logger "Receive: obj=GenerateReqInput()" part with text rather than input_ids.

[9dian]

sglang 0.4.3 logger sample is as follows:

[2025-03-03 17:53:04] INFO: 10.27.1.1:65179 - "POST /v1/chat/completions HTTP/1.1" 200 OK
[2025-03-03 17:53:04] Receive: obj=GenerateReqInput(text=None, input_ids=[151646, 198, 5405, 1614, 25, 5538, 25713, 3795, 16, 25, 18, 17, 65, 198, 5405, 2400, 25, 220, 17, 15, 17, 20, 12, 15, 18, 12, 15, 18, 51, 15, 24, 25, 20, 18, 25, 16, 24, 13, 23, 21, 16, 57, 271, 2610, 525, 264, 10950, 17847, 13, 151644, 100633, 47815, 101562, 107380, 82894, 101437, 100968, 3837, 104719, 101914, 102513, 100371, 11319, 151645], input_embeds=None, image_data=None, sampling_params={'temperature': 0.1, 'max_new_tokens': None, 'min_new_tokens': 0, 'stop': None, 'stop_token_ids': None, 'top_p': 0.9, 'top_k': -1, 'min_p': 0.0, 'presence_penalty': 0.0, 'frequency_penalty': 0.0, 'repetition_penalty': 1.0, 'regex': None, 'ebnf': None, 'n': 1, 'no_stop_trim': False, 'ignore_eos': False, 'skip_special_tokens': True}, rid='ced00776101841e180bf04c8dbdc4ec2', return_logprob=False, logprob_start_len=-1, top_logprobs_num=0, return_text_in_logprobs=True, stream=True, log_metrics=True, modalities=[], lora_path=None, session_params=None, custom_logit_processor=None)
[2025-03-03 17:53:04 TP0] Prefill batch. #new-seq: 1, #new-token: 63, #cached-token: 1, cache hit rate: 1.41%, token usage: 0.00, #running-req: 0, #queue-req: 0
[2025-03-03 17:53:05 TP0] Decode batch. #running-req: 1, #token: 97, token usage: 0.00, gen throughput (token/s): 1.58, #queue-req: 0
[2025-03-03 17:53:06 TP0] Decode batch. #running-req: 1, #token: 137, token usage: 0.00, gen throughput (token/s): 55.91, #queue-req: 0

It show input_ids=[151646, 198, 5405, 1614, 25, 5538, 25713, 3795, 16, 25, 18, 17, 65, 198, 5405, 2400, 25, 220, 17, 15, 17, 20, 12, 15, 18, 12, 15, 18, 51, 15, 24, 25, 20, 18, 25, 16, 24, 13, 23, 21, 16, 57, 271, 2610, 525, 264, 10950, 17847, 13, 151644, 100633, 47815, 101562, 107380, 82894, 101437, 100968, 3837, 104719, 101914, 102513, 100371, 11319, 151645] in the log, could you please add text as substutition?

[minleminzui]

thanks for raising your issue, You can do this yourself using transformers

from transformers import AutoTokenizer


tokenizer = AutoTokenizer.from_pretrained("your model name")

input_ids = [151646, 198, 5405, 1614, 25, 5538, 25713, 3795, 16, 25, 18, 17, 65, 198, 5405, 2400, 25, 220, 17, 15, 17, 20, 12, 15, 18, 12, 15, 18, 51, 15, 24, 25, 20, 18, 25, 16, 24, 13, 23, 21, 16, 57, 271, 2610, 525, 264, 10950, 17847, 13, 151644, 100633, 47815, 101562, 107380, 82894, 101437, 100968, 3837, 104719, 101914, 102513, 100371, 11319, 151645]

text = tokenizer.decode(input_ids)

print(text)

[9dian]

Why not make the log more readable, then improve the Product usability...