Search results

Stale router replay instances caused by temporary GPTModel decoder in Qwen3VLGPTModel

Status: Open.

#156 In ISEEKYAN/mbridge;

· kimchou825 opened on Jul 7, 2026

get_model crashes on Megatron-Core >= 0.18 — ModelType.encoder_and_decoder was removed (util.py:52,83)

Status: Open.

#155 In ISEEKYAN/mbridge;

· chengcuiping opened on Jul 6, 2026

[BUG] get_thd_partitioned_indices not supported by Megatron

Status: Open.

#140 In ISEEKYAN/mbridge;

· yxh-y opened on May 18, 2026

Support for gradient_checkpointing in Vision Encoder to Reduce Memory Usage with High-Resolution Images on verl

Status: Open.

#114 In ISEEKYAN/mbridge;

· yangshurong opened on Apr 2, 2026

IndexError: too many indices for tensor of dimension 2

Status: Open.

#107 In ISEEKYAN/mbridge;

· wade0604 opened on Mar 27, 2026

NameError: name 'PretrainedConfig' is not defined

Status: Open.

#104 In ISEEKYAN/mbridge;

· wade0604 opened on Mar 25, 2026

After training the qwen3.5 based on Transformers 5.2.0, loading the model fails.

Status: Open.

#99 In ISEEKYAN/mbridge;

· JoyTim-777 opened on Mar 21, 2026

How can I measure memory profile with tools?

Status: Open.

#95 In ISEEKYAN/mbridge;

· robotsp opened on Mar 16, 2026

qwen3.5 27b save checkpoint oom

Status: Open.

#94 In ISEEKYAN/mbridge;

· mamazi0131 opened on Mar 13, 2026

[Question] MLP_MAPPING order issue resulting in missing expert_bias

Status: Open.

#88 In ISEEKYAN/mbridge;

· BeerTai opened on Mar 3, 2026

[Feature]GLM-5 support?

Status: Open.

#81 In ISEEKYAN/mbridge;

· wuyaoxuehun opened on Feb 14, 2026

Will mbridge support Phi from Microsoft in the future?

Status: Open.

#73 In ISEEKYAN/mbridge;

· Xchange7 opened on Jan 27, 2026