Introduction to Language Models

#deepLearning/llm/1

加载模型

当使用 LLM 时，需要加载两个 model 1. 调用的模型本身 2. tokenizer

from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    "microsoft/Phi-3-mini-4k-instruct",
    device_map="cuda",
    torch_dtype="auto",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-4k-instruct")

print(model)

Phi3ForCausalLM(
  (model): Phi3Model(
    (embed_tokens): Embedding(32064, 3072, padding_idx=32000)
    (embed_dropout): Dropout(p=0.0, inplace=False)
    (layers): ModuleList(
      (0-31): 32 x Phi3DecoderLayer(
        (self_attn): Phi3Attention(
          (o_proj): Linear(in_features=3072, out_features=3072, bias=False)
          (qkv_proj): Linear(in_features=3072, out_features=9216, bias=False)
          (rotary_emb): Phi3RotaryEmbedding()
        )
        (mlp): Phi3MLP(
          (gate_up_proj): Linear(in_features=3072, out_features=16384, bias=False)
          (down_proj): Linear(in_features=8192, out_features=3072, bias=False)
          (activation_fn): SiLU()
        )
        (input_layernorm): Phi3RMSNorm()
        (resid_attn_dropout): Dropout(p=0.0, inplace=False)
        (resid_mlp_dropout): Dropout(p=0.0, inplace=False)
        (post_attention_layernorm): Phi3RMSNorm()
      )
    )
    (norm): Phi3RMSNorm()
  )
  (lm_head): Linear(in_features=3072, out_features=32064, bias=False)
)

# 每一个模型的 special token 都是不一样的（但是大同小异）
tokenizer.special_tokens_map

{'bos_token': '<s>',
 'eos_token': '<|endoftext|>',
 'unk_token': '<unk>',
 'pad_token': '<|endoftext|>'}

Tokenizer负责将输入文本分割为token，然后将其提供给生成模型 # 构建 prompt 和 tokenizer 加载模型后，如何使用 3. 使用 generate() 4. 调用 pipeline

# The prompt (user input / query)
messages = [
    {"role": "user", "content": "Create a funny joke about chickens."}
]

# 构造格式
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

print(text)
print("====" * 10)
print(model_inputs)

<|user|>
Create a funny joke about chickens.<|end|>
<|assistant|>

========================================
{'input_ids': tensor([[32010,  6204,   263,  2090,  1460,  2958,   446,  1048,   521,   860,
           575, 29889, 32007, 32001]], device='cuda:0'), 'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]], device='cuda:0')}

# 推理
generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=512
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

# 解码
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)

Why did the chicken join the band? Because it had the drumsticks!

使用 transformers 的 pipeline 简化流程

from transformers import pipeline

# Create a pipeline
# return_full_text=False 表示只返回新生成的文本，不包含输入的prompt  
# return_full_text=True 则会返回完整文本，包含输入的prompt和生成的新文本
generator = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    return_full_text=False,
    max_new_tokens=500,
    do_sample=False
)

# The prompt (user input / query)
messages = [
    {"role": "user", "content": "Create a funny joke about chickens."}
]

# Generate output
output = generator(messages)
print(output[0]["generated_text"])

Why did the chicken join the band? Because it had the drumsticks!