Update TGI image versions

[xiaotia3]

Description

• Update TGI CPU/rocm version to v2.4.1
• Update TGI Gaudi version to v2.3.1

Versions in AvatarChatbot and ChatQnA-ROCm are not updated due to known issues.

Issues

Related to #1563

Type of change

List the type of change like below. Please delete options that are not relevant.

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds new functionality)
• Breaking change (fix or feature that would break existing design and interface)
• Others (enhancement, documentation, validation, etc.)

Dependencies

N/A

Tests

git grep text-generation-inference: | grep -v 2.4.1
.github/workflows/scripts/update_images_tag.sh:dict["ghcr.io/huggingface/text-generation-inference"]="docker://ghcr.io/huggingface/text-generation-inference:latest-intel-cpu"

git grep /tgi-gaudi: | grep -v 2.3.1

[github-actions]

Dependency Review

✅ No vulnerabilities or license issues found.

Scanned Files

None

[xiguiw]

@xiaotia3

There are conflicts.
would you please resolve it?

[xiguiw]

@xiaotia3

There are new conflicts.
Please help to resolve it.

[Copilot]

Pull Request Overview

This PR updates the TGI image versions across multiple Compose files for various platforms. In particular, it upgrades the Intel CPU images to version 2.4.1, the AMD/GPU (ROCm) images to version 2.4.1, and the Gaudi image to version 2.3.1.

• Upgrade Intel CPU images for ChatQnA, AvatarChatbot, and AudioQnA from 2.4.0 to 2.4.1
• Upgrade AMD/GPU (ROCm) images in ChatQnA, AvatarChatbot, AudioQnA, and AgentQnA from versions 2.3.1 (and in one case 3.0.0) to 2.4.1
• Update documentation in corresponding README.md files to reflect the new image tags

Reviewed Changes

Copilot reviewed 56 out of 61 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
ChatQnA/docker_compose/intel/cpu/xeon/compose_tgi.yaml	Updated Intel CPU image version for TGI service
ChatQnA/docker_compose/intel/cpu/xeon/compose_faqgen_tgi.yaml	Updated Intel CPU image version for TGI FAQ service
ChatQnA/docker_compose/intel/cpu/xeon/README.md	Updated documentation with new TGI CPU image version
ChatQnA/docker_compose/amd/gpu/rocm/*.yaml & README.md	Updated AMD/GPU (ROCm) image tags to version 2.4.1
AvatarChatbot/docker_compose/intel/hpu/gaudi/*.yaml & README.md	Updated Gaudi image to version 2.3.1
AvatarChatbot/docker_compose/intel/cpu/xeon/*.yaml & README.md	Updated Intel CPU image to version 2.4.1
AvatarChatbot/docker_compose/amd/gpu/rocm/compose.yaml	Updated AMD/GPU (ROCm) image from 2.3.1 to 2.4.1
AudioQnA/docker_compose/intel/cpu/xeon/compose_tgi.yaml	Updated Intel CPU image to version 2.4.1
AudioQnA/docker_compose/amd/gpu/rocm/README.md	Updated AMD/GPU (ROCm) image tag to version 2.4.1
AgentQnA/docker_compose/amd/gpu/rocm/compose.yaml	Updated image version from 3.0.0-rocm to 2.4.1-rocm
AgentQnA/docker_compose/amd/gpu/rocm/README.md	Updated documentation with new AMD/GPU (ROCm) image tag

Files not reviewed (5)

• AudioQnA/tests/test_compose_tgi_on_gaudi.sh: Language not supported
• AudioQnA/tests/test_compose_tgi_on_xeon.sh: Language not supported
• AvatarChatbot/tests/test_compose_on_gaudi.sh: Language not supported
• AvatarChatbot/tests/test_compose_on_rocm.sh: Language not supported
• AvatarChatbot/tests/test_compose_on_xeon.sh: Language not supported

Comments suppressed due to low confidence (1)

AgentQnA/docker_compose/amd/gpu/rocm/compose.yaml:4

• The image tag is downgraded from 3.0.0-rocm to 2.4.1-rocm. Please confirm if this downgrade is intentional or if the intended version should be maintained.

image: ghcr.io/huggingface/text-generation-inference:3.0.0-rocm

[chensuyue]

#1843

[xiguiw]

TGI Gaudi update to 2.3.1

log from
#1848

whiper load failed.
Transformer version mismatched?

  warnings.warn(
Traceback (most recent call last):
  File "/home/user/comps/third_parties/whisper/src/whisper_server.py", line 116, in <module>
    asr = WhisperModel(
  File "/home/user/comps/third_parties/whisper/src/whisper_model.py", line 27, in __init__
    from optimum.habana.transformers.modeling_utils import adapt_transformers_to_gaudi
  File "/usr/local/lib/python3.10/dist-packages/optimum/habana/__init__.py", line 19, in <module>
    from .transformers import (
  File "/usr/local/lib/python3.10/dist-packages/optimum/habana/transformers/__init__.py", line 20, in <module>
    from .trainer import GaudiTrainer
  File "/usr/local/lib/python3.10/dist-packages/optimum/habana/transformers/trainer.py", line 122, in <module>
    from .training_args import GaudiTrainingArguments
  File "/usr/local/lib/python3.10/dist-packages/optimum/habana/transformers/training_args.py", line 39, in <module>
    from transformers.training_args import (
ImportError: cannot import name '_VALID_DICT_FIELDS' from 'transformers.training_args' (/usr/local/lib/python3.10/dist-packages/transformers/training_args.py)
/usr/local/lib/python3.10/dist-packages/pydantic/_internal/_fields.py:160: UserWarning: Field "model_name_or_path" has conflict with protected namespace "model_".

[lianhao]

The whisper-gaudi issue is being tracked by opea-project/GenAIComps#1608