fd4940b410cf1081d7
+size 154408451

training/shard_00000.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77501e942dd3ccfbbd17e85e7425d36f0d60d030b5bf970bc57cfa782aab34fd
+size 387144259

training/shard_00001.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0459073717754236cdb97da167feadd9c9ee3c28472e74aba2b7ddcf5d23a6f
+size 386303105

training/shard_00002.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:761dde82942e846dff6b077db059fcde163920d4fee3d0bc45e5b6d7e493730e
+size 388056800

training/shard_00003.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1c849b52ebcace713d14dedd95b0854cfe70c75a508f839a39eab0a7b854fa3
+size 387300326

training/shard_00004.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89e02a291cf5d1762870b52c63d4550ee788d461f3e36eec2933a4a658c69f11
+size 388306599

training/shard_00005.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c962c231add258299f5134a9f1e94c85de103eab3487b7dc157a5032053a0c5
+size 388389400

training/shard_00006.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22050399286f637765d3df29e9073dd1035281396f556a41b7014792e1a72e44
+size 388897666

training/shard_00007.jsonl.gz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:349298a64c4de844649a5c7c802eea3275597cce9c48bdf02c1054d5a4ccfa8a
+size 388123106