chore(deps): update dependency compressed-tensors to v0.18.0

[renovate]

This PR contains the following updates:

Package	Change	Age	Confidence
compressed-tensors	==0.17.0 → ==0.18.0

---

Release Notes

vllm-project/compressed-tensors (compressed-tensors)

v0.18.0: Compressed-Tensors v0.18.0

Compare Source

What's Changed

• Add Buildkite pipeline for test-check workflow on H100 duo and L4 solo by @​deepak-kumar-neu in #​722
• Rename upload step label in pipeline.yml by @​dsikka in #​725
• [Converters] Support loading safetensors on meta device by @​kylesayrs in #​726
• [Offload] Use explicit class patching for load_offloaded_model by @​kylesayrs in #​728
• Ignore non-linear MoE gate layers by @​kylesayrs in #​733
• [Offload] Skip device-map entries with no local module in dispatch_with_map by @​waynehacking8 in #​737
• fix model chkpt name mismatch for ignore by @​HDCharles in #​739
• Add runner selection dropdown for manual Buildkite triggers by @​deepak-kumar-neu in #​744
• CT support for humming by @​HDCharles in #​732
• [Offload] Force offloading of buffers by @​kylesayrs in #​731
• [Offload] Better error for disk offloaded transformers models by @​kylesayrs in #​729
• Add tqdm to apply_quantization_config by @​kylesayrs in #​730
• Revert default GPU runner to H100duo by @​deepak-kumar-neu in #​745
• Add XPU support to Buildkite pipeline and scripts by @​chensuyue in #​736
• [Bugfix] Fix cast_to_fp4 per-rank torch.compile recompilation (#​734) by @​Yatimai in #​741
• [Offloading] Speedup to_accelerate by only instantiating OffloadedWeightsLoader once by @​kylesayrs in #​752
• [Bugfix] Enforce block_structure length validation for string format by @​amogh-nagri-11 in #​755
• [Bugfix] [Distributed] Fix disk distributed module compression by @​kylesayrs in #​750
• Fix tied transform weight detection for offloaded models by @​ZewenShen-Cohere in #​751
• Add psutil as required dependency by @​timkpaine in #​740
• [Bugfix] Fix ignore list for linearized MoE layers by @​kylesayrs in #​759
• [Bugfix] Validate positive block structure dimensions by @​harivilasp in #​763
• Fix MTP tensor copy from indexed checkpoints by @​SorenDreano in #​756
• [Offload] [1/2] Disambiguate synchronous __setitem__/offload from update_offload by @​kylesayrs in #​709
• Revert "[Offload] [1/2] Disambiguate synchronous `setitem/offload… by @​kylesayrs in #​767
• CT humming schemes by @​HDCharles in #​760
• fix pack_fp4_to_uint8 faster/less mem by @​HDCharles in #​775
• [Bugfix] Remove unreachable GROUP+dynamic activation validation branch by @​ishrith-gowda in #​772
• Treat imatrix weight observer as calibration data dependent by @​dshane1903 in #​710
• Fix NVFP4 packing for float32 inputs by @​yiliu30 in #​781
• [XPU] Add torch.cuda linter by @​kylesayrs in #​771
• add wNaM helper and presets by @​HDCharles in #​785
• [Quantization] Only apply quantization schemes to Linear, Embedding, and Attention layers by @​kylesayrs in #​753
• [ModelCompressor] QuantizationConfig.merge, allow compress on models with pre-existing quant by @​brian-dellabetta in #​779
• [Performance] Triton implementation for cast_to_fp4 by @​kylesayrs in #​774
• [offload] wrap offload calls in try/finally by @​brian-dellabetta in #​790
• save_mtp_tensors_to_checkpoint graceful failure by @​kylesayrs in #​788
• feat: cap shared memory budget by mmap descriptor limit by @​soyr-redhat in #​798
• [bugfix] Enable on MacOS by @​brian-dellabetta in #​800
• Fix KV cache quantization for composite multimodal models by @​yasu-oh in #​780
• [QuantizationConfig] drop propagation of quant config format to scheme format by @​brian-dellabetta in #​799
• Add SECURITY.md with vulnerability reporting policy by @​arijitroy003 in #​794
• fix: runtime bugs in quantization and utils by @​soyr-redhat in #​793
• Deprecate actorder group by @​kylesayrs in #​769
• [Bugfix] Do not throw "unexpected non-targeted tensor" validation error for ignored tensors by @​kylesayrs in #​692
• [Misc] Add position for tqdms by @​kylesayrs in #​783
• Triton implementation for quantize kernel in forward helpers by @​ElizaWszola in #​777
• Configure CodeRabbit to reduce verbosity by @​kylesayrs in #​770
• Fix Triton kernel support for non-contiguous x by @​ElizaWszola in #​807
• ImplBackend and triton import utils by @​kylesayrs in #​804
• Update documentation link for compression formats by @​dsikka in #​713
• [XPU] Fp4 pack triton acceleration by @​kylesayrs in #​814
• fix: MXFP4 exclude weight_global_scale from compression_param_names by @​kylesayrs in #​809
• Disable triton kernels by @​kylesayrs in #​817
• [nvfp4] Fix lm-eval failure by @​brian-dellabetta in #​820
• fix: add missing per-tensor calibration in FP8 compress/decompress test by @​kylesayrs in #​821

New Contributors

• @​deepak-kumar-neu made their first contribution in #​722
• @​waynehacking8 made their first contribution in #​737
• @​chensuyue made their first contribution in #​736
• @​amogh-nagri-11 made their first contribution in #​755
• @​ZewenShen-Cohere made their first contribution in #​751
• @​timkpaine made their first contribution in #​740
• @​harivilasp made their first contribution in #​763
• @​SorenDreano made their first contribution in #​756
• @​ishrith-gowda made their first contribution in #​772
• @​dshane1903 made their first contribution in #​710
• @​soyr-redhat made their first contribution in #​798
• @​yasu-oh made their first contribution in #​780
• @​arijitroy003 made their first contribution in #​794
• @​ElizaWszola made their first contribution in #​777

Full Changelog: vllm-project/compressed-tensors@0.17.0...0.18.0

v0.17.1: Compressed-Tensors v0.17.1

Compare Source

What's Changed

• Add Buildkite pipeline for test-check workflow on H100 duo and L4 solo by @​deepak-kumar-neu in #​722
• Rename upload step label in pipeline.yml by @​dsikka in #​725
• [Converters] Support loading safetensors on meta device by @​kylesayrs in #​726
• [Offload] Use explicit class patching for load_offloaded_model by @​kylesayrs in #​728
• Ignore non-linear MoE gate layers by @​kylesayrs in #​733

New Contributors

• @​deepak-kumar-neu made their first contribution in #​722

Full Changelog: vllm-project/compressed-tensors@0.17.0...0.17.1

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.