feat: add Qwen3.6-35B-A3B VLM finetune recipe#1882
Merged
snowmanwwg merged 1 commit intomainfrom Apr 16, 2026
Merged
Conversation
HuiyingLi
requested review from
ZhiyuLi-Nvidia,
adil-a,
akoumpa,
hemildesai,
jgerh,
pthombre and
snowmanwwg
as code owners
Adds a ready-to-run MedPix-VQA fine-tuning recipe for `Qwen/Qwen3.6-35B-A3B` under the existing `qwen3_5_moe` architecture (same custom model impl). Verified on 8×H100: 100 steps complete, loss 1.86 → ~1.5, peak mem 64 GiB/GPU. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> Signed-off-by: HuiyingLi <willwin.lee@gmail.com>
HuiyingLi
force-pushed
the
huiyingl/add_qwen3_6_35b_config
branch
from
7a16517 to
f6a4f99
Compare
Contributor
Author
|
/ok to test f6a4f99 |
hemildesai
approved these changes
Apr 16, 2026
snowmanwwg
approved these changes
Apr 16, 2026
linnanwang
pushed a commit
that referenced
this pull request
Apr 24, 2026
Adds a ready-to-run MedPix-VQA fine-tuning recipe for `Qwen/Qwen3.6-35B-A3B` under the existing `qwen3_5_moe` architecture (same custom model impl). Verified on 8×H100: 100 steps complete, loss 1.86 → ~1.5, peak mem 64 GiB/GPU. Signed-off-by: HuiyingLi <willwin.lee@gmail.com> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
examples/vlm_finetune/qwen3_5_moe/qwen3_6_35b.yaml— a MedPix-VQA fine-tuning recipe forQwen/Qwen3.6-35B-A3B(next-gen Qwen3 MoE, sameqwen3_5_moearch).README.md, a row indocs/model-coverage/latest-models.md, and entries indocs/model-coverage/vlm/qwen/qwen3-5-vl.md(available models + example recipes).Test plan
Note: requires the collate_fn fix in #1799 for the default MedPix recipe (
local_batch_size: 1, max_length: 2048) to avoid occasional batches where the only sample exceedsmax_length.🤖 Generated with Claude Code