Convert examples to ruff-format (#18400)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-05-26 16:57:54 +00:00
committed by GitHub
parent e7523c2e03
commit 27bebcd897
83 changed files with 2529 additions and 2405 deletions
@@ -18,8 +18,7 @@ Run:
"""
import torch
from transformers import (AutoModelForCausalLM, AutoTokenizer,
PreTrainedTokenizer)
from transformers import AutoModelForCausalLM, AutoTokenizer, PreTrainedTokenizer
from vllm import LLM
@@ -32,27 +31,29 @@ def init_tokenizer_and_llm(model_name: str):
return tokenizer, embedding_layer, llm
def get_prompt_embeds(chat: list[dict[str,
str]], tokenizer: PreTrainedTokenizer,
embedding_layer: torch.nn.Module):
token_ids = tokenizer.apply_chat_template(chat,
add_generation_prompt=True,
return_tensors='pt')
def get_prompt_embeds(
chat: list[dict[str, str]],
tokenizer: PreTrainedTokenizer,
embedding_layer: torch.nn.Module,
):
token_ids = tokenizer.apply_chat_template(
chat, add_generation_prompt=True, return_tensors="pt"
)
prompt_embeds = embedding_layer(token_ids).squeeze(0)
return prompt_embeds
def single_prompt_inference(llm: LLM, tokenizer: PreTrainedTokenizer,
embedding_layer: torch.nn.Module):
chat = [{
"role": "user",
"content": "Please tell me about the capital of France."
}]
def single_prompt_inference(
llm: LLM, tokenizer: PreTrainedTokenizer, embedding_layer: torch.nn.Module
):
chat = [{"role": "user", "content": "Please tell me about the capital of France."}]
prompt_embeds = get_prompt_embeds(chat, tokenizer, embedding_layer)
outputs = llm.generate({
"prompt_embeds": prompt_embeds,
})
outputs = llm.generate(
{
"prompt_embeds": prompt_embeds,
}
)
print("\n[Single Inference Output]")
print("-" * 30)
@@ -61,34 +62,26 @@ def single_prompt_inference(llm: LLM, tokenizer: PreTrainedTokenizer,
print("-" * 30)
def batch_prompt_inference(llm: LLM, tokenizer: PreTrainedTokenizer,
embedding_layer: torch.nn.Module):
chats = [[{
"role": "user",
"content": "Please tell me about the capital of France."
}],
[{
"role": "user",
"content": "When is the day longest during the year?"
}],
[{
"role": "user",
"content": "Where is bigger, the moon or the sun?"
}]]
def batch_prompt_inference(
llm: LLM, tokenizer: PreTrainedTokenizer, embedding_layer: torch.nn.Module
):
chats = [
[{"role": "user", "content": "Please tell me about the capital of France."}],
[{"role": "user", "content": "When is the day longest during the year?"}],
[{"role": "user", "content": "Where is bigger, the moon or the sun?"}],
]
prompt_embeds_list = [
get_prompt_embeds(chat, tokenizer, embedding_layer) for chat in chats
]
outputs = llm.generate([{
"prompt_embeds": embeds
} for embeds in prompt_embeds_list])
outputs = llm.generate([{"prompt_embeds": embeds} for embeds in prompt_embeds_list])
print("\n[Batch Inference Outputs]")
print("-" * 30)
for i, o in enumerate(outputs):
print(f"Q{i+1}: {chats[i][0]['content']}")
print(f"A{i+1}: {o.outputs[0].text}\n")
print(f"Q{i + 1}: {chats[i][0]['content']}")
print(f"A{i + 1}: {o.outputs[0].text}\n")
print("-" * 30)