Browse Source

Merge c9564a647a into b8b2f766e5

pull/319/merge
Craig Probus 3 weeks ago committed by GitHub
parent
commit
3af3d3741b
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
  1. 38
      week8/pricer_service2.py

38
week8/pricer_service2.py

@ -5,6 +5,7 @@ from modal import App, Volume, Image
app = modal.App("pricer-service")
image = Image.debian_slim().pip_install("huggingface", "torch", "transformers", "bitsandbytes", "accelerate", "peft")
image.add_local_python_source("hello", "llama")
secrets = [modal.Secret.from_name("hf-secret")]
# Constants
@ -17,26 +18,25 @@ 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
CACHE_DIR = "/cache"
QUESTION = "How much does this cost to the nearest dollar?"
PREFIX = "Price is $"
@app.cls(image=image, secrets=secrets, gpu=GPU, timeout=1800)
#NOTE: Use the pre-configured hf-hub-cache Volume...
hf_cache_volume = Volume.from_name("hf-hub-cache")
@app.cls(
image=image.env({"HF_HUB_CACHE": CACHE_DIR}),
secrets=secrets,
gpu=GPU,
timeout=1800,
volumes={CACHE_DIR: hf_cache_volume}
)
class Pricer:
@modal.build()
def download_model_to_folder(self):
from huggingface_hub import snapshot_download
import os
os.makedirs(MODEL_DIR, exist_ok=True)
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
@ -48,20 +48,17 @@ 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 = AutoTokenizer.from_pretrained(BASE_MODEL)
self.tokenizer.pad_token = self.tokenizer.eos_token
self.tokenizer.padding_side = "right"
self.base_model = AutoModelForCausalLM.from_pretrained(
BASE_DIR,
BASE_MODEL,
quantization_config=quant_config,
device_map="auto"
)
self.fine_tuned_model = PeftModel.from_pretrained(self.base_model, FINETUNED_DIR, revision=REVISION)
self.fine_tuned_model = PeftModel.from_pretrained(self.base_model, FINETUNED_MODEL, revision=REVISION)
@modal.method()
def price(self, description: str) -> float:
@ -85,5 +82,4 @@ class Pricer:
@modal.method()
def wake_up(self) -> str:
return "ok"
return "ok"
Loading…
Cancel
Save