juliensimon

User: write a fictional technical discussion between Alice, a Senior Machine Learning Engineer working at a Fortune 500 company in the Telco industry, and Bob, an Arcee.ai pre-sales engineer. Alice is a potential customer of Arcee.ai, and she’s trying to figure out if Arcee Cloud is a good fit to help her build a customer support chatbot to offload their existing call centers. Bob should focus on understanding the customer pain points and see how Arcee can help answer them. Bob should explain how continuous pre-training, model merging and instruction fine-tuning should help Alice tailor her models on company and customer data with a high level of accuracy.

Alice: Hi Bob, thanks for taking the time to meet with me. I've been looking into how to improve our customer support chatbot. We currently rely heavily on call centers, and I've been considering using an AI-powered chatbot instead. Arcee Cloud caught my eye, but I'm still a bit skeptical about its capabilities. Could you tell me more about how Arcee Cloud

juliensimon

User: Please write a marketing pitch for a new SaaS AI platform called Arcee Cloud. We will send this pitch by email to business and technical decision-makers, so make it sound exciting and convincing. The contact email is sales@arcee.ai. Feel free to use emojis as appopriate. Arcee Cloud makes it simple for enterprise users to tailor open-source small language models to their own domain knowledge, in order to build high-quality, cost-effective and secure AI solutions.

Scribe: 🌟 Introducing Arcee Cloud: The Future of AI for Every Enterprise 🌟

Are you ready to unlock the full potential of Artificial Intelligence in your enterprise? Look no further than **Arcee Cloud** - the cutting-edge SaaS AI platform that redefines the game with its unique and powerful approach.

🚀 **Elevate Your AI Game** 🚀

Arcee Cloud empowers you to craft bespoke AI solutions tailored to your unique domain knowledge, making it easier than ever to integrate AI into your business strategy. Our platform is designed to be simple, intuitiv

juliensimon

LANGUAGE PRETRAINING

python run_clm.py \
    --model_name_or_path gpt2 \
    --dataset_name wikitext \
    --dataset_config_name wikitext-103-raw-v1 \
   --num_train_epochs 10 \
    --per_device_train_batch_size 8 \
    --per_device_eval_batch_size 8 \
    --do_train \

juliensimon

from datasets import load_dataset
from sentence_transformers.losses import CosineSimilarityLoss

from setfit import SetFitModel, SetFitTrainer

dataset = load_dataset("yelp_polarity")
print(dataset)

# Select N examples per class (8 in this case)
train_ds = dataset["train"].shuffle(seed=42).select(range(8 * 2))

juliensimon

import time

import numpy as np
import torch
from transformers import pipeline


def benchmark(pipeline, data, iterations=1000):
    # Warmup
    for i in range(100):

juliensimon

### CREATE NOTEBOOK INSTANCE

export HOME=/home/ec2-user

# Install and enable Git LFS
curl -s https://packagecloud.io/install/repositories/github/git-lfs/script.rpm.sh | sudo bash
sudo yum install git-lfs -y 
git lfs install

juliensimon

Here are the setup instructions. Please reply in the thread if you have questions or issues.
Using your own AWS account, log in to the AWS console at https://console.aws.amazon.com/sagemaker.
Select the « Ireland » region in the top-right corner
Go to « Notebook / Notebook instances »
Click on « Create notebook instance »
« Notebook instance name »
          Type a name for your instance, e.g « workshop-instance »
« Notebook instance type »
          Select « ml.t2.medium ». No need for anything bigger
« IAM role »

juliensimon

last_epoch = all_preds[epochs-1].detach().numpy()
predicted_class = np.argmax(last_epoch, axis=-1)
print(predicted_class)

[0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 1 0 1 0 1 1 1 1 1 1 0 1 0 1 1 1]

juliensimon

optimizer = torch.optim.Adam(net.parameters(), lr=0.001)
all_preds = []
epochs = 50

for epoch in range(epochs):
    preds = net(G, inputs)
    all_preds.append(preds)
    # we only compute loss for labeled nodes
    loss = F.cross_entropy(preds[labeled_nodes], labels)
    # PyTorch accumulates gradients by default, we need to zero them

juliensimon

inputs = torch.eye(node_count)
labeled_nodes = torch.tensor([0, 33])
labels = torch.tensor([0,1])