[FEATURE] Support latest version of transformers

[ParagEkbote]

Is your feature request related to a problem? Please describe.

Currently, transformer v.4.51.0 is only supported due to constraints of llmcompressor and gliner as seen by this pipdeptree output:

transformers==4.51.0
├── llmcompressor==0.6.0.1 [requires: transformers>4.0,<=4.52.4]
│   └── pruna==0.2.10 [requires: llmcompressor]
├── whisper-s2t==1.3.1 [requires: transformers]
│   └── pruna==0.2.10 [requires: whisper-s2t==1.3.1]
├── gliner==0.2.22 [requires: transformers>=4.38.2,<=4.51.0]
│   └── pruna==0.2.10 [requires: gliner]
├── DeepCache==0.1.1 [requires: transformers]
│   └── pruna==0.2.10 [requires: DeepCache]
├── optimum==1.27.0 [requires: transformers>=4.29]
│   └── whisper-s2t==1.3.1 [requires: optimum]
│       └── pruna==0.2.10 [requires: whisper-s2t==1.3.1]
├── pruna==0.2.10 [requires: transformers]
├── compressed-tensors==0.10.2 [requires: transformers]
│   └── llmcompressor==0.6.0.1 [requires: compressed-tensors==0.10.2]
│       └── pruna==0.2.10 [requires: llmcompressor]
└── hqq==0.2.7.post1 [requires: transformers>=4.36.1]
    └── pruna==0.2.10 [requires: hqq==0.2.7.post1]

This leads to key errors for some models like SmolLM3-3B and GLM-4.5V. While we can resolve this using pip install --upgrade transformers, it could be better to add support to prevent these silent bugs. Also, since transformers v5 is right around the corner, this error could become more prominent. WDYT?

cc: @davidberenstein1957 , @johannaSommer

Describe the solution you'd like

Please support latest version of transformers for better model architecture coverage.

[davidberenstein1957]

@ParagEkbote I agree we should do this. Although we commonly try to control this, as new releases commonly break older algorithms, so we prefer to do this manually instead of fully relaxing the contraints.

[ParagEkbote]

I'm trying to resolve this in #417, the version has been updated to v.4.55.0 in the PR without breaking changes. WDYT?

cc: @davidberenstein1957