Improve WOQ model saving and loading

[changwangss]

Type of Change

Code Changes

1. Change all algo name, RTN to Rtn, AWQ to Awq, TEQ to Teq, GPTQ no change, AUTOROUND to AutoRound
2. For rtn/awq/teq algo, backend get scales and zeros from INC, not computed by backend.
3. Saving: checkpoint weightformat, config.json and quantize_config.json to Transformers_like API request.

python run_generation.py --model facebook/opt-125m --woq --bits 4 --output_dir "saved_results"

4. Loading: Transformers_like

python run_generation.py --model "saved_results"

Local validation

Rtn

# int4_clip
python run_generation.py --model facebook/opt-125m --woq --bits 4
python run_generation.py --model facebook/opt-125m --woq --bits 4  --scheme "asym"
# fp8_e5m2
python run_generation.py --model facebook/opt-125m --woq --bits 8 --weight_dtype "fp8_e5m2"

Awq

python run_generation.py --model facebook/opt-125m --woq --woq_algo "Awq" --bits 4 --scheme "asym"

Teq

python run_generation.py --model facebook/opt-125m --woq --woq_algo "Teq" --bits 4 --scheme "asym"

GPTQ

python run_generation.py --model facebook/opt-125m --woq --woq_algo "GPTQ" --bits 4 --scheme "asym" --desc_act --blocksize 32 --nsamples 32 --max_input_length 256 --static_groups

AutoRound

python run_generation.py --model facebook/opt-125m --woq --woq_algo "AutoRound" --bits 4 --scheme "asym" --calib_len 1024 --nsamples 128

Description

detail description
JIRA ticket: xxx

Expected Behavior & Potential Risk

the expected behavior that triggered by this PR

How has this PR been tested?

how to reproduce the test (including hardware information)

Dependency Change?

any library dependency introduced or removed

[github-actions]

⚡ Required checks status: All passing 🟢

Groups summary

🟢 Format Scan Tests workflow

Check ID	Status	Error details
format-scan (pylint)	success		✅
format-scan (bandit)	success		✅
format-scan (clangformat)	success		✅
format-scan (cloc)	success		✅
format-scan (cpplint)	success		✅

These checks are required after the changes to intel_extension_for_transformers/llm/operator/csrc/dispatcher/include/bestla_packq_impl.hpp, intel_extension_for_transformers/llm/operator/csrc/dispatcher/src/bestla_packq_impl.cpp, intel_extension_for_transformers/llm/operator/csrc/qbits_ut/test_packq.py, intel_extension_for_transformers/llm/quantization/gptq_utils.py, intel_extension_for_transformers/llm/quantization/nn/modules.py, intel_extension_for_transformers/llm/quantization/utils.py, intel_extension_for_transformers/transformers/modeling/modeling_auto.py, intel_extension_for_transformers/transformers/utils/config.py.

🟢 Optimize Unit Test workflow

Check ID	Status	Error details
optimize-unit-test-baseline	success		✅
optimize-unit-test-PR-test	success		✅
Genreate-OptimizeUT-Report	success		✅

These checks are required after the changes to intel_extension_for_transformers/transformers/modeling/modeling_auto.py, intel_extension_for_transformers/transformers/utils/config.py, intel_extension_for_transformers/llm/quantization/gptq_utils.py, intel_extension_for_transformers/llm/quantization/nn/modules.py, intel_extension_for_transformers/llm/quantization/utils.py, intel_extension_for_transformers/llm/operator/csrc/dispatcher/include/bestla_packq_impl.hpp, intel_extension_for_transformers/llm/operator/csrc/dispatcher/src/bestla_packq_impl.cpp, intel_extension_for_transformers/llm/operator/csrc/qbits_ut/test_packq.py, tests/CI/test_quantization.py, tests/CI/test_weight_only.py.

🟢 NeuralChat Unit Test

Check ID	Status	Error details
neuralchat-unit-test-baseline	success		✅
neuralchat-unit-test-PR-test	success		✅
Generate-NeuralChat-Report	success		✅

These checks are required after the changes to intel_extension_for_transformers/llm/quantization/gptq_utils.py, intel_extension_for_transformers/llm/quantization/nn/modules.py, intel_extension_for_transformers/llm/quantization/utils.py, intel_extension_for_transformers/transformers/modeling/modeling_auto.py, intel_extension_for_transformers/transformers/utils/config.py.

🟢 Chat Bot Test workflow

Check ID	Status	Error details
call-inference-llama-2-7b-chat-hf / inference test	success		✅
call-inference-mpt-7b-chat / inference test	success		✅

These checks are required after the changes to intel_extension_for_transformers/llm/quantization/gptq_utils.py, intel_extension_for_transformers/llm/quantization/nn/modules.py, intel_extension_for_transformers/llm/quantization/utils.py, intel_extension_for_transformers/transformers/modeling/modeling_auto.py, intel_extension_for_transformers/transformers/utils/config.py.

---

Thank you for your contribution! 💜

Note
This comment is automatically generated and updates for 360 minutes every 180 seconds. If you have any other questions, contact VincyZhang or XuehaoSun for help.

[a32543254]

LGTM

[a32543254]

After this pr, we can load a HF gptq/ awq model for both CPU and GPU to do inference?

[changwangss]

nf4, fp4, int4_fullrange is waiting kernel ready.
int8 issue will be fixed later.
docs and readme will be updated by next PR.