Does this framework currently support only the three Qwen3 models listed in the documentation? Does it support training for the Qwen3.5 and Qwen3.6 model series? If not, are there any plans to add support for them in the future?

[Flowerskill]

[PerryLink]

Checked against main @ 005e03b (last push 2026-07-09): the config tree supports exactly four models - config/dspark/{dspark_qwen3_4b, dspark_qwen3_8b, dspark_qwen3_14b, dspark_gemma4_12b}.py, with the same set under config/dflash and config/eagle3. So the documentation's three Qwen3 sizes are accurate, and Gemma4-12B is the fourth supported model. For the other questions in the tracker: deepseek-v4-flash-0731 (#89) and qwen3.8-flash-next (#86) have no config and no modeling glue in the tree, so they are not supported as of that sha; adding one needs a new model config plus (depending on architecture) modeling changes. Whether more models are planned is not stated in the repo【未核实】.