Consolidate Llama model usage in tests (#13094)
This commit is contained in:
@@ -26,12 +26,12 @@ os.environ["TOKENIZERS_PARALLELISM"] = "true"
|
||||
# Test FP8 checkpoint w. fp8_e4m3 kv-cache scaling factors.
|
||||
("fp8_e4m3", "meta-llama/Llama-3.2-1B-Instruct",
|
||||
"nm-testing/Llama-3.2-1B-Instruct-FP8-KV"),
|
||||
# Test FP16 checkpoint w. fp8_e5m2 kv-cache.
|
||||
# Test BF16 checkpoint w. fp8_e5m2 kv-cache.
|
||||
("fp8_e5m2", "meta-llama/Llama-3.2-1B-Instruct",
|
||||
"meta-llama/Llama-3.2-1B-Instruct"),
|
||||
# Test FP16 checkpoint w. fp8_e4m3 kv-cache scaling factors in json.
|
||||
("fp8_e4m3", "meta-llama/Llama-2-7b-chat-hf",
|
||||
"meta-llama/Llama-2-7b-chat-hf")
|
||||
# Test BF16 checkpoint w. fp8_e4m3 kv-cache scaling factors in json.
|
||||
("fp8_e4m3", "meta-llama/Llama-3.2-1B-Instruct",
|
||||
"meta-llama/Llama-3.2-1B-Instruct")
|
||||
])
|
||||
# Due to low-precision numerical divergence, we only test logprob of 4 tokens
|
||||
@pytest.mark.parametrize("max_tokens", [4])
|
||||
|
||||
@@ -141,7 +141,7 @@ _TEXT_GENERATION_EXAMPLE_MODELS = {
|
||||
"JAISLMHeadModel": _HfExamplesInfo("inceptionai/jais-13b-chat"),
|
||||
"JambaForCausalLM": _HfExamplesInfo("ai21labs/AI21-Jamba-1.5-Mini",
|
||||
extras={"tiny": "ai21labs/Jamba-tiny-dev"}), # noqa: E501
|
||||
"LlamaForCausalLM": _HfExamplesInfo("meta-llama/Meta-Llama-3-8B"),
|
||||
"LlamaForCausalLM": _HfExamplesInfo("meta-llama/Llama-3.2-1B-Instruct"),
|
||||
"LLaMAForCausalLM": _HfExamplesInfo("decapoda-research/llama-7b-hf",
|
||||
is_available_online=False),
|
||||
"MambaForCausalLM": _HfExamplesInfo("state-spaces/mamba-130m-hf"),
|
||||
|
||||
Reference in New Issue
Block a user