[Feature]: AutoDeploy: fake llm-models on cw-dfw

[lucaslie]

🚀 The feature, motivation and pitch

Let's update our fake llm-models dir on cw-dfw to be compatible all of AD's integration tests (see tests/integration/defs/accuracy and tests/integration/defs/examples)

Right now, we have all models and datasets in the llm-models-fake but seems there is some discrepancy due to those checkpoints in the accuracy.

We have seen multiple times that these tests pass in CI but not on cw-dfw with the llm-models-fake directory

Alternatives

No response

Additional context

No response

Before submitting a new issue...

• Make sure you already searched for relevant issues, and checked the documentation and examples for answers to frequently asked questions.

[lucaslie]

Useful mapping from HF checkpoint names to llm-models subdir:

TensorRT-LLM/tests/test_common/llm_data.py

Lines 28 to 46 in ff3a494

	HF_ID_TO_LLM_MODELS_SUBDIR = {
	"meta-llama/Meta-Llama-3.1-8B-Instruct": "llama-3.1-model/Llama-3.1-8B-Instruct",
	"meta-llama/Llama-3.1-8B-Instruct": "llama-3.1-model/Llama-3.1-8B-Instruct",
	"meta-llama/Llama-3.1-8B": "llama-3.1-model/Meta-Llama-3.1-8B",
	"TinyLlama/TinyLlama-1.1B-Chat-v1.0": "llama-models-v2/TinyLlama-1.1B-Chat-v1.0",
	"meta-llama/Llama-4-Scout-17B-16E-Instruct": "llama4-models/Llama-4-Scout-17B-16E-Instruct",
	"mistralai/Mixtral-8x7B-Instruct-v0.1": "Mixtral-8x7B-Instruct-v0.1",
	"mistralai/Mistral-Small-3.1-24B-Instruct-2503": "Mistral-Small-3.1-24B-Instruct-2503",
	"Qwen/Qwen3-30B-A3B": "Qwen3/Qwen3-30B-A3B",
	"Qwen/Qwen2.5-3B-Instruct": "Qwen2.5-3B-Instruct",
	"microsoft/Phi-3-mini-4k-instruct": "Phi-3/Phi-3-mini-4k-instruct",
	"deepseek-ai/DeepSeek-V3": "DeepSeek-V3",
	"deepseek-ai/DeepSeek-R1": "DeepSeek-R1/DeepSeek-R1",
	"ibm-ai-platform/Bamba-9B-v2": "Bamba-9B-v2",
	"nvidia/NVIDIA-Nemotron-Nano-12B-v2": "NVIDIA-Nemotron-Nano-12B-v2",
	"nvidia/NVIDIA-Nemotron-Nano-31B-A3-v3": "NVIDIA-Nemotron-Nano-31B-A3-v3",
	"nvidia/Nemotron-Nano-3-30B-A3.5B-dev-1024": "Nemotron-Nano-3-30B-A3.5B-dev-1024",
	"yuhuili/EAGLE3-LLaMA3.1-Instruct-8B": "EAGLE3-LLaMA3.1-Instruct-8B",
	}

[nvchenghaoz]

Fixed -
======================================================================================================================================================== 8 passed, 3 warnings in 875.46s (0:14:35) =========================================================================================================================================================

Two issues I found,

1. ckpt mismatch
2. The ccdv/cnn_dailymail requires the 3.0.0 folder instead of the original folder. Something wrong but once I copy the 3.0.0 into the ccdv/cnn_dailymaill folder the test passed.