update test_models.py to conform to new quantization config

This commit is contained in:
Sunny Liu
2025-04-21 11:34:37 -04:00
parent ffd4ef1ece
commit 9be971d47c

View File

@@ -21,8 +21,10 @@ class TestModelsUtils:
"base_model": "JackFram/llama-68m", "base_model": "JackFram/llama-68m",
"model_type": "LlamaForCausalLM", "model_type": "LlamaForCausalLM",
"tokenizer_type": "LlamaTokenizer", "tokenizer_type": "LlamaTokenizer",
"load_in_8bit": True, "quantization": {
"load_in_4bit": False, "backend": "bnb",
"bits": 8,
},
"adapter": "lora", "adapter": "lora",
"flash_attention": False, "flash_attention": False,
"sample_packing": True, "sample_packing": True,