saimachi

///////////////////////////////////////////////////////////////////////////////
// Author:      Sai Machiraju
// Course:      ECE8893 - Parallel Programming for FPGAs
// Filename:    real_matmul.cpp
// Description: Perform matrix multiplication with real values
//
// Note:        You are free to modify this code to optimize your design.
///////////////////////////////////////////////////////////////////////////////

#include "real.h"

saimachi

import numpy as np
import pynq

overlay = pynq.Overlay('matmul.bit')

top_ip = overlay.real_matmul_0

a_matrix = np.random.randint(10, size=(100, 150))
b_matrix = np.random.randint(10, size=(150, 200))

saimachi

+---------------------------------------------------------------------------------------+------+------+---------+-----------+----------+---------+------+----------+---------+-----------+-------------+-------------+-----+
|                                        Modules                                        | Issue|      | Latency |  Latency  | Iteration|         | Trip |          |         |           |             |             |     |
|                                        & Loops                                        | Type | Slack| (cycles)|    (ns)   |  Latency | Interval| Count| Pipelined|  BRAM   |    DSP    |      FF     |     LUT     | URAM|
+---------------------------------------------------------------------------------------+------+------+---------+-----------+----------+---------+------+----------+---------+-----------+-------------+-------------+-----+
|+ real_matmul                                                                          |     -|  0.00|   787647|  7.876e+06|         -|   7

saimachi

+---------------------------------------------------------+------+------+---------+-----------+----------+---------+-------+----------+----------+---------+-----------+-----------+-----+
|                         Modules                         | Issue|      | Latency |  Latency  | Iteration|         |  Trip |          |          |         |           |           |     |
|                         & Loops                         | Type | Slack| (cycles)|    (ns)   |  Latency | Interval| Count | Pipelined|   BRAM   |   DSP   |     FF    |    LUT    | URAM|
+---------------------------------------------------------+------+------+---------+-----------+----------+---------+-------+----------+----------+---------+-----------+-----------+-----+
|+ real_matmul                                            |     -|  0.00|  3310043|  3.310e+07|         -|  3310044|      -|        no|  82 (29%)|   6 (2%)|  2026 (1%)|  2923 (5%)|    -|
| + real_matmul_Pipeline_MAT_A_ROWS_MAT_A_COLS            |     -|  0.00|    15004|  1.50

saimachi

///////////////////////////////////////////////////////////////////////////////
// Author:      <>
// Course:      ECE8893 - Parallel Programming for FPGAs
// Filename:    real_matmul.cpp
// Description: Perform matrix multiplication with real values
//
// Note:        You are free to modify this code to optimize your design.
///////////////////////////////////////////////////////////////////////////////

#include "real.h"

saimachi

{
  "input_model": {
    "type": "PyTorchModel",
    "config": {
      "hf_config": {
        "model_name": "model-cache/microsoft/phi-1_5",
        "task": "text-generation",
        "model_loading_args": {
          "trust_remote_code": true
        }

saimachi

import csv
import json
from transformers import AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained('microsoft/phi-1_5')
prompt_str = '### Wikipedia comment: {text}\n### Classifications:\n{classifications}'

if __name__ == '__main__':
    # JSONL format
    output = []

saimachi