The previous commit introduced the single-device budget but wired it
into only the downloaded-group and Hub search sites. The LM Studio,
custom-folder, local-dir, live-search and exported-GGUF expanders
reachable from the Images/Video pickers still measured against the
summed multi-GPU total, as did the size-based GGUF row badge, so those
paths could still recommend a quant that OOMs on a single device. All
GgufVariantExpander call sites in HubModelPicker now share
expanderGpuGb, and the row badge derives the same task-scoped budget.
The generate callback omitted loraCapable from its dependencies, so when
an auto-compile flips supports_lora off mid-session the memoized handler
still sent the previously selected adapters and the next generation
failed with the backend's LoRA-not-supported error instead of omitting
adapters the UI had already hidden.
HubModelPicker's GGUF variant expanders, format lists and Hub row fit
hints measured against the summed multi-GPU total. When the picker is
task-scoped (Images/Video) the loaders place the whole pipeline on one
device, so a variant could be recommended as fitting and then OOM at
load; those sites now share the single-device budget the group fit gate
already uses, while chat pickers keep the summed total since llama.cpp
splits layers across devices.
The dataset labeling grid's Remove button relied on group-hover with no
group parent, leaving it permanently invisible to mouse users; the image
wrapper now carries the group class.
The studio dependency spec resolves gguf 0.19.0, whose gguf/utility.py
legitimately sends an HF_TOKEN Authorization header from the authenticated
Hugging Face download helper used by convert_hf_to_gguf; main's baseline
entry covers a different gguf version so the evidence hash differs.
Verified locally: the full studio shard scan exits 0 with the updated
baseline and zero unsuppressed CRITICAL or HIGH findings.
Register the dims the forward actually compiled with: image-conditioned
workflows (img2img, inpaint, upscale, edit) run at the input image's size,
not the slider's, so recording the slider values marked never-compiled
shapes as covered and warm restarts kept paying compile for the real one.
Validate a request-supplied transformer_prequant_path (existence plus the
UNSLOTH_ALLOW_LOCAL_PREQUANT_PATH allowlist) before treating prequant as
available at the resident-fit re-check: an unusable path skipped the dense
fit check up front and then fell back to materializing dense bf16 after
the previous pipeline was evicted, recreating the post-eviction OOM path.
Shared as usable_prequant_source, also used by the auto-policy planner.
A generation runs on a backend daemon thread and survives a page reload,
but the mount effect only probed the load progress, so reloading during a
generate showed an idle page that never picked up the finished clip until
a manual refresh. Hoist the generate poll loop out of handleGenerate and
re-enter it on mount when generate-progress reports an active job; merge a
terminal completed record into the gallery to cover the race with the
mount gallery fetch.
New releases of huggingface-hub (1.23.0) and openai (2.45.0) shifted or
added polling loops that the C2 polling/beaconing check flags, failing
all three pip scan-packages shards (studio 1, hf-stack 1, extras 3 new
CRITICAL findings) org-wide including on main.
Regenerated with scan_packages.py --write-baseline per CI shard (same
shard-to-requirements mapping and --with-deps as security-audit.yml)
and merged. All entries were manually reviewed at the resolved versions:
- huggingface-hub hf_api.py: create_repo 409-concurrency retry loop
body changed in 1.23.0; refreshed evidence hash. The loop POSTs to
the canonical Hub endpoint and retries only on a specific conflict
error. Benign client retry.
- openai beta/threads/runs/runs.py: create_and_poll run-status helper
refactored in 2.45.0 (Assistants deprecation annotations); refreshed
evidence hash. Documented polling helper against api.openai.com.
- openai beta/responses/responses.py: new beta websocket client whose
__aiter__ yields server events until the connection closes. New
entry; standard event-stream iterator, not beaconing.
- openai resources/responses/responses.py: evidence line number
refreshed only, hash unchanged.
The two dropped entries are the pre-refactor hashes of the same two
loops above; they no longer occur at the resolved versions. Verified
locally: all three shards exit 0 with 0 unsuppressed CRITICAL/HIGH
(hf-stack 120, studio 151, extras 99 suppressed).
* Auto-detect completion masking markers with template table fallback
Studio's train_on_completions previously relied only on the hardcoded
MODEL_TO_TEMPLATE_MAPPER / TEMPLATE_TO_RESPONSES_MAPPER tables and
silently disabled masking when a model was not in the table, so unmapped
models (LFM2-8B-A1B, DeepSeek, and others) trained on full sequences
without telling the user. Several mapped templates (glm, mistral, llama,
starling, zephyr, qwen3-thinking) also carried markers that mask every
assistant token, which made every row drop in the post-masking filter.
Both training callsites (CUDA trainer.py and MLX worker.py) now share
utils.datasets.completion_masking.apply_completion_masking:
- Try unsloth_zoo chat template auto-detection first; it raises loudly
when the template cannot be parsed and never masks the EOS token.
- gpt-oss models keep their manual markers so non-final assistant
<|end|> tokens stay trained, matching current behavior.
- If auto-detection raises, fall back to the template table exactly as
before.
- If the table also misses, emit an explicit user-visible warning that
completion masking could not be applied and full-sequence training
will occur, instead of a quiet log line.
The >30 percent dropped-rows safety net in trainer.py now guards the
auto path as well. Table consumers for inference and chat templates are
unchanged. Validated against one representative tokenizer for every
template in TEMPLATE_TO_RESPONSES_MAPPER plus the unmapped models:
no template regresses; unit tests cover the four decision paths.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Restrict masking fallback to marker detection failures
The auto branch wrapped the whole train_on_responses_only call, so a real
failure while applying the masking (dataset map, tokenization) was treated
as a detection miss and training silently proceeded on full sequences.
Detect markers separately via get_chat_template_parts (test seam via
detect_fn), then apply them with errors propagating, matching the manual
path. Tokenizers with preset unsloth marker attrs skip detection and call
bare so zoo reuses the stored parts.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Fail the run when applying completion masking raises
The helper already falls back internally on detection failures and returns
applied=False on a double miss, so an exception reaching the callsites is a
real failure applying the masking. Remove the callsite catches that
downgraded it to full-sequence training; the run now fails visibly instead.
Also use the explicit re-export alias form in utils/datasets/__init__.py for
the two new names, satisfying the import-hoist source lint.
* Import completion masking from its submodule
The import-hoist source lint counts only real name loads, so package-level
re-exports of the two new names cannot satisfy it. Import
apply_completion_masking from utils.datasets.completion_masking directly at
both callsites and leave utils/datasets/__init__.py untouched.
* Completion masking: gpt-oss renames and MLX raw/alpaca parity
Renamed or private gpt-oss checkpoints are name-detected as gpt-oss but miss
the exact-name table; default them to the gpt-oss template markers instead of
falling through to full-sequence training.
Gate the MLX masking call on not raw_text_mode and format_type != alpaca,
mirroring the CUDA path: raw/CPT text has no chat turns to mask and
Alpaca-rendered text lacks the tokenizer's chat markers.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Define raw_text_mode outside the MLX feature-detect block
With an older zoo lacking the append_eos config field, the masking
gate referenced raw_text_mode before assignment. Hoist the assignment
above the feature detection so both consumers see it.
* Gate MLX masking on the formatter's resolved format
format_type auto can resolve to alpaca or raw text; the masking skip
checked only the requested value, so auto-detected Alpaca data got
chat-template markers applied to rendered prompt text. Track the
final_format returned by format_and_template_dataset and gate on it,
matching the CUDA path.
* Unwrap the mlx-lm TokenizerWrapper before marker checks
The wrapper delegates plain reads to the wrapped HF tokenizer but hides
underscore attrs, so preset unsloth markers were invisible and detection
relied on the loader's call patch. Unwrap to the real tokenizer first,
as the zoo MLX resolver does.
* Tighten masking comments
* gpt-oss: auto-detect markers first like every other template
The quantized and BF16 gpt-oss checkpoints ship a chat template without
the channel final header, so the pinned manual markers match nothing
there and masking trained zero tokens. Auto-detection derives markers
from whichever template the checkpoint ships and keeps the final
terminator trained; the manual gpt-oss markers remain the detection
failure fallback, including for renamed checkpoints.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Tighten comments
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Studio: route models by CONFIG_MAPPING_NAMES instead of hardcoded tables
A model whose model_type is absent from an overlay's transformers cannot load
there, so a new MoE arch not yet in the tier tables gets routed to default and
fails (e.g. lfm2_moe, deepseek_v4). Add a static resolver that parses each
overlay's CONFIG_MAPPING_NAMES straight from source (AST only, no import, no
network, no trust_remote_code) and picks the lowest tier that ships the
model_type. Runs after the existing checks and only ever upgrades default, so
no existing routing changes and new archs no longer need a table edit.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Studio router: harden the CONFIG_MAPPING_NAMES resolver
- Resolve the default tier map from the base install, skipping any .venv_t5_*
sidecar on sys.path, so an in-process 5.x activation cannot make a 5.x-only
model look loadable by 4.x.
- Do not cache an overlay whose sidecar dir is absent, so a later call re-reads
it once provisioned instead of serving a stale empty map.
- Also collect model types added via CONFIG_MAPPING_NAMES.update({...}) and
**{...} unpacking, not just the literal assignment (5.10 uses both).
- Wrap the AST walk in the try/except so a malformed source can never crash tier
resolution.
- Feed the mapping fallback from _load_config_json so a config served from the
hub cache during a transient outage still routes new architectures.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Tighten comments
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
LFM2-8B-A1B and any other lfm2_moe checkpoint were missing from the
transformers tier tables, so they fell through to the default 4.57.x
sidecar, which does not register lfm2_moe and errors with
"not supported yet in transformers==4.57.6". Only lfm2_vl was listed.
Add Lfm2MoeForCausalLM / lfm2_moe to the 5.3.0 tier (lfm2_moe is
registered in transformers 5.3.0). get_transformers_tier now returns
530 for LFM2-8B-A1B and the model loads and trains as expected.
The image bench drove pipe() directly with no cache reset between the
warmup and measured prompts, while the production backend clears the
FBCache residuals before each generation. diffusers keys those residuals
on the long-lived transformer and never resets them itself, so step 0 of
each measured prompt compared its first-block residual against the
previous prompt's final one, a state production never runs. Mirror the
backend's _reset_step_cache (best-effort, no-op for uncached configs and
for SDXL's unet) inside _generate so both warmup and measured passes
start clean, and note that pre-fix FBCache rows may overstate results.
diffusion_compile_cache: auto mode now saves the Mega-cache bundle after the
first compiled generation (UNSLOTH_DIFFUSION_COMPILE_CACHE_SAVE=0 opts out), so
users get warm restarts without the distributor env; a bundle hit starts clean
(no pointless rewrite of the just-loaded artifacts) and explicit mode 1/on keeps
the distributor-style re-save. New register_shape + manifest shape coverage: a
STATIC compile produces new artifacts per (width, height, batch), so the
generate path registers each generation's shape and an uncovered shape
re-dirties the context, growing the bundle to cover every shape the session
used. Measured (B200, real backend): Qwen-Image deferred gen-3 hitch 29.1 ->
22.2 s warm with bit-identical output (7.9 MB bundle, ~0.5 s save); SDXL gen-3
115.7 -> 24.7 s and a mid-session 768px recompile 65.8 -> 12.6 s (bundle 63.6 ->
98.7 MB after the 768 re-save).
diffusion_speed: U-Net denoisers (UNet2DConditionModel; no _repeated_blocks, so
the regional compile never reached them) now get a whole-module STATIC
torch.compile on the default tier, plus fused QKV projections and a compiled VAE
decode. Measured on SDXL (30 steps / 7.0 / 1024px, 4 prompts, LPIPS vs the
bit-exact reference): 6.16 -> 3.14 s end to end (1.96x) at LPIPS 0.035, steady
state 0.70-0.88 s/image through the real backend. Rejected on measurement:
dynamic=True whole-module (366 s compile for 39.3 ms/step vs static's 73 s for
26.9), regional BasicTransformerBlock only (45.0 ms/step; ResNet convs stay
eager), max-autotune + inductor flags (25.9 ms/step for a 445 s warmup),
channels-last UNet alone (neutral). DiT tiers unchanged: fused QKV measured
exactly neutral under the regional compile (Qwen-Image 6.53 vs 6.52 s), so it
stays max-only there, and the DiT VAE decode stays eager (a few % of a DiT
generation). compiled_shapes_are_static tells the cache layer which loads are
per-shape (max tier, U-Net whole-module).
diffusion: register each generation's shape with the compile cache before the
save, pass pipe.unet to the cache fingerprint when the pipe has no transformer,
and correct the transformer_quant resolved reason on dense loads (it claimed a
GGUF transformer was loaded on every non-quantized pipeline load).
Tests: 333 passing across the related suites (speed 42, compile_cache 27, cache
40, precision 20, backend, base_precision, transformer_quant, memory); ruff
clean. Full measurement record: outputs/image_optim_round2_audit.md.
Reject an image whose name differs from an existing one only by extension
case (cat.PNG vs cat.png): the stems are exactly equal, so on a
case-sensitive filesystem both files land and both resolve to one cat.txt
caption sidecar, silently sharing and corrupting the caption. Stem case
variants (Pic.png vs pic.png) stay exempt: they are one file on
case-insensitive filesystems and write separate sidecars on Linux.
Treat an empty precision_modes list on a DiT family as the backend's
deliberate cannot-train signal (a non-bf16 CUDA GPU fails the trainer's
preflight for every mode) instead of falling back to the full mode list:
the precision selector disables and the start button reads not supported,
so the form no longer offers a run that always 400s. An absent field still
means an older backend and keeps the fallback.
Hide the Images page Reapply button when no reload target is known: a
resident GGUF or single_file model discovered by refresh carries no
checkpoint filename in status, so clicking was a silent no-op. A resident
full pipeline keeps the button (it reloads by repo id alone).
A load request that is rejected up front (validation error, gated repo,
training guard) leaves the previously loaded model resident, but both the
Images and Video pages had already pointed lastLoad at the failed pick, so
Reapply and the resident-default seeding retried the wrong model. Snapshot
the prior target before the optimistic assignment and restore it (plus the
video page's canReapply flag) when the start POST rejects; successful loads
and failures after the background load starts behave as before.
Require the model_index.json to sit at the snapshot ROOT before flagging a
cached repo as pipeline-loadable: CachedFileInfo.file_name is the basename,
so the previous name match also claimed nested copies (subdir/model_index.json)
and the picker then sent a from_pretrained load that fails only after the GPU
handoff. Scope by file_path against the revision's snapshot_path.
Validate the maximum derived seed before the multi-run image loop: an explicit
seed near 2**53-1 plus the per-run offset (base + i*batchSize) exceeded the
backend cap and 422'd a later run after earlier images had already generated.
Route local single-file .safetensors picks on the Images and Video pages
through the single_file load path (parent dir + basename), matching the local
GGUF branch: the pipeline route rejects a bare file with no model_index.json,
and only after evicting the resident model.
Applies the video round-2 accuracy findings to the image diffusion stack and fixes
two real image-path bugs found while measuring. All numbers B200, production
settings (family default steps/guidance, 1024px, seed 42, 4 fixed prompts), LPIPS
(AlexNet) via the new scripts/image_speedmem_bench.py, which drives the production
lever functions in the loader's own order.
- inductor precision parity: emulate_precision_casts=True on the regional-compile
path (fused pointwise kernels keep fp32 intermediates where eager rounds to bf16
between ops). Pairwise LPIPS of the compiled tier vs the same-stack eager tier:
Qwen-Image 0.019 to 0.006 at identical speed (72.4 vs 72.5 ms/step), FLUX.1-dev
0.046 to 0.029 at +2% step time (69.8 vs 68.3, reproduced), FLUX.2-klein-4B
0.018 to 0.017 at identical speed. Snapshot/restored with the other process-wide
backend flags so an off load never inherits it.
- cache x compile composition: re-point each cache hook's fn_ref.original_forward
at a torch.compile'd wrapper of the same bound method (armed only where the
speed layer compiled the block; restored before every disable_cache and before
the partial-hook cleanup). Qwen-Image FBCache computed steps 91.8 to 71.2 ms
(back at the uncached compiled rate), 1.21x end to end (7.36 to 6.06 s per 4
images); FLUX.1-dev already traced through its FBCache hook and is measured
neutral (same-process armed vs unarmed latents bit-identical). Skip counts
within noise (13 vs 11 of 76; pairwise LPIPS 0.005).
- FBCache mid-session toggle crash: diffusers 0.39 caches the HookRegistry child
list on first cache_context use, so an uncached generation followed by a
20+-step generation (the auto toggle path) enabled hooks the context never
reached and crashed with "No context is set" (reproduced live on FLUX.1-dev).
Invalidate the stale child cache after every enable_cache.
- TE fp8_dynamic zero-row guard: torchao per-row fp8 derives a per-output-channel
scale from the row amax, so an all-zero weight row is 0/0 = NaN. SDXL's
text_encoder_2 (OpenCLIP bigG) ships exactly such a row, and every explicit
fp8_dynamic SDXL render came out black; keep zero-row Linears dense (LPIPS
0.976 black to 0.096 working). Other families' encoders have no such rows and
are byte-identical.
- No AUTO TE quant exists on the image branch (text_encoder_quant defaults dense,
explicit-only), so the video round's auto-dense retune has no image analogue;
the explicit lever's cost is now measured (TE fp8_dynamic alone, LPIPS vs
bit-exact: Qwen-Image 0.038, FLUX.1-dev 0.084, SDXL 0.096; no speed win, VRAM
-6.5 GB on Qwen-Image) for the docs.
Tests: 96 passing across the cache/speed/precision suites (11 new arming, 2
child-registry, 2 zero-row, 4 inductor-flag); ruff clean.
Step 0 now reads "Preparing (text encoding + warmup)..." on the video and
images pages: text encoding and warmup run before the first scheduler tick,
so the bar otherwise sits on "step 0/N" for up to a minute at 720p.
Generation progress polls are also wired to a visibilitychange listener for
their lifetime: background tabs clamp setInterval to one second and can
suspend it entirely after a few minutes, so returning to the tab now fires
one immediate poll (overlap-guarded) instead of showing a stale label until
the next throttled tick. The listener is removed with the interval on the
terminal phase, generation end, and unmount.
* CI: retry transient HTTP timeouts in Studio smoke probes
The post() helper in the Studio inference smoke workflows does a single
urlopen with a 240s timeout against the local Studio server. On shared
runners this sporadically hits TimeoutError while the server is stalled,
failing the whole job for a transport hiccup; the same flake has recurred
across unrelated PRs on Linux and Windows (JSON/images and tool-calling
jobs) and passes on rerun.
Retry the probe up to 3 times on transport-level failures only
(TimeoutError, ConnectionError, non-HTTP URLError), 15s apart. HTTP
status errors still surface immediately, so genuine server failures are
unaffected. post_sse() is left unchanged: it has a 600s budget and has
not flaked.
* CI: retry only short probes so worst case fits the job budget
Some json-images calls pass timeout=600; three attempts there could spend
30 minutes in one step and hit the job's timeout-minutes instead of failing
with the Python error. Retry (3 attempts) only when timeout <= 300s, which
covers the observed flaky 180-240s probes; longer probes keep the pre-PR
single attempt.
* CI: give long smoke probes one capped retry
Round two of bounding the retries: timeout>300s probes previously got a
single attempt, so a transient stall in the 600s JSON-mode probes still
failed on first occurrence. Give them one retry with the attempt timeout
capped at 300s. Worst cases stay inside timeout-minutes: 240s probes
12.5 min, one 600s probe 15.25 min, the Windows JSON job's two long
probes 30.5 min against its 35 minute budget.
POST /video/generate previously held the response open for the whole
generation (multi-minute for 720p), so in --secure mode the Cloudflare
quick tunnel's ~100s origin-response cap returned a 524 while the server
kept generating, and the frontend treated the run as failed.
Generation now follows the same return-at-once pattern as /video/load:
begin_generate validates synchronously (409 on no model or on a second
concurrent generate via a new busy sentinel) and runs the existing
generate + gallery-persist pipeline, with the route's exact error
mapping, on a daemon thread. GET /video/generate-progress gains optional
terminal fields: phase completed carries the saved gallery record, phase
failed a client-safe error; active only drops together with a terminal
phase. The cancel event is registered before the worker starts so
/video/generate/cancel keeps working across the whole job.
VideoGenerateResponse becomes an accepted acknowledgement (status
started, video kept as an always-null compat field). The video page
fires the POST, then drives completion off the progress poll it already
runs (completed prepends the clip, failed surfaces the error, the
cancelled sentinel stays toast-free). The API-key training-start guards
now also probe the video backend for an in-flight background clip, since
it is no longer visible as an in-flight HTTP request to the keep-warm
counter.
Route tests keep the fake backend for load/generate/status but inherit
the real job machinery, covering immediate accept, concurrent 409, the
terminal completed record, sanitized/ValueError/cancelled failures, and
cancel of a running job.
Tag cached diffusion repos that ship no model_index.json with single_file in the
cached-models listing, and keep them out of the task-scoped On Device pickers
unless the curated catalog carries their artifact: the selection fall-through
loads uncataloged rows as a full pipeline and from_pretrained fails on a
single-file checkpoint repo after the GPU handoff.
Refresh diffusion status after a successful generation run on the Images page.
Speed Auto compiles the transformer on the third LoRA-free generation and flips
supports_lora to false; without the refresh the LoRA picker stayed enabled and
the next LoRA generation failed on the backend.
Match upload passthrough exact paths with trailing slashes normalized: the
trailing-slash variant of /api/train/diffusion/dataset reaches MaxBodyMiddleware
before the router's redirect_slashes 307, so it fell through to the default
/api/train body cap and 413ed large uploads. JSON sub-routes keep extra path
components after normalization and stay on the small cap.
Run backend.start_training off the event loop with asyncio.to_thread so the
synchronous diffusion/video unload calls (which wait on engine generation
locks) cannot freeze concurrent requests; guard against overlapping starts
with a _start_in_progress compare-and-set under the service lock.
Resolve bare diffusion dataset names directly under datasets_root() before
falling back to the generic resolver, so an unrelated LLM upload file or
recipe folder sharing the name cannot shadow the image dataset.
Reject exact duplicate filenames within one multipart upload batch: two
parts staged to the same destination would let the later tmp.replace
silently discard the earlier file. Case variants stay exempt per the
existing stem-guard contract.
Require an instance prompt in the train panel when only some images have
captions, since backend discovery silently skips uncaptioned images.
The catalog fit budget used gpu.memoryTotalGb, which sums VRAM across
every GPU. That sum is right for the chat/llama.cpp path (tensor-split
shards across cards) but wrong for the diffusion/video catalog: those
backends place the whole pipeline on a single device (pipe.to or cpu
offload, never device_map), so on a multi-GPU host the fit toggle and
bare-group-click routing credited VRAM no single card has. On a 4x24 GB
plus 128 GB RAM host the 114 GB Wan A14B bf16 group passed the toggle
(0.7*96 + 0.7*128 budget) and a click would OOM, the exact load the
toggle exists to prevent. Expose maxDeviceMemoryGb (largest single
device) from use-gpu-info and use it for deviceBudget; the chat path
keeps the sum. Single-GPU hosts are unchanged.
* Clarify in README that -H 0.0.0.0 starts a public Cloudflare tunnel
* Hedge tunnel URL wording and restore trusted-network caution
* Tighten the 0.0.0.0 tunnel note
* Drop trust-the-network caution from tunnel note
* Restore trusted-network note on the raw-bind sentence
* Use Cloudflare's quick tunnel terminology and consolidate the trust warning
* Studio: resolve the repo-root MTP drafter after the MTP/ GGUF rename
The Gemma 4 QAT GGUF repos renamed the higher-precision MTP/ subdir
copies from gemma-4-...-<quant>-MTP.gguf to mtp-gemma-4-...-<quant>.gguf,
so their basenames now start with the same mtp- prefix as the small
repo-root drafter (mtp-gemma-4-E4B-it.gguf).
The drafter selectors filtered candidates by a mtp- basename prefix and
took the first in sort order. With the new names the MTP/ copies also
match, and because MTP/ (uppercase) sorts before the lowercase root file,
selection flipped to the large BF16 copy under MTP/ instead of the root
drafter both functions document they should pick.
Restrict both selectors, and the companion byte estimate, to root-level
mtp-*.gguf so the MTP/ copies stay explicit-selection only:
- core/inference/llama_cpp.py _pick_mtp (loader auto-download)
- hub/utils/gguf_plan.py preferred_mtp_sibling (Hub variant plans)
- routes/inference.py _remote_gguf_companion_bytes (VRAM headroom)
Also reuse a drafter already in the local cache before downloading, so a
device that already holds a copy on disk does not re-fetch it.
Old-scheme names keep working (they have no root-level mtp- sibling to
mis-select). Adds regression tests for the new naming, both selection
paths, and the on-disk reuse.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Studio: gate MTP drafter cache reuse to offline mode
Reuse the cached drafter only when HF is offline. Online, route back
through _download_companion_gguf/hf_hub_download so the current revision
is checked (etag) and a changed drafter is refetched, matching the
offline-only cross-snapshot reuse already used for the main GGUF. This
avoids pairing freshly downloaded weights with a stale cached draft.
Make the reuse tests offline and add an online-skips-reuse test.
* Studio: prefer a root MTP drafter across all cached snapshots
Offline reuse scanned snapshots one at a time and returned the first
snapshot that held any drafter, only preferring root within it. A newer
partial snapshot with just the MTP/ copy could shadow the small root
drafter in an older snapshot. Collect drafters across all snapshots and
prefer any repo-root file before an MTP/ copy.
* Studio: keep newest-first snapshot order when reusing cached drafters
Collecting root candidates and sorting by absolute snapshot path could
pick a drafter from an older snapshot. _iter_hf_cache_snapshots yields
newest first and the main GGUF is resolved in that order, so preserve it
(root still preferred over MTP/ copies) to avoid pairing a fresh main
weight with a stale drafter revision.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Restore dropped FP8 weight_scale_inv tensors on load
Some block-scale FP8 checkpoints (for example Qwen3.6-27B-FP8, issue #6200) load
with transformers leaving an mlp.gate_proj as a plain bf16 Linear instead of an
fp8 module. Its raw quantized values are read into the bf16 weight and the
weight_scale_inv is dropped as an unexpected key, so the weight is used un-scaled
and the base model is garbage (perplexity around 2 million).
After load, for every checkpoint weight_scale_inv whose live weight is not fp8,
dequantize the orphaned weight in place using the block scale from the checkpoint
index. Modules that were converted correctly keep an fp8 weight and are skipped,
so healthy checkpoints and single-file checkpoints are a no-op.
Verified on Qwen3.6-27B-FP8: 64 gate_proj scales restored, perplexity 2028902 to
8.9. No-op on Qwen3-8B-FP8 (all scales already live).
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Harden FP8 weight_scale_inv restore from review
- Skip restore when the model has no fp8 weights, so an intentionally
dequantized load (load_in_16bit) is never re-scaled and corrupted.
- Thread revision, subfolder and cache_dir through the index and shard
downloads so scales come from the same snapshot as the weights.
- Cover unsharded single-file model.safetensors checkpoints (no index).
- Handle transposed block-scale layouts and skip on a true grid mismatch
instead of applying a wrong scale.
- Match text-only VLM loads where the language_model prefix was stripped.
- Restore on the FastLanguageModel text path too, not only vision.
- Handle a scalar weight_block_size; per-tensor error handling so one bad
tensor cannot abort the rest or hide a partial mutation.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Address second review round on FP8 scale restore
- Bound peak memory: dequantize block views in place with the fp32 scale
broadcast instead of materializing a full expanded scale and fp32 copy,
so a near-VRAM-limit load is not pushed into OOM by the repair.
- Restore on the sequence-classification load path too.
- Cover more VLM key remappings (language_model.model.* to
model.language_model.*) when matching modules.
- Skip the restore for variant loads (variant=...) rather than risk
applying default-checkpoint scales to variant weights.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Align FP8 scale restore revision with the loaded weights and warn on disk-offloaded layers
In llama.py the CausalLM/SequenceClassification weight loads resolve model_name on its
default branch (revision is not forwarded there), so read the dropped weight_scale_inv
tensors from the same default branch instead of the requested revision, avoiding rescaling
default-branch weights with scales from another revision.
In loader_utils.py a disk-offloaded layer keeps its weight on the meta device until the
offload hook materializes it, so the scale cannot be applied in place. Skip such layers
explicitly and print a warning rather than silently leaving them unscaled.
* Tighten comments in the FP8 scale restore path
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
parse_direct_linux_release_bundle and direct_linux_release_plan are no
longer reached by any live code path. Fork Linux installs resolve through
_fork_manifest_release_plans -> _linux_published_attempts, and the upstream
(ggml-org) path uses direct_upstream_release_plan. The dead parser also
called _resolve_linux_bundle_profile, which no longer exists, so its CUDA
branch would raise NameError if ever executed.
Drop both functions and the obsolete TestDirectLinuxNvidiaCpuGate; its live
equivalent TestLinuxPublishedAttemptsNvidiaCpuGate already covers the
NVIDIA no-silent-CPU behaviour.
* scripts: refresh scan_packages allowlist baseline
Regenerate scripts/scan_packages_baseline.json against the current
resolved dependency set so the blocking pip scan-packages gate matches
what the scanner now finds. Refreshes evidence hashes for benign
findings whose code shifted lines (unsloth-zoo mlx loader, gguf/mlx
test /tmp fixtures) and adds two mainstream-library entries that were
newly surfaced (torch inductor codecache base64+subprocess compile
cache, torch testing common_utils socket import). Stale entries whose
matching code changed and no longer triggers are dropped.
All entries remain CRITICAL/HIGH findings manually judged benign;
matched on (package, file, check, evidence_hash).
* ci(security-audit): re-run scan when the allowlist baseline changes
The security-audit pull_request trigger listed the scanners but not
their allowlist baselines, so a baseline-only edit never re-ran the
scan that consumes it. A refreshed baseline could therefore merge
without CI confirming its evidence hashes match what the scanner finds.
Add scan_packages_baseline.json and scan_npm_packages_baseline.json to
the paths filter so baseline changes are validated on their own PR.
* Keep native RoPE scaling when extending context; carry rope_theta for linear
When max_seq_length exceeds a model's native window, the loader overwrote the
model's rope_scaling with linear scaling. For models that already ship a scaled
RoPE (llama3/yarn/longrope) that is far worse for long context, and on
transformers v5 the linear dict omitted rope_theta (v5 keeps it under
rope_parameters), so the rotary base fell back to 10000 and broke past ~8K tokens.
Keep the native scaling and just widen the window; only synthesize linear for
plain-RoPE models, and carry rope_theta so v5 keeps the real base.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Only preserve native llama3 when extending context; keep linear fallback otherwise
The patched attention constructor (patch_llama_rope_scaling) rebuilds only linear,
llama3 and longrope and its longrope branch reads a top-level
original_max_position_embeddings, so preserving yarn or a nested-only longrope config
would raise during construction on transformers <= 4.47.1. Keep only llama3 native;
yarn/longrope/other types fall back to the linear override, still carrying rope_theta.
* Correct long-context extension comment to match llama3-only preservation
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Fix fast_gemv crash on compressed-tensors FP8 models
Loading a compressed-tensors FP8 checkpoint (for example
unsloth/Llama-3.2-1B-Instruct-FP8-Block) with fast_inference=False and
running a forward crashed with 'Parameter object has no attribute absmax'
inside fast_gemv.
A compressed-tensors CompressedLinear exposes an already dequantized bf16
weight at forward time while keeping a weight_scale Parameter. The quant
state resolution in get_lora_parameters/get_lora_parameters_bias fell back
to that weight_scale, so a bf16 weight was routed into the bitsandbytes
fast_gemv/fast_dequantize path, which expects a bitsandbytes QuantState
with an absmax attribute.
Only fall back to weight_scale_inv/weight_scale when the weight is still
fp8. A decompressed bf16 weight then resolves to no quant state and flows
through the normal bf16 path, which already handles bias and the LoRA
backward. Real fp8 and bitsandbytes 4bit weights are unchanged.
* Skip the fast_gemv dispatch test before importing unsloth when bitsandbytes is absent
* Fix Windows installer torch index override
* Clear inherited uv index env vars for pinned installs in studio/setup.ps1 (#6898)
* Harden setup.ps1 index-var clearing to truly remove vars (#6898)
* Apply UV_DEFAULT_INDEX torch index fix to Linux/Mac install.sh (#6898)
* Neutralize all uv index env vars for pinned torch installs (#6898)
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: Daniel Han <danielhanchen@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Studio: add Vulkan llama.cpp support
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Address gemini's feedback
* Studio: move the Vulkan VRAM probe into a standalone script
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Improve Vulkan probe error reporting
* Resolve llama-server symlink so Vulkan build is detected
* Drop unreachable Vulkan fallback in GPU free-memory dispatcher
* Skip the Intel GPU probe when NVIDIA or ROCm is present
* Reserve host RAM headroom for Vulkan integrated GPUs
* Add a `UNSLOTH_FORCE_VULKAN` environment variable
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Honor GGML_VK_VISIBLE_DEVICES, reserve discrete Vulkan VRAM headroom, and clear Intel GPU on --cpu-fallback
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Route Intel and forced-Vulkan hosts to the upstream Vulkan prebuilt, add arm64 Vulkan, keep Vulkan out of RAG auto-detect
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Clear the fork release pin when routing a Vulkan host to the upstream repo
* Gate auto-Vulkan routing on no physical NVIDIA so hidden CUDA devices aren't used
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Pin Vulkan launches with --device Vulkan<i> instead of the raw GGML_VK_VISIBLE_DEVICES index space
* Let user --device override the Vulkan pin, and gate direct Vulkan asset picks on no physical NVIDIA
* Update RAG auto-backend test mocks for the _resolve_auto binary and Vulkan probes
* Keep the add_dll_directory handle alive through the Vulkan probe DLL loads
* Revert RAG auto Vulkan guard, guard multi-backend Vulkan detection, and preserve forced Vulkan across updates
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Use getattr for RTLD_GLOBAL in the Vulkan probe CDLL mode
* Skip CUDA/ROCm APU and datacenter GPU tuning on Vulkan builds
On a Vulkan llama.cpp build gpu_indices are ggml compact ordinals, not
CUDA/ROCm physical ids, so _amd_apu_wants_unified_memory and
_apply_datacenter_env were reading the wrong device. On a mixed AMD APU
plus discrete GPU host that could raise a spurious system-RAM shortfall
and block a valid discrete-GPU load. Gate all three call sites on
not is_vulkan_backend; the Vulkan path already reserves iGPU host
headroom and the backend ignores GGML_CUDA_* anyway.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Tighten Vulkan-guard comment in load_model
* Reduce comments in Vulkan support to be more succinct
* Resolve shell-wrapper llama-server entrypoint to the real lib dir
create_exec_entrypoint falls back to a #!/bin/sh wrapper at the install
root when it cannot symlink into build/bin. _find_llama_server_binary
returns that root entrypoint, but Path.resolve() does not follow a shell
wrapper, so _llama_lib_dir returned the install root and _is_vulkan_backend
missed libggml-vulkan.so -- silently skipping the Vulkan probe and --device
pin on an otherwise valid Vulkan install. Follow the wrapper's exec target
to build/bin. Regression test: test_shell_wrapper_entrypoint_resolves_to_real_lib_dir.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Lee Jackson <130007945+Imagineer99@users.noreply.github.com>
Co-authored-by: danielhanchen <danielhanchen@gmail.com>
* Studio /v1/messages: accept thinking and unknown content blocks
The Anthropic-compatible /v1/messages endpoint modeled a message's content as
Union[str, list[{text|image|tool_use|tool_result}]], so any other block type
made Pydantic reject the whole request with
`messages.N.content.str: Input should be a valid string`. Resuming a Claude
session commonly replays assistant turns that carry `thinking` (extended
thinking) blocks, and sometimes a null content for a tool-only turn, both of
which tripped this and returned a 400.
Accept them:
- Add a permissive AnthropicUnknownBlock fallback (any block whose type is not
one of the four known ones), so thinking/redacted_thinking/provider-specific/
future blocks validate. A validator keeps known types on their typed models,
so a malformed known block (e.g. a tool_use without id) still fails cleanly.
- Coerce a null message (and tool_result) content to "" so the converter's
`for block in content` stays safe.
The converter already drops block types it does not translate, so a thinking
block is not forwarded to the model.
* Studio /v1/messages: keep user content validation strict
Make the thinking/null leniency role-aware so it never silently drops real
user input. Assistant turns (replayed history) still accept unknown/thinking
blocks and coerce a null tool-only turn to empty. User turns keep the strict
boundary: a null user content is rejected, and a content block the converter
cannot translate is rejected instead of being dropped into an empty prompt.
Also remove an empty file committed by accident.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Studio /v1/messages: coalesce resumed user turns and tighten content checks
- The /v1/messages count and generation paths now coalesce the adjacent user
turns that dropping an empty or null assistant turn can leave behind, so a
strict GGUF chat template no longer 400s on non-alternating roles.
- A user content block with a non-string type (list / dict) is rejected as a
clean 400 instead of raising TypeError and escaping as a 500.
- The assistant null-to-empty coercion only applies to an explicit null; an
assistant turn that omits content entirely still fails required-field
validation instead of being silently coerced to an empty string.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Studio /v1/messages: tighten comments
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* unsloth start: add --resume to persist and reopen agent sessions
`unsloth start <agent>` launches a coding agent whose home is a throwaway
temp dir wiped on exit, so codex/openclaw/hermes/pi (which relocate their
whole home there) cannot resume a conversation after you quit. opencode and
claude keep their session data in a fixed user dir, so they already resume.
Add an opt-in --resume/--no-resume flag: it routes the launch to the stable
Unsloth agents dir (the same one --no-launch already uses) so the session
survives the exit, never touching the user's own ~/.<agent>. A bare --resume
also reopens the last conversation via the agent's native flag (codex
`resume --last`, opencode/claude/pi `--continue`). The default is unchanged:
a plain launch still uses a temp dir and persists nothing.
Add a dispatch-only `resume` job to the Local Agent Guides CI that drives the
real launch path and asserts the split: codex/pi are wiped without --resume
and persist with it, while opencode/claude persist either way.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* unsloth start: rename --resume to --persist
The session flag collided with agents' own resume flags. `unsloth start
claude --resume <id>` used to forward `--resume <id>` straight to Claude
(which keeps its history in ~/.claude regardless), so a boolean --resume on
unsloth start would have swallowed the session id and turned it into a stray
prompt. Name the persistence flag --persist instead, so every agent's native
resume flag (claude --resume <id>, codex resume, opencode --continue, ...)
still passes through untouched. Behavior is otherwise identical: --persist
keeps a launched agent's session under the Unsloth agents dir, and a bare
--persist reopens the last conversation.
Add a regression test that `--resume <id>` passes through verbatim, and in the
CI resume experiment skip the redundant second pass for opencode/claude (they
persist either way, and a second CPU turn only risks a timeout).
* unsloth start: correct --persist help and drop the buggy auto-resume
Reword the --persist help to be accurate: claude and opencode keep sessions in
the user's own stores and resume regardless, so --persist only stabilizes the
otherwise-ephemeral relocated home of codex/openclaw/hermes/pi. Drop the
bare-launch auto-append of native resume tokens: it errored on a first launch
with no prior session, and was inconsistent between launch and no-launch.
--persist now only keeps the session dir; resume via the agent's own command
(e.g. `unsloth start codex --persist resume`), which now finds it.
In the CI resume experiment, fail the pass when the launched turn exits
non-zero, so a write-then-error is not misread as PERSISTED.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
The main merge left listStoredChatThreads imported twice -- once as a standalone import from
the deep utils path and once via the @/features/chat barrel (which re-exports it) -- tripping
TS2300 'Duplicate identifier' and failing the Tauri frontend build. Keep the barrel import,
grouped with the other chat imports.
* Silence torch._check_is_size FutureWarning and shim it if torch removes it
bitsandbytes 4-bit dequant calls torch._check_is_size, which torch
deprecated with a FutureWarning ("Use _check(i >= 0) instead") that prints
on every bnb-4bit load. Silence that warning in suppress_cuda_printf, and
add fix_torch_check_is_size so a future torch that removes _check_is_size
gets it shimmed to _check(i >= 0) (honoring the max bound) and bitsandbytes
keeps working.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Tighten fix_torch_check_is_size docstring
Lead with what the shim does and drop the redundant line; two lines
instead of three, same intent.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Mirror dynamo/inductor config sets into defaults so torch 2.12 worker threads honor them
torch 2.12 stores config user overrides in ContextVars, so direct
assignments like torch._dynamo.config.recompile_limit = 1024 no longer
reach the autograd engine worker threads. Gradient checkpointing
recomputes fullgraph-compiled gpt-oss kernels inside backward on those
threads, which then read the default recompile limit of 8 and raise
FailOnRecompileLimitHit at step 0 of GRPO/SFT. Mirror direct config
assignments into the process-global entry defaults on torch >= 2.12,
restoring the torch <= 2.11 cross-thread semantics while leaving the
context-scoped config.patch API untouched.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Keep config.patch thread-local when mirroring dynamo/inductor sets
config.patch(...) also assigns through ConfigModule.__setattr__, so the
default-mirror was leaking its scoped, thread-local writes into the
process-global entry default. Track patch enter/exit with a per-thread
depth counter (wrapping ConfigModule.patch) and skip mirroring while
inside a patch, so only genuine direct assignments restore the torch
2.11 cross-thread semantics and config.patch stays context-local.
* Also keep config.load_config thread-local when mirroring config sets
load_config restores a saved dynamo/inductor config by calling setattr
per key, which the default-mirror would otherwise leak process-wide just
like config.patch did. Wrap load_config with the same per-thread depth
counter (renamed to _scoped_depth) so both scoped writers skip the mirror
and stay context-local, while genuine direct assignments still restore the
torch 2.11 cross-thread default.
* Drop the pre-existing override replay from the config thread fix
The replay was redundant: this runs from _gpu_init before unsloth sets any
dynamo/inductor config, so the __setattr__ wrapper already mirrors every
later assignment (recompile_limit included). It could also read a value
that belonged to a config.patch context still active at import time and
write that thread-local override into the global default. Removing it keeps
the cross-thread fix and drops the now-unused _inductor.config import.
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
