unsloth/studio/backend
Daniel Han 2d026a1184
Studio: reset quantized KV cache to f16 when the flash-attn-off crash-recovery fallback fires (#7390)
* Studio: reset quantized KV cache to f16 when flash-attn-off fallback fires

Studio force-enables --flash-attn on for GGUF launches. On a hard startup
or first-decode crash it retries via _with_flash_attn_off, which flipped FA
off but left --cache-type-k/-v untouched. A quantized KV cache (q8_0, q4_0,
q4_1, q5_0, q5_1, iq4_nl) requires flash attention in llama.cpp, so the retry
itself aborted at init with 'V cache quantization requires flash_attn' instead
of recovering.

Reset any quantized --cache-type-k/-v to f16 in the FA-off fallback path so
the retry can actually launch. Non-quantized types (f16, bf16, f32) run fine
without flash attention and are left unchanged. Handles long and short flag
forms and both space and equals syntax, rewriting in place to preserve list
length. Adds pytest coverage.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* Studio: FA-off fallback resets only the quantized V cache and drops env-only V cache

Only the V cache requires flash attention in llama.cpp; a quantized K cache
runs fine without it. Restrict the FA-off crash-recovery reset to the V axis
(main and draft) so a memory-constrained config keeps its quantized K cache
instead of risking an OOM on the recovery. Also drop an inherited quantized V
cache set purely through the environment (LLAMA_ARG_CACHE_TYPE_V /
LLAMA_ARG_SPEC_DRAFT_CACHE_TYPE_V) at the FA-off retry sites, which the argv
rewrite cannot reach, so the child falls back to the f16 default rather than
aborting.

* Studio: normalize underscore V-cache aliases in the FA-off fallback

llama.cpp rewrites '_' to '-' for any '--' long option before matching,
so a pass-through --cache_type_v q8_0 enables a quantized V cache just
like --cache-type-v. The FA-off crash-recovery reset only matched the
hyphenated spelling, so the underscore alias slipped through and the
retry still aborted with "V cache quantization requires flash_attn".
Canonicalize the flag name the same way before matching (short flags and
the type value are untouched).

* [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>
2026-07-25 04:10:44 -07:00
..
assets feat(studio): add DoRA support to studio (#7315) 2026-07-24 03:24:16 -07:00
auth Studio: reject whitespace-only passwords (#7341) 2026-07-23 00:44:37 -07:00
core Studio: reset quantized KV cache to f16 when the flash-attn-off crash-recovery fallback fires (#7390) 2026-07-25 04:10:44 -07:00
hub Studio: add local speech-to-text dictation engine (#7095) 2026-07-23 01:39:03 -07:00
loggers Studio: quiet noisy logs, log real progress, and speed up Windows/macOS dataset prep (#7087) 2026-07-15 06:49:52 -07:00
models feat(studio): add DoRA support to studio (#7315) 2026-07-24 03:24:16 -07:00
picker Studio: add configurable model download location (#7274) 2026-07-23 01:34:38 -07:00
plugins Replace standalone Studio wording with Unsloth (#7221) 2026-07-19 00:47:04 -07:00
requirements unsloth start/run: tool-call flags, positional model, and grouped help (#7328) 2026-07-23 01:44:57 -07:00
routes Unsloth Studio (desktop): fix canvas preview, download file button, toast placement, and model-load typing lag (#7391) 2026-07-24 22:23:41 -05:00
state Studio: shareable per-checkpoint preview links (#6486) 2026-06-24 06:31:53 -07:00
storage fix(studio): persist connection model selections for remote clients (#7298) 2026-07-23 19:11:50 -07:00
tests Studio: reset quantized KV cache to f16 when the flash-attn-off crash-recovery fallback fires (#7390) 2026-07-25 04:10:44 -07:00
utils Studio: scan HF cache snapshot loads by their repo id (#7398) 2026-07-24 02:12:00 -07:00
__init__.py Final cleanup 2026-03-12 18:28:04 +00:00
_platform_compat.py Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
cloudflare_tunnel.py Studio: fix stuck composer prompt on first send and unreachable --secure Cloudflare links (#7340) 2026-07-23 00:39:14 -07:00
colab.py fix(studio/colab): restore blank Colab iframe embed (#7344) (#7349) 2026-07-24 02:23:24 -07:00
main.py fix(studio): stop false MTP/vision capability reports (#7332) 2026-07-24 02:13:52 -07:00
mcp_server.py Replace standalone Studio wording with Unsloth (#7221) 2026-07-19 00:47:04 -07:00
run.py Studio: reject whitespace-only passwords (#7341) 2026-07-23 00:44:37 -07:00
startup_banner.py Replace standalone Studio wording with Unsloth (#7221) 2026-07-19 00:47:04 -07:00