Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
83 commits
Select commit Hold shift + click to select a range
6c9345a
Add Apple Silicon MLX routing
Manan17 Apr 9, 2026
462fb93
Add Apple Silicon MLX routing
Manan17 Apr 9, 2026
c588fe6
mlx with studio
Manan17 Apr 14, 2026
4cc55b7
mlx with studio
Manan17 Apr 14, 2026
48632f8
updating temporary install.sh
Manan17 Apr 14, 2026
2abf4cb
updating temporary install.sh
Manan17 Apr 14, 2026
1bf5853
adding t_v5 path
Manan17 Apr 14, 2026
4e5b229
adding t_v5 path
Manan17 Apr 14, 2026
728d08c
fixing vision training
Manan17 Apr 14, 2026
902fa4f
fixing vision training
Manan17 Apr 14, 2026
297adec
adding chat
Manan17 Apr 14, 2026
c420327
adding chat
Manan17 Apr 14, 2026
e654ac1
minor
mmathew23 Apr 18, 2026
a91f4f2
minor
mmathew23 Apr 18, 2026
77f53af
Adding export and fixing training issues, inference with lora adaptors
Manan17 Apr 19, 2026
67fdd3f
Adding export and fixing training issues, inference with lora adaptors
Manan17 Apr 19, 2026
f1673fc
Merge remote-tracking branch 'origin/fix/mlxvlmcompile' into mlx-appl…
Manan17 Apr 19, 2026
a2f2f40
Merge remote-tracking branch 'origin/fix/mlxvlmcompile' into mlx-appl…
Manan17 Apr 19, 2026
f08953d
fix: MLX worker pass load_in_4bit, override is_vlm based on dataset, …
Manan17 Apr 21, 2026
3dba5ec
fix: MLX worker pass load_in_4bit, override is_vlm based on dataset, …
Manan17 Apr 21, 2026
e1f096f
Merge mlx-apple-silicon into main
Manan17 Apr 21, 2026
4e42bfd
Merge mlx-apple-silicon into main
Manan17 Apr 21, 2026
de85036
update install.sh to point to main branch
Manan17 Apr 21, 2026
a258a23
update install.sh to point to main branch
Manan17 Apr 21, 2026
be99b87
fix: export returns 3 values (success, message, output_path) matching…
Manan17 Apr 21, 2026
9b811ad
fix: export returns 3 values (success, message, output_path) matching…
Manan17 Apr 21, 2026
a253b0f
fix(mlx): show training-process peak memory in Studio UI, not system-…
Manan17 Apr 28, 2026
07714f3
fix(mlx): show training-process peak memory in Studio UI, not system-…
Manan17 Apr 28, 2026
9b49c3c
fix(mlx): make is_bfloat16_supported detect M1/M2 (no native bf16)
Manan17 Apr 28, 2026
fd14a7d
fix(mlx): make is_bfloat16_supported() detect M1/M2 (no native bf16)
Manan17 Apr 28, 2026
dca214d
feat(mlx): wire training_type="Full Finetuning" through MLX worker
Manan17 Apr 28, 2026
4a425a4
feat(mlx): wire training_type="Full Finetuning" through MLX worker
Manan17 Apr 28, 2026
0ba5366
fix(mlx): pass save_method='merged_16bit' from Studio's export page
Manan17 Apr 29, 2026
8b014ec
fix(mlx): pass save_method='merged_16bit' from Studio's export page
Manan17 Apr 29, 2026
e0ab0b1
fix(studio): pass private to MLX push, return 3-tuples consistently
Manan17 Apr 29, 2026
8405e32
fix(studio): pass private to MLX push, return 3-tuples consistently
Manan17 Apr 29, 2026
dfdcf5d
studio wirings
mmathew23 Apr 30, 2026
1a9dbe1
studio wirings
mmathew23 Apr 30, 2026
b42426e
Merge pull request #5 from Manan17/feat/quant_config
Manan17 Apr 30, 2026
7c91739
Merge pull request #5 from Manan17/feat/quant_config
Manan17 Apr 30, 2026
477bd7e
fix(mlx): wire train_on_completions for VLM via per-template lookup
Manan17 Apr 30, 2026
a0d8822
fix(mlx): wire train_on_completions for VLM via per-template lookup
Manan17 Apr 30, 2026
2f4e038
wire in lora rslora, init lora weights, random_state
mmathew23 Apr 30, 2026
bb23d6f
wire in lora rslora, init lora weights, random_state
mmathew23 Apr 30, 2026
a9edbfa
loftq studio error message fix
mmathew23 Apr 30, 2026
f71656a
loftq studio error message fix
mmathew23 Apr 30, 2026
4d58b95
handle unknown optim and lr scheduler
mmathew23 Apr 30, 2026
d38f9bc
handle unknown optim and lr scheduler
mmathew23 Apr 30, 2026
f08b021
Merge pull request #6 from Manan17/update/peftkwargs
Manan17 May 1, 2026
300a58e
Merge pull request #6 from Manan17/update/peftkwargs
Manan17 May 1, 2026
b511646
Merge remote-tracking branch 'upstream/main'
Manan17 May 1, 2026
fc0298c
Merge remote-tracking branch 'upstream/main'
Manan17 May 1, 2026
ae8b5d6
feat(mlx): pass finetune_language/attention/mlp/vision flags to FastM…
Manan17 May 1, 2026
f0c00eb
feat(mlx): pass finetune_language/attention/mlp/vision flags to FastM…
Manan17 May 1, 2026
72d17a5
feat(mlx,ux): auto-imply finetune_language_layers when user picks att…
Manan17 May 2, 2026
0e62b46
feat(mlx,ux): auto-imply finetune_language_layers when user picks att…
Manan17 May 2, 2026
47d992b
fix(mlx): wire top_k, repetition_penalty, and VLM top_p through to ml…
Manan17 May 2, 2026
59b4232
fix(mlx): wire top_k, repetition_penalty, and VLM top_p through to ml…
Manan17 May 2, 2026
b76e0eb
feat(mlx): map format_type to MLX save_method, reuse local save dir f…
Manan17 May 3, 2026
5d51e08
feat(mlx): map format_type to MLX save_method, reuse local save dir f…
Manan17 May 3, 2026
d799e2e
Merge branch 'unslothai:main' into main
Manan17 May 4, 2026
f1a378c
Merge branch 'unslothai:main' into main
Manan17 May 4, 2026
8c51668
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] May 4, 2026
a1cd51c
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] May 4, 2026
760714c
restore install
Manan17 May 4, 2026
1856e32
restore install
Manan17 May 4, 2026
f40554d
fix(mlx): restore FastVisionModel as a distinct class
Manan17 May 5, 2026
6dfb9db
fix(mlx): restore FastVisionModel as a distinct class
Manan17 May 5, 2026
b1d5215
Merge branch 'main' into main
Imagineer99 May 5, 2026
bf56f80
Merge branch 'main' into main
Imagineer99 May 5, 2026
0b1baa1
Merge remote-tracking branch 'origin/main' into
danielhanchen May 5, 2026
dc0ca40
Studio: harden MLX training and export, restore GPU init guards
danielhanchen May 5, 2026
7b3b20d
Studio: regression tests for MLX training/export and GPU init ldconfi…
danielhanchen May 5, 2026
55bc5b7
Merge branch 'main' from origin fork into studio MLX work
danielhanchen May 5, 2026
7fdd4ce
Merge regression test suite into studio MLX harden branch
danielhanchen May 5, 2026
7d4e896
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] May 5, 2026
48ec7db
fix(mlx): cap inference memory + release wired on unload + tame worke…
Manan17 May 6, 2026
605bfaf
Merge branch 'main' of origin into Studio MLX work
danielhanchen May 6, 2026
df0cc6e
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] May 6, 2026
a159966
tests/studio: regression tests for the _IS_MLX dispatch gate
danielhanchen May 6, 2026
6468373
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] May 6, 2026
cad3597
Add AGPL-3.0 SPDX header to Studio MLX regression tests
danielhanchen May 6, 2026
4940597
Wrap MLX submodule imports with friendly install hint
danielhanchen May 6, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -228,3 +228,4 @@ setup_leo.sh
server.pid
*.log
package-lock.json
llama.cpp/
6 changes: 6 additions & 0 deletions install.sh
Original file line number Diff line number Diff line change
Expand Up @@ -1721,6 +1721,12 @@ else
fi
fi

# ── Install mlx-vlm on Apple Silicon (optional, for VLM training) ──
if [ "$OS" = "macos" ] && [ "$_ARCH" = "arm64" ]; then
substep "installing mlx-vlm (VLM training support)..."
run_install_cmd "install mlx-vlm" uv pip install --python "$_VENV_PY" mlx-vlm
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Make mlx-vlm install non-fatal on Apple Silicon

This block is labeled optional, but because install.sh runs with set -e, a transient uv pip install mlx-vlm failure aborts the entire installer on macOS arm64. That turns an optional VLM capability into a hard install dependency and can prevent Studio setup from completing for users who only need text/chat flows.

Useful? React with 👍 / 👎.

fi

# ── Run studio setup ──
tauri_log "STEP" "Running Studio setup"
# When --local, use the repo's own setup.sh directly.
Expand Down
4 changes: 2 additions & 2 deletions pyproject.toml
Original file line number Diff line number Diff line change
Expand Up @@ -89,7 +89,7 @@ huggingfacenotorch = [
]
huggingface = [
"unsloth[huggingfacenotorch]",
"unsloth_zoo>=2026.5.1",
"unsloth_zoo>=2026.4.8",
Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P1 Badge Raise the unsloth-zoo floor for MLX modules

This dependency now allows unsloth_zoo==2026.4.8, but the new Apple-Silicon import path immediately imports unsloth_zoo.mlx_trainer and unsloth_zoo.mlx_loader (unsloth/__init__.py, and the MLX training worker does the same). A fresh install can therefore resolve an older zoo build that satisfies this spec but lacks those MLX entry points, causing import unsloth or MLX training/inference startup to fail before users can recover. Keep the lower bound at the first zoo release that ships these modules (and mirror it in colab-new).

Useful? React with 👍 / 👎.

"torchvision",
"unsloth[triton]",
]
Expand Down Expand Up @@ -579,7 +579,7 @@ colab-ampere-torch220 = [
"flash-attn>=2.6.3 ; ('linux' in sys_platform)",
]
colab-new = [
"unsloth_zoo>=2026.5.1",
"unsloth_zoo>=2026.4.8",
"packaging",
"tyro",
"transformers>=4.51.3,!=4.52.0,!=4.52.1,!=4.52.2,!=4.52.3,!=4.53.0,!=4.54.0,!=4.55.0,!=4.55.1,!=4.57.0,!=4.57.4,!=4.57.5,!=5.0.0,!=5.1.0,<=5.5.0",
Expand Down
260 changes: 174 additions & 86 deletions studio/backend/core/export/export.py

Large diffs are not rendered by default.

Loading