brad/LLM_Engineering_OLD
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge pull request #261 from kellewic/community-contributions-kellewic

Browse Source 
Changed to use modal.Volume instead of modal.build()
This commit is contained in:
Ed Donner
2025-04-30 14:53:44 -04:00
committed by GitHub
parent 6b9a0a76a6 79948e3b71
commit 0360cc4d75
1 changed files with 28 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

59
week8/pricer_service2.py
View File

@@ -1,15 +1,22 @@
import modal
from modal import App, Volume, Image
from pathlib import PurePosixPath
# Setup - define our infrastructure with code!
app = modal.App("pricer-service")
image = Image.debian_slim().pip_install("huggingface", "torch", "transformers", "bitsandbytes", "accelerate", "peft")
secrets = [modal.Secret.from_name("hf-secret")]
secrets = [modal.Secret.from_name("huggingface-secret")]
# Constants
image = modal.Image.debian_slim().pip_install(
"huggingface", "torch", "transformers", "bitsandbytes",
"accelerate", "peft", "huggingface_hub[hf_transfer]"
).env({"HF_HUB_ENABLE_HF_TRANSFER": "1"})
# This is where we cache model files to avoid redownloading each time a container is started
hf_cache_vol = modal.Volume.from_name("hf-cache", create_if_missing=True)
GPU = "T4"
# Keep N containers active to avoid cold starts
MIN_CONTAINERS = 0
BASE_MODEL = "meta-llama/Meta-Llama-3.1-8B"
PROJECT_NAME = "pricer"
HF_USER = "ed-donner" # your HF name here! Or use mine if you just want to reproduce my results.
@@ -17,30 +24,28 @@ RUN_NAME = "2024-09-13_13.04.39"
PROJECT_RUN_NAME = f"{PROJECT_NAME}-{RUN_NAME}"
REVISION = "e8d637df551603dc86cd7a1598a8f44af4d7ae36"
FINETUNED_MODEL = f"{HF_USER}/{PROJECT_RUN_NAME}"
MODEL_DIR = "hf-cache/"
BASE_DIR = MODEL_DIR + BASE_MODEL
FINETUNED_DIR = MODEL_DIR + FINETUNED_MODEL
# Mount for cache location
MODEL_DIR = PurePosixPath("/models")
BASE_DIR = MODEL_DIR / BASE_MODEL
FINETUNED_DIR = MODEL_DIR / FINETUNED_MODEL
QUESTION = "How much does this cost to the nearest dollar?"
PREFIX = "Price is $"
@app.cls(image=image, secrets=secrets, gpu=GPU, timeout=1800)
@app.cls(image=image, secrets=secrets, gpu=GPU, timeout=1800, min_containers=MIN_CONTAINERS, volumes={MODEL_DIR: hf_cache_vol})
class Pricer:
@modal.build()
def download_model_to_folder(self):
@modal.enter()
def setup(self):
import torch
from huggingface_hub import snapshot_download
import os
os.makedirs(MODEL_DIR, exist_ok=True)
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig
from peft import PeftModel
# Download and cache model files to the volume
snapshot_download(BASE_MODEL, local_dir=BASE_DIR)
snapshot_download(FINETUNED_MODEL, revision=REVISION, local_dir=FINETUNED_DIR)
@modal.enter()
def setup(self):
import os
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig, set_seed
from peft import PeftModel
# Quant Config
quant_config = BitsAndBytesConfig(
load_in_4bit=True,
@@ -48,9 +53,8 @@ class Pricer:
bnb_4bit_compute_dtype=torch.bfloat16,
bnb_4bit_quant_type="nf4"
)
# Load model and tokenizer
self.tokenizer = AutoTokenizer.from_pretrained(BASE_DIR)
self.tokenizer.pad_token = self.tokenizer.eos_token
self.tokenizer.padding_side = "right"
@@ -65,11 +69,8 @@ class Pricer:
@modal.method()
def price(self, description: str) -> float:
import os
import re
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig, set_seed
from peft import PeftModel
import re, torch
from transformers import set_seed
set_seed(42)
prompt = f"{QUESTION}\n\n{description}\n\n{PREFIX}"
@@ -83,7 +84,3 @@ class Pricer:
match = re.search(r"[-+]?\d*\.\d+|\d+", contents)
return float(match.group()) if match else 0
@modal.method()
def wake_up(self) -> str:
return "ok"