完美demo解析异常

[CodeChook]

完美的demo能解析出json文件，但接着会报错.
官匹和hltv的demo没问题.
另外web如果能加一个保存和读取解析出的json文件的功能会方便一些

[Gary2005]

我更新过了，试试现在这个呢？

[CodeChook]

好像还是不行，我看完美的demo解析出的json里有很多NaN，报错也是NaN的错误，
像这样：Unexpected token 'N', ..." "X": NaN, "... is not valid JSON

[Gary2005]

有完整日志吗

[CodeChook]

阶段: 结果收集中
进度: 0%
回合进度: -
结果已生成，正在整理数据...
日志:
已接收任务，准备开始分析...
提示: 解析 demo 通常 30~60s，模型推理通常不到 10 分钟。
[warn] CUDA not available, falling back to CPU
Loaded alive from C:\Users\CC\cs-net\cs-net-models\alive\latest_alive_in_the_end.pt
Loaded kill from C:\Users\CC\cs-net\cs-net-models\nxt_kill\latest_nxt_kill.pt
Loaded death from C:\Users\CC\cs-net\cs-net-models\nxt_death\latest_nxt_death.pt
Loaded win from C:\Users\CC\cs-net\cs-net-models\win_rate\latest_winrate.pt
Loaded duel from C:\Users\CC\cs-net\cs-net-models\duel\latest_duel.pt

Extracting states by round: 0%| | 0/24 [00:00<?, ?it/s]
Extracting states by round: 4%|▍ | 1/24 [00:04<01:35, 4.16s/it]Error processing a round: 'Pandas' object has no attribute 'weapon_name'

Extracting states by round: 8%|▊ | 2/24 [00:04<00:45, 2.07s/it]
Extracting states by round: 12%|█▎ | 3/24 [00:07<00:53, 2.53s/it]
Extracting states by round: 17%|█▋ | 4/24 [00:11<00:58, 2.91s/it]
Extracting states by round: 21%|██ | 5/24 [00:15<01:02, 3.30s/it]
Extracting states by round: 25%|██▌ | 6/24 [00:19<01:03, 3.53s/it]
Extracting states by round: 29%|██▉ | 7/24 [00:23<01:05, 3.87s/it]
Extracting states by round: 33%|███▎ | 8/24 [00:29<01:09, 4.34s/it]
Extracting states by round: 38%|███▊ | 9/24 [00:32<01:02, 4.15s/it]
Extracting states by round: 42%|████▏ | 10/24 [00:36<00:54, 3.92s/it]
Extracting states by round: 46%|████▌ | 11/24 [00:39<00:49, 3.78s/it]
Extracting states by round: 50%|█████ | 12/24 [00:41<00:37, 3.11s/it]Error processing a round: Bomb carrier not found for tick 77767 in round 12

Extracting states by round: 54%|█████▍ | 13/24 [00:41<00:24, 2.26s/it]
Extracting states by round: 58%|█████▊ | 14/24 [00:44<00:22, 2.28s/it]
Extracting states by round: 62%|██████▎ | 15/24 [00:48<00:26, 2.94s/it]
Extracting states by round: 67%|██████▋ | 16/24 [00:51<00:23, 2.96s/it]
Extracting states by round: 71%|███████ | 17/24 [00:56<00:24, 3.45s/it]Error processing a round: Bomb carrier not found for tick 104860 in round 17

Extracting states by round: 75%|███████▌ | 18/24 [00:56<00:15, 2.58s/it]
Extracting states by round: 79%|███████▉ | 19/24 [01:00<00:15, 3.05s/it]
Extracting states by round: 83%|████████▎ | 20/24 [01:05<00:14, 3.67s/it]
Extracting states by round: 88%|████████▊ | 21/24 [01:09<00:11, 3.76s/it]
Extracting states by round: 92%|█████████▏| 22/24 [01:12<00:06, 3.48s/it]Error processing a round: Bomb carrier not found for tick 136639 in round 22

Extracting states by round: 96%|█████████▌| 23/24 [01:13<00:02, 2.61s/it]
Extracting states by round: 100%|██████████| 24/24 [01:17<00:00, 3.18s/it]
Extracting states by round: 100%|██████████| 24/24 [01:17<00:00, 3.24s/it]

rounds: 0%| | 0/24 [00:00<?, ?it/s]C:\Users\CC\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\nn\modules\transformer.py:515: UserWarning: The PyTorch API of nested tensors is in prototype stage and will change in the near future. We recommend specifying layout=torch.jagged when constructing a nested tensor, as this layout receives active development, has better operator coverage, and works with torch.compile. (Triggered internally at C:\actions-runner_work\pytorch\pytorch\pytorch\aten\src\ATen\NestedTensorImpl.cpp:182.)
output = torch._nested_tensor_from_mask(

rounds: 4%|▍ | 1/24 [00:08<03:14, 8.44s/it]
rounds: 12%|█▎ | 3/24 [00:14<01:31, 4.38s/it]
rounds: 17%|█▋ | 4/24 [00:19<01:35, 4.75s/it]
rounds: 21%|██ | 5/24 [00:26<01:45, 5.55s/it]
rounds: 25%|██▌ | 6/24 [00:33<01:45, 5.84s/it]
rounds: 29%|██▉ | 7/24 [00:42<01:59, 7.00s/it]
rounds: 33%|███▎ | 8/24 [00:59<02:39, 9.98s/it]
rounds: 38%|███▊ | 9/24 [01:10<02:33, 10.26s/it]
rounds: 42%|████▏ | 10/24 [01:19<02:20, 10.03s/it]
rounds: 46%|████▌ | 11/24 [01:28<02:04, 9.58s/it]
rounds: 50%|█████ | 12/24 [01:35<01:47, 8.93s/it]
rounds: 58%|█████▊ | 14/24 [01:40<00:58, 5.80s/it]
rounds: 62%|██████▎ | 15/24 [01:51<01:05, 7.23s/it]
rounds: 67%|██████▋ | 16/24 [01:59<00:59, 7.45s/it]
rounds: 71%|███████ | 17/24 [02:10<00:58, 8.29s/it]
rounds: 79%|███████▉ | 19/24 [02:21<00:34, 7.00s/it]
rounds: 83%|████████▎ | 20/24 [02:34<00:33, 8.41s/it]
rounds: 88%|████████▊ | 21/24 [02:42<00:25, 8.45s/it]
rounds: 92%|█████████▏| 22/24 [02:47<00:15, 7.57s/it]
rounds: 100%|██████████| 24/24 [02:53<00:00, 5.52s/it]
rounds: 100%|██████████| 24/24 [02:53<00:00, 7.23s/it]
Results saved to C:\Users\CC\cs-net\demo_analysis\outputs\634f7f7217994fda8b86db6a92f40a3a.json

分析失败: Unexpected token 'N', ..." "X": NaN,
"... is not valid JSON

这是我从web应用截的，是这个吗

[CodeChook]

我把所有NaN换成0，能读取了，但有一到两个人的数据缺失

[CodeChook]

哦不只一两个人，有些小局很多人没数据

[Gary2005]

很奇怪，能发一下demo文件吗。我没遇到过这种情况

[CodeChook]

链接: https://pan.baidu.com/s/1iCyzxao7lRbJezKN_fSsFw?pwd=1pe5 提取码: 1pe5
里面有两个demo和一个解析出的json文件

[Gary2005]

你能在游戏客户端正常播放吗？我试着播放的时候游戏画面一直在不正常抖动，我怀疑demo文件损坏了

[CodeChook]

确实，那个2026516demo会抖，但另一个叫w的demo没问题。我刚刚又试了那个叫w的demo，游戏里播放正常，但解析也是一样有问题

[CodeChook]

我发现出现“Error processing a round: Bomb carrier not found for tick 77767 in round 12”的小局都没有数据，json里是"error_round_13": "No states extracted"

[Gary2005]

你其他完美demo能正常运行吗，我觉得应该是这两个文件损坏了，所以没办法正常解析

[CodeChook]

但我的其它demo都可以在游戏里正常播放，只有那个demo可能是压缩的时候损坏的

[heipizhu4]

https://drive.google.com/file/d/18JtCBmB8_gvwnIPpvddPRuLDx095d5qd/view?usp=sharing
以上为demo，我这边遇到了类似的问题，同样是有nan
在将nan改为0后，报错Analysis failed: echarts is not defined
我猜测是有玩家掉线 且 有人玩bot搞出来的问题？