transformers/docs/source/en/model_doc/llama2.md

This model was published in HF papers on 2023-07-18 and contributed to Hugging Face Transformers on 2023-07-18.

Llama 2

Llama 2 is a family of large language models, Llama 2 and Llama 2-Chat, available in 7B, 13B, and 70B parameters. The Llama 2 model mostly keeps the same architecture as Llama, but it is pretrained on more tokens, doubles the context length, and uses grouped-query attention (GQA) in the 70B model to improve inference.

Llama 2-Chat is trained with supervised fine-tuning (SFT), and reinforcement learning with human feedback (RLHF) - rejection sampling and proximal policy optimization (PPO) - is applied to the fine-tuned model to align the chat model with human preferences.

You can find all the original Llama 2 checkpoints under the Llama 2 Family collection.

Tip

Click on the Llama 2 models in the right sidebar for more examples of how to apply Llama to different language tasks.

The example below demonstrates how to generate text with [Pipeline], [AutoModel], and how to chat with Llama 2-Chat from the command line.

from transformers import pipeline


pipeline = pipeline(
    task="text-generation",
    model="meta-llama/Llama-2-7b-hf",
    device=0
)
pipeline("Plants create energy through a process known as")

from transformers import AutoModelForCausalLM, AutoTokenizer


tokenizer = AutoTokenizer.from_pretrained(
    "meta-llama/Llama-2-7b-hf",
)
model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-7b-hf",
    device_map="auto",
    attn_implementation="sdpa"
)
input_ids = tokenizer("Plants create energy through a process known as", return_tensors="pt").to(model.device)

output = model.generate(**input_ids, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))

transformers chat meta-llama/Llama-2-7b-chat-hf --dtype auto --attn_implementation flash_attention_2

Quantization reduces the memory burden of large models by representing the weights in a lower precision. Refer to the Quantization overview for more available quantization backends.

The example below uses torchao to only quantize the weights to int4.

# pip install torchao
from transformers import AutoModelForCausalLM, AutoTokenizer, TorchAoConfig


quantization_config = TorchAoConfig("int4_weight_only", group_size=128)
model = AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-2-13b-hf",
    device_map="auto",
    quantization_config=quantization_config
)

tokenizer = AutoTokenizer.from_pretrained("meta-llama/Llama-2-13b-hf")
input_ids = tokenizer("Plants create energy through a process known as", return_tensors="pt").to(model.device)

output = model.generate(**input_ids, cache_implementation="static")
print(tokenizer.decode(output[0], skip_special_tokens=True))

Use the AttentionMaskVisualizer to better understand what tokens the model can and cannot attend to.

from transformers.utils.attention_visualizer import AttentionMaskVisualizer


visualizer = AttentionMaskVisualizer("meta-llama/Llama-2-7b-hf")
visualizer("Plants create energy through a process known as")

Notes

• Setting config.pretraining_tp to a value besides 1 activates a more accurate but slower computation of the linear layers. This matches the original logits better.

• The original model uses pad_id = -1 to indicate a padding token. The Transformers implementation requires adding a padding token and resizing the token embedding accordingly.

  tokenizer.add_special_tokens({"pad_token":"<pad>"})
  # update model config with padding token
  model.config.pad_token_id
  

• It is recommended to initialize the embed_tokens layer with the following code to ensure encoding the padding token outputs zeros.

  self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.config.padding_idx)
  

• The tokenizer is a byte-pair encoding model based on SentencePiece. During decoding, if the first token is the start of the word (for example, "Banana"), the tokenizer doesn't prepend the prefix space to the string.

• Don't use the dtype parameter in [~AutoModel.from_pretrained] if you're using FlashAttention-2 because it only supports fp16 or bf16. You should use Automatic Mixed Precision, set fp16 or bf16 to True if using [Trainer], or use torch.autocast.

LlamaConfig

autodoc LlamaConfig

LlamaTokenizer

autodoc LlamaTokenizer - get_special_tokens_mask - save_vocabulary

LlamaTokenizerFast

autodoc LlamaTokenizerFast - get_special_tokens_mask - update_post_processor - save_vocabulary

LlamaModel

autodoc LlamaModel - forward

LlamaForCausalLM

autodoc LlamaForCausalLM - forward

LlamaForSequenceClassification

autodoc LlamaForSequenceClassification - forward