Sugato Ray sugatoray

🎯

Focusing

I am a Physicist turned Data Scientist + ML Practitioner. Research Interests: Data Science, ML, DL, Statistics, Math, Computing, LLMs.

sugatoray / export_locally.py

Created October 15, 2024 14:32 — forked from tomaarsen/export_locally.py

Export Sentence Transformer models to ONNX (+ optimization, quantization) & OpenVINO

	# requires sentence_transformers>=3.2.0
	from sentence_transformers import SentenceTransformer, export_optimized_onnx_model, export_dynamic_quantized_onnx_model

	# The model to export to ONNX (+ optimize, quantize), OpenVINO
	model_id = "mixedbread-ai/mxbai-embed-large-v1"
	# Where to save the exported models locally
	output_dir = model_id.replace("/", "-")

	onnx_model = SentenceTransformer(model_id, backend="onnx", model_kwargs={"export": True})
	onnx_model.save_pretrained(output_dir)

sugatoray / prompt.txt

Created October 6, 2024 14:53 — forked from philschmid/prompt.txt

	Begin by enclosing all thoughts within <thinking> tags, exploring multiple angles and approaches.
	Break down the solution into clear steps within <step> tags. Start with a 20-step budget, requesting more for complex problems if needed.
	Use <count> tags after each step to show the remaining budget. Stop when reaching 0.
	Continuously adjust your reasoning based on intermediate results and reflections, adapting your strategy as you progress.
	Regularly evaluate progress using <reflection> tags. Be critical and honest about your reasoning process.
	Assign a quality score between 0.0 and 1.0 using <reward> tags after each reflection. Use this to guide your approach:

	0.8+: Continue current approach
	0.5-0.7: Consider minor adjustments
	Below 0.5: Seriously consider backtracking and trying a different approach

sugatoray / pipeline_parallel.py

Created October 2, 2024 17:20 — forked from 3outeille/pipeline_parallel.py

Self contained example of how pipeline parallel works (AFAB and 1F1B) in 200 LOC

	#VERBOSE=0 torchrun --nproc_per_node 3 self_contained_pp_LOC.py
	import os, random, numpy as np, torch, torch.nn as nn, torch.distributed as dist, torch.nn.functional as F
	from torch.optim import AdamW
	from torch.utils.data import DataLoader, DistributedSampler
	from datasets import load_dataset
	from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer

	STEP, local_rank, world_size, verbose = 0, int(os.environ["LOCAL_RANK"]), int(os.environ["WORLD_SIZE"]), os.environ.get("VERBOSE", "0") == "1"

	def set_all_seed(seed):

sugatoray / l3min.py

Created August 16, 2024 05:18 — forked from awni/l3min.py

	"""
	A minimal, fast example generating text with Llama 3.1 in MLX.

	To run, install the requirements:

	pip install -U mlx transformers fire

	Then generate text with:

	python l3min.py "How tall is K2?"

sugatoray / mlx_lm_openai.md

Created August 16, 2024 05:14 — forked from awni/mlx_lm_openai.md

MLX LM with the OpenAI Python Package

Install MLX LM and openai:

pip install mlx-lm openai

sugatoray / mlx_lm_openai.md

Created August 16, 2024 05:13 — forked from awni/mlx_lm_openai.md

Install MLX LM and openai:

pip install mlx-lm openai

sugatoray / mflux_steps.md

Created August 16, 2024 03:37 — forked from awni/mflux_steps.md

This can take a little while:

# Make sure you have git-lfs installed (https://git-lfs.com)
git lfs install

git clone https://huggingface.co/black-forest-labs/FLUX.1-schnell

sugatoray / chunking-regex.ts

Created August 15, 2024 12:04 — forked from hanxiao/testRegex.js

Use regex to do chunking by using all semantic cues

	// Used in https://jina.ai/tokenizer (Aug. 14th version)
	// Define variables for magic numbers
	const MAX_HEADING_LENGTH = 6;
	const MAX_HEADING_CONTENT_LENGTH = 200;
	const MAX_HEADING_UNDERLINE_LENGTH = 200;
	const MAX_HTML_HEADING_ATTRIBUTES_LENGTH = 100;
	const MAX_LIST_ITEM_LENGTH = 200;
	const MAX_NESTED_LIST_ITEMS = 5;
	const MAX_LIST_INDENT_SPACES = 7;
	const MAX_BLOCKQUOTE_LINE_LENGTH = 200;

sugatoray / great_tables_co2.py

Created May 1, 2024 04:44 — forked from chalg/great_tables_co2.py

Showcase electricity consumption data for 2023 from ElectricityMaps via Python great_tables

	from great_tables import GT, md, html, system_fonts
	import pandas as pd

	power_cie_prepared_tbl = pd.read_csv("./data/2023_cie_power_cons.csv")

	# Create a Great Tables object
	ciep_gt_tbl = GT(data=power_cie_prepared_tbl)

	# Apply wider color ranges & formatting
	gt_tbl = ciep_gt_tbl \

sugatoray / .gitconfig

Created August 1, 2023 17:56 — forked from johnpolacek/.gitconfig

My current .gitconfig aliases