unsloth/unsloth
Daniel Han 6df6f82557 torchcodec: warn when torch outruns the last lockstep row
The ABI-stable branch clears torch >= 2.11 paired with torchcodec >= 0.12,
but everything else fell through to the lockstep matrix, which has no row
past torch 2.11. So torch 2.12 or 2.13 with torchcodec 0.10 or 0.11 hit
`allowed is None` and returned no hint at all, even though upstream pins
0.11 to torch 2.11 exactly and 0.10 to torch 2.10 exactly. torchcodec
ships no `Requires-Dist: torch`, so pip cannot catch it either and the
user only sees a raw "Could not load libtorchcodec" at first use.

Split the lookup into three cases: a torch minor below the table stays
silent as before, a torch minor with a row keeps the existing per-minor
pin and audio-torch2xx extra hint, and a torch minor at or past the
ABI-stable floor with no row now points at `torchcodec>=0.12.0`.

Mirror the same three-way split in scripts/notebook_validator.py's
R-INST-004 so the two stay in step, and cover both with tests plus a
regression test that torch below the table still produces no finding.
2026-07-27 13:28:15 +00:00
..
dataprep fix(dataprep): don't emit a degenerate chunk for empty text (#7183) 2026-07-23 00:56:51 -07:00
kernels Propagate fp8 block_size before the early return in get_lora_parameters_bias (#7189) 2026-07-17 16:30:45 -07:00
models avoid Hub metadata probe when loading tokenizers with local_files_only (#7482) 2026-07-27 05:59:48 -07:00
optimizers Reduce and tighten code comments and docstrings repo-wide (#6095) 2026-06-08 23:09:51 -07:00
registry fix(registry): don't register deepseek models at import time (#7227) 2026-07-19 03:37:23 -07:00
utils Experimental: correct varlen sample packing for hybrid linear-attention models (#7249) 2026-07-20 00:57:02 -07:00
__init__.py fix(mlx): relax context-store timeout by default (#7141) 2026-07-17 17:13:19 -03:00
_auto_install.py Upgrade setuptools and wheel in the auto-install command (#6282) 2026-06-13 03:59:15 -07:00
_compressed_quantize.py Fix export-time trust_remote_code bypass in FP8/INT8/GGUF-LoRA export (#6869) 2026-07-05 05:16:39 -07:00
_gpu_init.py Silence torch._check_is_size FutureWarning and shim it if torch removes it (#7023) 2026-07-09 02:26:36 -07:00
chat_templates.py fix(chat_templates): bind loop_messages when default_system_message is None (#7199) 2026-07-19 06:33:48 -07:00
device_type.py AMD: enable ROCm torch on gfx906 (MI50 / Radeon VII) on Linux (#7354) 2026-07-27 05:22:19 -07:00
import_fixes.py torchcodec: warn when torch outruns the last lockstep row 2026-07-27 13:28:15 +00:00
ollama_template_mappers.py Re-apply #4939: updated models template mappers (#4950) 2026-04-15 07:52:12 -07:00
save.py avoid Hub metadata probe when loading tokenizers with local_files_only (#7482) 2026-07-27 05:59:48 -07:00
tokenizer_utils.py Replace standalone Studio wording with Unsloth (#7221) 2026-07-19 00:47:04 -07:00
trainer.py Experimental: correct varlen sample packing for hybrid linear-attention models (#7249) 2026-07-20 00:57:02 -07:00