Unit Tests for On Device Sampling

[quic-sanising]

This PR adds the following Unit Tests for On Device Sampling:

1. test_sampler_transform: Test if SamplerTransform adds nodes at the output of a QEffForCausalLM model to enable the sampling of next tokens at the device (instead of the host) and returns the next tokens and/or probability distributions.
2. test_greedy_sampling: Test greedy sampling with QPC compiled with and without On Device Sampling.
3. test_random_sampling: Test random sampling with QPC compiled with and without On Device Sampling.

[quic-rishinr]

@quic-sanising can you add a small feature description under /docs/source/quick_start.md supported feature section? also provide the example script link in the description

[quic-sanising]

@quic-sanising can you add a small feature description under /docs/source/quick_start.md supported feature section? also provide the example script link in the description

Done

[quic-amitraj]

Please fix lint error.

[quic-sanising]

Please fix lint error.

@quic-amitraj The lint failures were happening because the linter is installing ruff v0.12.2 whereas the .pre-commit-config.yaml file has an older version of v0.5.2.

To fix the errors, we need to either install ruff v0.5.2 in the linter or update the .pre-commit-config.yaml file to version v0.12.2.

[ochougul]

why?

[quic-sanising]

Refer to my comment above: #463 (comment)

[ochougul]

why?

[quic-sanising]

Some of the inference tests in the later stages of CI were failing. From the logs, prompt was being passed twice to the same function. Not sure how this is not causing problems in the main branch.

prompts = 'My name is', device_id = None, runtime_ai100 = True
kwargs = {'prompt': 'My name is', 'prompts_txt_file_path': 'examples/prompts.txt'}

>           return QEfficient.cloud_ai_100_exec_kv(
                tokenizer,
                self.qpc_path,
                prompt=prompts,
                device_id=device_id,
                generation_len=generation_len,
                is_tlm=self.is_tlm,
                **kwargs,
            )
E           TypeError: QEfficient.generation.text_generation_inference.cloud_ai_100_exec_kv() got multiple values for keyword argument 'prompt'

@quic-rishinr please feel free to comment.

[ochougul]

are the print statements intentional here?

[quic-sanising]

Yes, they are intentional. But let me know if you want any of them removed.

[quic-hemagnih]

Anything pending for this PR? @ochougul @quic-sanising

[quic-sanising]

Anything pending for this PR? @ochougul @quic-sanising

Nothing from my end. Please go ahead with the merge if the CI has passed.