scaleswe_v2: refreshed release with current v1 tasks, uniform card and trace statistics

.gitattributes

@@ -31592,3 +31592,4 @@ termigen_v1/cached_images/bundles-v1/26815ec125eb548e8d2538cb4c92801113e8ba66a04
 termigen_v1/cached_images/bundles-v1/aed4cb2f7ceddd540c4b55dc23b611dd6527a887cbb8d4978ed4572ea1c31b0c/build_windows_dll_hijacking_medium-1abf78ca205a.sif/build_windows_dll_hijacking_medium-1abf78ca205a.sif filter=lfs diff=lfs merge=lfs -text
 termigen_v1/cached_images/bundles-v1/4923c868dac769cad9a8ef491db983d5daa10d68b881106fe9da5527b7448818/build_work_stealing_scheduler_medium-7f9e5f8dd1c4.sif/build_work_stealing_scheduler_medium-7f9e5f8dd1c4.sif filter=lfs diff=lfs merge=lfs -text
 termigen_v1/cached_images/bundles-v1/24dd24246b8fb450cb7caf53dfdeab1591d75bcbc4b6446685d77c5d9922926e/build_wrk_http_benchmarking_hard-5d31b0456350.sif/build_wrk_http_benchmarking_hard-5d31b0456350.sif filter=lfs diff=lfs merge=lfs -text
+scaleswe_v2/metadata/selection.json filter=lfs diff=lfs merge=lfs -text

scaleswe_v2/README.md

@@ -1,67 +1,78 @@
-# ScaleSWE v2
+# Scale-SWE v2
 
-## Dataset statistics
+## Dataset stats
 
-- **17,202 original tasks → 16,786 repaired v1 tasks → 292 kept in v2.** All kept tasks were patched in v1; v2 changes no task payloads.
-- **16,494 v1 tasks excluded for now:** no confirmed successful attempt in this snapshot, including unfinished and unattempted tasks. Recover excluded payloads from [v1](../scaleswe_v1); no duplicate archive Parquet.
-- 17,202 original Docker images; no consolidation. V2 uses the original [Docker Hub images](https://hub.docker.com/r/aweaiteam/scaleswe/tags) and reuses v1 Apptainer caches for 13 kept tasks.
+- **v1 tasks evaluated:** 7,656 of 16,786 so far
+- **Kept:** 1,240 (solved at least once by either model)
+- **Excluded:** 15,546 (not solved so far, including 9,130 not yet evaluated)
+- **Attempts:** pass@8 with Qwen3-30B Instruct and Qwen3-30B Thinking (89,061 graded so far)
 
-## Dataset characteristics
+## Pass@8 filtering
 
-Real repository issues with regression tests and reference fixes. Targets repository-level software engineering and debugging.
-
-<details><summary>Domains and capabilities</summary>
-
-Domains: software engineering, scientific computing and libraries. Capabilities: repository navigation, debugging, code editing, dependency management and regression testing.
+| Model | Tasks evaluated | Attempts | Tasks solved at least once |
+| --- | ---: | ---: | ---: |
+| Qwen3-30B Instruct | 7,656 | 60,832 | 1,105 |
+| Qwen3-30B Thinking | 3,564 | 28,229 | 375 |
+| Either model | 7,656 | 89,061 | 1,240 |
 
-</details>
+Snapshot of 2026-10-17. Instruct: 7,279 tasks have all 8 attempts; this run is still being completed. Thinking was stopped: 3,306 tasks have all 8 attempts, and its partial results are kept.
 
-## Validation and filtering
+### Successful attempts among kept tasks
 
-- **Environment reliability and rule-based task quality:** retains all repairs in [v1](../scaleswe_v1), including build/setup, reference-solution and no-op verification, and static-check repairs.
-- **Pass@8 filtering:** keep tasks solved at least once by either Qwen3-30B-A3B-Instruct-2507 or Qwen3-30B-A3B (Thinking). Evaluated content hashes match every kept v1 task.
-- **Snapshot (15 October 2026, 21:18 UTC):** 258 complete, 1,648 started but incomplete, and 14,880 without a recorded attempt. Complete means eight scored attempts per model. Includes **228 solved tasks with partial results**.
-- Instruct solved **218** tasks; Thinking solved **99**; their union is **292**.
+| Successful attempts | Instruct tasks | Thinking tasks |
+| --- | ---: | ---: |
+| 0/8 | 135 | 865 |
+| 1/8 | 552 | 202 |
+| 2/8 | 232 | 78 |
+| 3/8 | 142 | 40 |
+| 4/8 | 77 | 27 |
+| 5/8 | 47 | 13 |
+| 6/8 | 29 | 8 |
+| 7/8 | 17 | 2 |
+| 8/8 | 9 | 5 |
 
-| Model | Complete | Started, incomplete | No recorded attempt |
-| --- | ---: | ---: | ---: |
-| Instruct | 1,364 | 122 | 15,300 |
-| Thinking | 740 | 4 | 16,042 |
+## Select tasks by successful-attempt count
 
-### Successful attempts among kept tasks
+You can select tasks by how often they were solved, for example all tasks Instruct solved exactly once:
 
-Counts below are observed successes, not estimates for unfinished attempts.
+```python
+import pyarrow.parquet as pq
 
-| Successful attempts | Instruct | Thinking |
-| --- | ---: | ---: |
-| 0 | 74 | 193 |
-| 1 | 110 | 49 |
-| 2 | 49 | 20 |
-| 3 | 23 | 16 |
-| 4 | 11 | 7 |
-| 5 | 9 | 3 |
-| 6 | 8 | 2 |
-| 7 | 7 | 0 |
-| 8 | 1 | 2 |
+subset = pq.read_table("tasks.parquet", filters=[("instruct_successes", "=", 1)])
+```
 
-### Select by successful attempts
+## Traces
 
-`*_successes`, `*_graded_attempts` and `*_missing_attempts` distinguish success counts from incomplete coverage. The target is eight attempts per model.
+`traces/` holds all 89,760 recorded attempts, successful and failed, one row per attempt (Zstandard-compressed Parquet). 0 have no saved conversation (`trajectory_found=false`). Columns include `task_id`, `content_sha256` (the task version the attempt ran on), `model`, `attempt_id`, `reward`, `retained_v2` and the full `trajectory` and `result` JSON.
 
 ```python
-from huggingface_hub import hf_hub_download
-import pyarrow.parquet as pq
+import json
+import pyarrow.dataset as ds
 
-path = hf_hub_download("FWeindel/validated-tasks", "scaleswe_v2/tasks.parquet", repo_type="dataset")
-tasks = pq.read_table(path, filters=[("instruct_successes", ">=", 1), ("instruct_successes", "<=", 2)])
+traces = ds.dataset("traces", format="parquet", exclude_invalid_files=True)
+failed = traces.to_table(filter=(ds.field("reward") == 0) & ds.field("trajectory_found"))
+conversation = json.loads(failed["trajectory"][0].as_py())
 ```
 
-`metadata/selection.json` records every v1 task's snapshot results and retention decision. `patch.py` reproduces the task Parquet from pinned v1 data.
+Trajectory statistics (plots, command distribution) are in [trace_stats](trace_stats/README.md).
 
-## Resources
+Not yet checked with the repository-leak checks added on 2026-10-17 (leftover git history and file timestamps that could reveal the solution).
 
-Task environments and cached-image references are inherited from v1. See its README for the Apptainer, Python, Harbor and bridge versions and environment repairs.
+## Runtime
 
-## Traces
+- **Versions:** Apptainer **1.5.4**, Python **3.12.14**, Harbor **0.8.1**.
+- **Implementation:** [Marin Harbor `7faf878c14b6`](https://github.com/marin-community/harbor/tree/7faf878c14b6d72737579ec936ebf5f74ba3194d) + [our runtime patches](https://github.com/MLI-lab/OT-next-data/tree/main/harbor_patches).
+
+| Model | Mean time per attempt | Median time per attempt | Attempts measured |
+| --- | ---: | ---: | ---: |
+| Instruct | 268.42 s | 95.90 s | 61,248 |
+| Thinking | 452.24 s | 168.32 s | 28,512 |
+
+Compute used so far (all jobs, including failed and cancelled ones):
+
+| Run | CPU hours | GPU hours | Wall-clock |
+| --- | ---: | ---: | ---: |
+| Instruct (continuing) | 50,901 | 4,242 | 141 h |
+| Thinking (stopped) | 61,618 | 5,117 | 140 h |
 
-[Compressed trace Parquets](traces) contain **17,003 snapshot attempt records**, including successes and failures for kept and excluded tasks. **107 conversation payloads are not included** and are marked `trajectory_found=false`; **8 attempts have no score**. Each row contains the task ID, model, attempt ID, snapshot reward, retention flag and available full trajectory and result JSON.
+`patch.py` rebuilds `tasks.parquet` from the current v1 release and `metadata/selection.json`.

scaleswe_v2/metadata/earlier-version-count.json

@@ -0,0 +1 @@
+{"kept_evaluated_on_earlier_version": 0}

scaleswe_v2/tasks.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9f28c04b81f21d208abfb73f812237445f9e2f78765d924a6afbdfd056a4e998
-size 1539367
+oid sha256:98d620bb1a065de850a16be818eceebf813c3f08ed9f955ccc53fcb2b65c8512
+size 6741332

scaleswe_v2/trace_stats/README.md

@@ -0,0 +1,34 @@
+# Trajectory statistics
+
+89,760 attempts across 7,656 tasks of scaleswe_v2, kept and excluded, each counted once. Task success counts are in the [dataset overview](../README.md).
+
+| Metric | qwen3-30b-instruct-2507 | qwen3-30b-thinking |
+| --- | ---: | ---: |
+| Recorded attempts | 61,248 | 28,512 |
+| Median agent turns | 15 | 17 |
+| Median input tokens per attempt | 91,966 | 128,586 |
+| Median output tokens per attempt | 5,101 | 14,999 |
+| Median agent runtime (seconds) | 68 | 139 |
+| Median verifier runtime (seconds) | 7 | 8 |
+
+Medians use recorded measurements only.
+
+## Plots
+
+![Trace measurements](plots/all-models-trace-distributions.png)
+
+![Trace outcomes](plots/all-models-trace-outcomes.png)
+
+![Most frequent commands](plots/all-models-trace-commands.png)
+
+Commands are counted as submitted, not as successful executions.
+
+Per-model plots: [qwen3-30b-instruct-2507](plots/qwen3-30b-instruct-2507-trace-commands.png), [qwen3-30b-instruct-2507](plots/qwen3-30b-instruct-2507-trace-distributions.png), [qwen3-30b-instruct-2507](plots/qwen3-30b-instruct-2507-trace-outcomes.png), [qwen3-30b-thinking](plots/qwen3-30b-thinking-trace-commands.png), [qwen3-30b-thinking](plots/qwen3-30b-thinking-trace-distributions.png), [qwen3-30b-thinking](plots/qwen3-30b-thinking-trace-outcomes.png).
+
+## Data
+
+- [Summary](data/summary.json): overall and per-model statistics.
+- [Per-task metrics](data/per_task.parquet) and [per-attempt metrics](data/per_attempt.parquet), with nested metrics stored as JSON.
+- [Provenance](data/provenance.json): source files and deduplication.
+
+Attempts without a saved conversation: 0; without a verifier reward: 699. Their missing measurements stay empty.

scaleswe_v2/trace_stats/data/per_attempt.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de71bf38fad17ffcc5611799e93c1085533a817d4910f364bd42d2dfe6498b49
+size 14868871

scaleswe_v2/trace_stats/data/per_task.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:966c9286f9ee037c741919f062a27eb40774a3e138fd9bff771e4c907401631c
+size 6379856

scaleswe_v2/traces/manifest.json

@@ -1,10 +1,10 @@
 {
-  "snapshot_utc": "2026-10-15T21:18:23.797911+00:00",
+  "snapshot_utc": "2026-10-17T19:07:51.895120+00:00",
   "counts": {
-    "attempts": 17003,
-    "missing_conversations": 107,
-    "successful_attempts": 679,
-    "ungraded_attempts": 8
+    "attempts": 89760,
+    "missing_conversations": 0,
+    "successful_attempts": 3103,
+    "ungraded_attempts": 699
   },
-  "selection": "Exactly the attempt IDs and rewards in assets/selection.json; discarded attempts and later results are excluded."
+  "selection": "Exactly the attempt IDs and rewards in metadata/selection.json; content_sha256 is the task version each attempt ran on."
 }

scaleswe_v2/traces/part-0001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:66bd85fb337b43143a51a5c7f529ddcc40ee25083772dce01857ca23b4c8d836
-size 2045423
+oid sha256:ef502616044ff96902890f2bdce0bcbccdca7f95b176097d9fd303f1cbbb710d
+size 2470560

scaleswe_v2/traces/part-0002.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aca89c67ed7be2283bbf157fb8306b2058d09f95fc763efdc23a1b4f14472aee
-size 1711156
+oid sha256:95fa7e12634b2cf396c4d67f5e7eb33cfc3e3841eb0b9d87d6714159eb470832
+size 1988997

scaleswe_v2/traces/part-0003.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:92b16b67e62e6c1c25dd96c7c26af7d20d2ef56af85f998187abec62605b5d2e
-size 1744278
+oid sha256:3d1b1916d77133fa5ea940bbd5dacb1980e73de19969c7446583557f93bbc0c6
+size 2064734

scaleswe_v2/traces/part-0004.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3df23e23648389d6edfd394d9ca5ae82d79d8843961509924dbe57f0610721d5
-size 1387097
+oid sha256:47226ff033106cb868d309e2bef4007d8fd6e6950480a79c2178cbda85410565
+size 2223483

scaleswe_v2/traces/part-0005.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e31ebf9a8d89a37c88c44a46bf814fd5e57eff3fc2a0e7a98939053e8b6e2516
-size 1294570
+oid sha256:dac4a55f0015a0342106e6b6259b5d8b33a59bf9fd484e890fd48f00abd74132
+size 1978728

scaleswe_v2/traces/part-0006.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:747a75758c0695fa644c3a2ef44d62e83f48a4bfcb987323b34bfbdf540c2605
-size 1986817
+oid sha256:fcafa978667e00779442fe72b8d1b310ed176dfb702584f8ed48cbe114271843
+size 1916013

scaleswe_v2/traces/part-0007.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:81bf86523f53b51d244e6778fef65443c8fac03f72d2f2526ee4b55f07c948c6
-size 1502824
+oid sha256:b24445412f374e0cfb1f4cfef9a87522b3e90a63745e5be8b9753af32fb94da1
+size 2306953

scaleswe_v2/traces/part-0008.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4cfa813fd286d5cf82516487616505319bf86b9d987d9fb50b7ccc26f3f97c41
-size 1716478
+oid sha256:0f4cce6e9ad6571b47277592713565f2b65a4e70b365f8cd3a576675bce5baa2
+size 1429597

scaleswe_v2/traces/part-0009.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:44472f3dc3f7687abc801271ebcf0c38a25733ac23a87412335d59fc3fa89cda
-size 1402271
+oid sha256:992fb32dd379a546af0295f484f65b299086d4c19e7ad33f9c0099a1c102b91d
+size 2241174

scaleswe_v2/traces/part-0010.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7dc777e747be5b0a272860764823daaac3b1b9f4247a00057ccf1bf8e28010c9
-size 1313352
+oid sha256:58c167c754e03920b5646b7ee55fec50310b2f6736e6aef71503409a4c08b5bf
+size 1977581

scaleswe_v2/traces/part-0011.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a3478cedd0c26d94d2fde05719d62e1fccdd899dc4c402ff44adaeb774f2b927
-size 3307026
+oid sha256:92c78206cad6961872575bb6d6a83c850c191a451b986c9530b79f75bf625fdd
+size 2015935

scaleswe_v2/traces/part-0012.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:44c35c04d567300fa2117b8dd4a01d69c5522fdee2e186ec7dcee7288dadb78e
-size 2292019
+oid sha256:4277a8f57a574e85196fc39a680231ed3646dcded0a54531f843fba025dd4b24
+size 2187567

scaleswe_v2/traces/part-0013.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3aebf92267e01300c8a5c70cd588863a71fe947095028260b08381b697350e84
-size 2060232
+oid sha256:739650e816ec04df504e9972dbdd9fdbdb7dd37e08f0df74c462e7a098c36b95
+size 2131407

scaleswe_v2/traces/part-0014.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d012274a9dcb71c1c41f3694bb3c8948caf5d72a495645a4422f760ee7dd637f
-size 2017317
+oid sha256:398109aef636a413577498648fec68195f6d0616c65d9d2a66e896756c75c9f7
+size 1952691

scaleswe_v2/traces/part-0015.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9e5add30a00fe1361f4f0b386b99f78f03db9978e65fd31b8a2483ffa0395266
-size 2056987
+oid sha256:11e75f452a72616344eb3685dc22ad80b701053eed0565f0177fc59c3b79fd93
+size 2196729

scaleswe_v2/traces/part-0016.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:9719ea01facd65b6268aa89505f42095e6d51a295d14e1170a41a49f2cd6a635
-size 1612494
+oid sha256:ffcfa7666b4e45333dfb775ced5286e0b83c4a95a29345d9fcf61dcef410551d
+size 1997448

scaleswe_v2/traces/part-0017.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:84c64dd7ed416f2724eca5b15f5ce12769e04a3dbcece7d348c9fea802c9165f
-size 1633493
+oid sha256:338b1ad60756837f7f20034b5377a2fe4648dd94bf6d0eb8b9081af79d252427
+size 1788481

scaleswe_v2/traces/part-0018.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d61a342fe88c63089f0795fe7f8d2f91ab0890028d193d30fa0d70b842310599
-size 1467368
+oid sha256:b454dff43c9c5cf479d8fd2b88197a813708dec40048d4a7a53ce81abbd8c27e
+size 2307153

scaleswe_v2/traces/part-0019.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f1356977e6e01d7b358b20ba0707094359db7e55cd63b676284f086083bf1f3
-size 1290311
+oid sha256:aa070dbaf682ec50380f43e7293d3ffb9daaa4bb3b386b3c5d9009471c81620f
+size 2251636

scaleswe_v2/traces/part-0020.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0293fd6a76e43eac4c2a6f01d72f7a898e5865acc0e236f68f7c35e42782cd9c
-size 1418784
+oid sha256:c9f7d86991839759d8ca299f1db238705e8f74ab9416168eb03c29985b476d7c
+size 2036658

scaleswe_v2/traces/part-0021.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:212bfa07a951453f0be6935e86b42500156a8401d68a886cc01e1b42b462a908
-size 1801750
+oid sha256:3f3e7c56ee88aabf6096508434094b3ec24436cba26e991f61bca11a4a3a029f
+size 1857033

scaleswe_v2/traces/part-0022.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0756fe46643e499e04baf1b81e08f44fbb234f6a6799cc1a7453ae64cc68dd2b
-size 2697476
+oid sha256:9b7f439c587a2a5493acb5664f907c350ba0655a2f679ad622c140d3034d2854
+size 2180942

scaleswe_v2/traces/part-0023.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2a0f638990bd5e1ab725733cc592556ada6a536aacb18be4224a0041dbc2e8d4
-size 1510719
+oid sha256:a4c03a52568494b8c57d3445865c4a739c0d7e039c03a2c0b9fcbe9a5575b658
+size 2329646

scaleswe_v2/traces/part-0024.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a0567f12a53da8beb783f849e59335ec9fc7e4d563dabbc890bf0c5e31135b0d
-size 1490404
+oid sha256:997085710fb8900f397d0593eea465c5494ec8078275b35ae788116847c028d8
+size 1736927

scaleswe_v2/traces/part-0025.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:81796699a545a333626b1c691a7e97ad6d38c8b4a5da3d9ea17db1d7a58d2a5a
-size 2580032
+oid sha256:4606c5a873b71b9d724ef728dec11649800c99db9a8f07bcf57e985f9891d949
+size 2084803

scaleswe_v2/traces/part-0026.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:33661cc475eac0724cd4d8195ff629f03ffa8672992a5acea15a5bdfb96099b7
-size 1622866
+oid sha256:66c6ac32b3c2c322ca92d1277c8d020131f462b1ca40dbdba2307e3a2e810218
+size 2580205

scaleswe_v2/traces/part-0027.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:44bc8c462dbb031e6b5e06f0109dbc76235568687ba643720d7f89ec864da643
-size 2051133
+oid sha256:129ae00860c6ad00b224d639ca62fcccf99c121901c9357bae880e364c879e04
+size 1800739

scaleswe_v2/traces/part-0028.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:465388c829e5d4ebd8b62f18afaee05f236e927f009ee6c7a531accf814478de
-size 1783725
+oid sha256:18a7618fb621edfbbf34af4219b0870e4bc573795b06c80d1fa16b470023e147
+size 1999511

scaleswe_v2/traces/part-0029.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1ce4c4b7e02d4ccafd9ebeabfbfd5d11d2a41afe75b827bc25d639f9563636b6
-size 1809752
+oid sha256:11378b967dd92b1e5a743418f401dfdf2ee568a0205b297b12f147a397c929d5
+size 1966187

scaleswe_v2/traces/part-0030.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5f1ebeae961f948b3a4a00265dc14bf2c09c82532ef7dd9b66b6e287bed1a954
-size 1583575
+oid sha256:da87a770cc92775cf012532409aeb2d61e96b05378f18fa1f5bca0ce9ee48520
+size 2017804