swanlab: swanlab api error ('Trace id: mj8hheoq8fb86zaj', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')

[Kenwwww]

确认项

• 我已确认我遇到的问题在文档中未有说明。
• 我已在 GitHub 上搜索过类似或已有的问题。
• 我已确认此问题并非由网络错误引起。

问题描述

我在使用Easyr1进行强化学习，采用swanlab进行中间的可视化追踪，基本没问题，但是日志中有若干条：
swanlab: swanlab api error ('Trace id: mj8hheoq8fb86zaj', 'POST
https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')
看起来是部分结果的格式问题，但是不知道怎么排查。请问下这个trace id可以怎么用于原因定位呢？或者可能是哪里的问题？

必须的复现代码

easyr1的配置代码如下，没啥特别的：
trainer:
total_epochs: 3
max_steps: null
project_name: agenticrl
experiment_name: agentic_rl
logger: ["swanlab"]
nnodes: 1
n_gpus_per_node: 1
max_try_make_batch: 20 # -1 means no limit

错误信息

(Runner pid=54046) swanlab: swanlab api error ('Trace id: mj8hh9u11brg13st', 'POST
(Runner pid=54046) https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')

[SAKURA-CAT]

看起来是一些格式问题，方便看看你swanlab.log的key是啥不？你应该log了一些媒体资源

[Kenwwww]

看起来是一些格式问题，方便看看你swanlab.log的key是啥不？你应该log了一些媒体资源

框架中和swanlab通讯的部分如下，我加了一个打印：

class SwanlabLogger(Logger):
    def __init__(self, config: dict[str, Any]) -> None:
        swanlab_key = os.getenv("SWANLAB_API_KEY")
        swanlab_dir = os.getenv("SWANLAB_DIR", "swanlab_log")
        swanlab_mode = os.getenv("SWANLAB_MODE", "cloud")
        if swanlab_key:
            swanlab.login(swanlab_key)

        swanlab.init(
            project=config["trainer"]["project_name"],
            experiment_name=config["trainer"]["experiment_name"],
            config={"UPPERFRAMEWORK": "EasyR1", "FRAMEWORK": "veRL", **config},
            logdir=swanlab_dir,
            mode=swanlab_mode,
        )

    def log(self, data: dict[str, Any], step: int) -> None:
        print('上传日志：',data)
        swanlab.log(data=data, step=step)

    def finish(self) -> None:
        swanlab.finish()

报错部分的日志切片如下：

(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bx5b41wa0ssbv', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bx7pv1ime17xi', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxa568td8kwm4', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxcm0cu71liko', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxf0hlgrxlgd6', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxhew1ei3wau8', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxjtuf4aqzdhb', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) 上传日志： {'global_seqlen/min': 150084, 'global_seqlen/max': 150084, 'global_seqlen/minmax_diff': 0, 'global_seqlen/balanced_min': 150084, 'global_seqlen/balanced_max': 150084, 'global_seqlen/mean': 150084.0, 'reward/overall': np.float64(0.3984375), 'reward/accuracy': np.float64(0.3984375), 'actor/kl_loss': np.float64(0.05709445849061012), 'actor/kl_coef': np.float64(0.01), 'actor/ppo_kl': np.float64(0.001341459111184425), 'actor/entropy_loss': np.float64(0.051496682378153004), 'actor/pg_clipfrac_higher': np.float64(0.0033462445562084516), 'actor/pg_clipfrac_lower': np.float64(0.0), 'actor/pg_loss': np.float64(-0.012466067882875602), 'actor/grad_norm': np.float64(2.037082076072693), 'perf/mfu_actor': np.float64(0.4622573132531507), 'perf/max_memory_allocated_gb': np.float64(14.689261436462402), 'perf/max_memory_reserved_gb': np.float64(22.498046875), 'perf/cpu_memory_used_gb': np.float64(53.581825256347656), 'actor/lr': np.float64(1e-06), 'critic/score/mean': 0.3984374701976776, 'critic/score/max': 1.0, 'critic/score/min': 0.0, 'critic/rewards/mean': 0.3984374701976776, 'critic/rewards/max': 1.0, 'critic/rewards/min': 0.0, 'critic/advantages/mean': 0.010471164248883724, 'critic/advantages/max': 0.4472030997276306, 'critic/advantages/min': -1.7888150215148926, 'critic/returns/mean': 0.010471164248883724, 'critic/returns/max': 0.4472030997276306, 'critic/returns/min': -1.7888150215148926, 'response_length/mean': 49.24687576293945, 'response_length/max': 63.0, 'response_length/min': 45.0, 'response_length/clip_ratio': 0.0, 'prompt_length/mean': 419.765625, 'prompt_length/max': 482.0, 'prompt_length/min': 329.0, 'prompt_length/clip_ratio': 0.0, 'timing_s/gen': 2.5821521158795804, 'timing_s/reward': 0.035545137943699956, 'timing_s/old': 3.8300256901420653, 'timing_s/ref': 3.386198187014088, 'timing_s/adv': 0.010195391019806266, 'timing_s/update_actor': 14.378616189816967, 'timing_s/step': 24.23459490505047, 'timing_per_token_ms/ref': 0.02256201984897849, 'timing_per_token_ms/gen': 0.1638525360669827, 'timing_per_token_ms/adv': 6.793123197546885e-05, 'timing_per_token_ms/reward': 0.0022555452721428997, 'timing_per_token_ms/update_actor': 0.0958037911424067, 'timing_per_token_ms/old': 0.025519213841196032, 'perf/total_num_tokens': 150084, 'perf/time_per_step': 24.23459490505047, 'perf/throughput': 6192.96508103474}

[SAKURA-CAT]

抱歉，没看出有啥问题🥺

你可以尝试一下设置一下SWANLAB_LOG_LEVEL=debug，我们会详细打印请求体，方便排查

[Kenwwww]

感觉不是swanlab的问题，是我把训练框架改写后的问题。  给大佬添麻烦了，我再自己探索下，目前这个问题没啥硬伤，没啥影响，谢谢佬关注。

…

---Original--- From: "Kang ***@***.***> Date: Thu, Dec 18, 2025 16:05 PM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [SwanHubX/SwanLab] swanlab: swanlab api error ('Trace id:mj8hheoq8fb86zaj', 'POST https://api.swanlab.cn/api/resources/presigned/put','400 Bad Request') (Issue #1379) SAKURA-CAT left a comment (SwanHubX/SwanLab#1379) 抱歉，没看出有啥问题🥺 你可以尝试一下设置一下SWANLAB_LOG_LEVEL=debug，我们会详细打印请求体，方便排查 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID: ***@***.***>

[mys210]

SwanHubX/self-hosted#85
@SAKURA-CAT
这有一个

[SAKURA-CAT]

SwanHubX/self-hosted#85 @SAKURA-CAT 这有一个

您用的私有化场景？不过看日志应该不是。https://api.swanlab.cn/api/resources/presigned/put这个请求是我们0.7.4新增的请求，用于媒体资源上传。您上传了媒体资源吗？根据您打印的data来看似乎没有

[Kenwwww]

SwanHubX/self-hosted#85 @SAKURA-CAT 这有一个

您用的私有化场景？不过看日志应该不是。https://api.swanlab.cn/api/resources/presigned/put这个请求是我们0.7.4新增的请求，用于媒体资源上传。您上传了媒体资源吗？根据您打印的data来看似乎没有

报错和我的一致，不过我的不是私有化场景。

[mys210]

我用的是私有化部署（v2.5.0），在0.7.4使用上传swanlab.Image失败，loss是正常上传的

[SAKURA-CAT]

看起来是一些格式问题，方便看看你swanlab.log的key是啥不？你应该log了一些媒体资源

框架中和swanlab通讯的部分如下，我加了一个打印：

class SwanlabLogger(Logger):
def init(self, config: dict[str, Any]) -> None:
swanlab_key = os.getenv("SWANLAB_API_KEY")
swanlab_dir = os.getenv("SWANLAB_DIR", "swanlab_log")
swanlab_mode = os.getenv("SWANLAB_MODE", "cloud")
if swanlab_key:
swanlab.login(swanlab_key)

    swanlab.init(
        project=config["trainer"]["project_name"],
        experiment_name=config["trainer"]["experiment_name"],
        config={"UPPERFRAMEWORK": "EasyR1", "FRAMEWORK": "veRL", **config},
        logdir=swanlab_dir,
        mode=swanlab_mode,
    )

def log(self, data: dict[str, Any], step: int) -> None:
    print('上传日志：',data)
    swanlab.log(data=data, step=step)

def finish(self) -> None:
    swanlab.finish()

报错部分的日志切片如下：

(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bx5b41wa0ssbv', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bx7pv1ime17xi', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxa568td8kwm4', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxcm0cu71liko', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxf0hlgrxlgd6', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxhew1ei3wau8', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) swanlab: swanlab api error ('Trace id: mj9bxjtuf4aqzdhb', 'POST https://api.swanlab.cn/api/resources/presigned/put', '400 Bad Request')                                                                                                                        
(Runner pid=4206) 上传日志： {'global_seqlen/min': 150084, 'global_seqlen/max': 150084, 'global_seqlen/minmax_diff': 0, 'global_seqlen/balanced_min': 150084, 'global_seqlen/balanced_max': 150084, 'global_seqlen/mean': 150084.0, 'reward/overall': np.float64(0.3984375), 'reward/accuracy': np.float64(0.3984375), 'actor/kl_loss': np.float64(0.05709445849061012), 'actor/kl_coef': np.float64(0.01), 'actor/ppo_kl': np.float64(0.001341459111184425), 'actor/entropy_loss': np.float64(0.051496682378153004), 'actor/pg_clipfrac_higher': np.float64(0.0033462445562084516), 'actor/pg_clipfrac_lower': np.float64(0.0), 'actor/pg_loss': np.float64(-0.012466067882875602), 'actor/grad_norm': np.float64(2.037082076072693), 'perf/mfu_actor': np.float64(0.4622573132531507), 'perf/max_memory_allocated_gb': np.float64(14.689261436462402), 'perf/max_memory_reserved_gb': np.float64(22.498046875), 'perf/cpu_memory_used_gb': np.float64(53.581825256347656), 'actor/lr': np.float64(1e-06), 'critic/score/mean': 0.3984374701976776, 'critic/score/max': 1.0, 'critic/score/min': 0.0, 'critic/rewards/mean': 0.3984374701976776, 'critic/rewards/max': 1.0, 'critic/rewards/min': 0.0, 'critic/advantages/mean': 0.010471164248883724, 'critic/advantages/max': 0.4472030997276306, 'critic/advantages/min': -1.7888150215148926, 'critic/returns/mean': 0.010471164248883724, 'critic/returns/max': 0.4472030997276306, 'critic/returns/min': -1.7888150215148926, 'response_length/mean': 49.24687576293945, 'response_length/max': 63.0, 'response_length/min': 45.0, 'response_length/clip_ratio': 0.0, 'prompt_length/mean': 419.765625, 'prompt_length/max': 482.0, 'prompt_length/min': 329.0, 'prompt_length/clip_ratio': 0.0, 'timing_s/gen': 2.5821521158795804, 'timing_s/reward': 0.035545137943699956, 'timing_s/old': 3.8300256901420653, 'timing_s/ref': 3.386198187014088, 'timing_s/adv': 0.010195391019806266, 'timing_s/update_actor': 14.378616189816967, 'timing_s/step': 24.23459490505047, 'timing_per_token_ms/ref': 0.02256201984897849, 'timing_per_token_ms/gen': 0.1638525360669827, 'timing_per_token_ms/adv': 6.793123197546885e-05, 'timing_per_token_ms/reward': 0.0022555452721428997, 'timing_per_token_ms/update_actor': 0.0958037911424067, 'timing_per_token_ms/old': 0.025519213841196032, 'perf/total_num_tokens': 150084, 'perf/time_per_step': 24.23459490505047, 'perf/throughput': 6192.96508103474}

你这个场景应该没有上传媒体资源比如Image吧？