Small dummy LLama2-type Model useable for Unit/Integration tests. Suitable for CPU only machines, see H2O LLM Studio for an example integration test.
Model was created as follows:
from transformers import AutoConfig, AutoTokenizer, AutoModelForCausalLM
repo_name = "h2oai/llama2-0b-unit-test"
model_name = "h2oai/h2ogpt-4096-llama2-7b-chat"
config = AutoConfig.from_pretrained(model_name)
config.hidden_size = 12
config.max_position_embeddings = 1024
config.intermediate_size = 24
config.num_attention_heads = 2
config.num_hidden_layers = 2
config.num_key_value_heads = 2
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_config(config)
print(model.num_parameters()) # 770_940
model.push_to_hub(repo_name, private=False)
tokenizer.push_to_hub(repo_name, private=False)
config.push_to_hub(repo_name, private=False)
Use the following configuration in H2O LLM Studio to run a complete experiment in 5 seconds using the default dataset and default settings otherwise:
Validation Size: 0.1
Data Sample: 0.1
Max Length Prompt: 32
Max Length Answer: 32
Max Length: 64
Backbone Dtype: float16
Gradient Checkpointing: False
Batch Size: 8
Max Length Inference: 16