mirror of
https://github.com/zkonduit/ezkl.git
synced 2026-01-13 00:08:12 -05:00
Compare commits
13 Commits
v16.2.5
...
ac/gen-ran
| Author | SHA1 | Date | |
|---|---|---|---|
|
0ecee16120 | ||
|
|
6752a05f19 | ||
|
|
8ba4a853ad | ||
|
|
343cdf21fb | ||
|
|
8d623e59ef | ||
|
|
6096df7853 | ||
|
|
adb9bf49f4 | ||
|
|
03aefb85eb | ||
|
|
e86caca8b6 | ||
|
|
c839a30ae6 | ||
|
|
352812b9ac | ||
|
|
d48d0b0b3e | ||
|
|
8b223354cc |
1
.github/workflows/pypi-gpu.yml
vendored
1
.github/workflows/pypi-gpu.yml
vendored
@@ -34,6 +34,7 @@ jobs:
|
||||
run: |
|
||||
mv pyproject.toml pyproject.toml.orig
|
||||
sed "s/ezkl/ezkl-gpu/" pyproject.toml.orig >pyproject.toml
|
||||
sed "s/0\\.0\\.0/${RELEASE_TAG//v}/" pyproject.toml.orig >pyproject.toml
|
||||
|
||||
- uses: actions-rs/toolchain@v1
|
||||
with:
|
||||
|
||||
9
.github/workflows/pypi.yml
vendored
9
.github/workflows/pypi.yml
vendored
@@ -233,6 +233,14 @@ jobs:
|
||||
python-version: 3.12
|
||||
architecture: x64
|
||||
|
||||
- name: Set pyproject.toml version to match github tag
|
||||
shell: bash
|
||||
env:
|
||||
RELEASE_TAG: ${{ github.ref_name }}
|
||||
run: |
|
||||
mv pyproject.toml pyproject.toml.orig
|
||||
sed "s/0\\.0\\.0/${RELEASE_TAG//v}/" pyproject.toml.orig >pyproject.toml
|
||||
|
||||
- name: Set Cargo.toml version to match github tag
|
||||
shell: bash
|
||||
env:
|
||||
@@ -242,7 +250,6 @@ jobs:
|
||||
sed "s/0\\.0\\.0/${RELEASE_TAG//v}/" Cargo.toml.orig >Cargo.toml
|
||||
mv Cargo.lock Cargo.lock.orig
|
||||
sed "s/0\\.0\\.0/${RELEASE_TAG//v}/" Cargo.lock.orig >Cargo.lock
|
||||
|
||||
- name: Install required libraries
|
||||
shell: bash
|
||||
run: |
|
||||
|
||||
4
Cargo.lock
generated
4
Cargo.lock
generated
@@ -2377,7 +2377,7 @@ dependencies = [
|
||||
[[package]]
|
||||
name = "halo2_gadgets"
|
||||
version = "0.2.0"
|
||||
source = "git+https://github.com/zkonduit/halo2#0654e92bdf725fd44d849bfef3643870a8c7d50b"
|
||||
source = "git+https://github.com/zkonduit/halo2#d7ecad83c7439fa1cb450ee4a89c2d0b45604ceb"
|
||||
dependencies = [
|
||||
"arrayvec 0.7.4",
|
||||
"bitvec",
|
||||
@@ -2394,7 +2394,7 @@ dependencies = [
|
||||
[[package]]
|
||||
name = "halo2_proofs"
|
||||
version = "0.3.0"
|
||||
source = "git+https://github.com/zkonduit/halo2#0654e92bdf725fd44d849bfef3643870a8c7d50b#0654e92bdf725fd44d849bfef3643870a8c7d50b"
|
||||
source = "git+https://github.com/zkonduit/halo2#ee4e1a09ebdb1f79f797685b78951c6034c430a6#ee4e1a09ebdb1f79f797685b78951c6034c430a6"
|
||||
dependencies = [
|
||||
"bincode",
|
||||
"blake2b_simd",
|
||||
|
||||
@@ -280,7 +280,10 @@ no-update = []
|
||||
|
||||
|
||||
[patch.'https://github.com/zkonduit/halo2']
|
||||
halo2_proofs = { git = "https://github.com/zkonduit/halo2#0654e92bdf725fd44d849bfef3643870a8c7d50b", package = "halo2_proofs" }
|
||||
halo2_proofs = { git = "https://github.com/zkonduit/halo2#ee4e1a09ebdb1f79f797685b78951c6034c430a6", package = "halo2_proofs" }
|
||||
|
||||
[patch.'https://github.com/zkonduit/halo2#0654e92bdf725fd44d849bfef3643870a8c7d50b']
|
||||
halo2_proofs = { git = "https://github.com/zkonduit/halo2#ee4e1a09ebdb1f79f797685b78951c6034c430a6", package = "halo2_proofs" }
|
||||
|
||||
[patch.crates-io]
|
||||
uniffi_testing = { git = "https://github.com/ElusAegis/uniffi-rs", branch = "feat/testing-feature-build-fix" }
|
||||
|
||||
@@ -2,7 +2,7 @@
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 1,
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [
|
||||
{
|
||||
@@ -54,7 +54,7 @@
|
||||
" gip_run_args.param_scale = 19\n",
|
||||
" gip_run_args.logrows = 8\n",
|
||||
" run_args = ezkl.gen_settings(py_run_args=gip_run_args)\n",
|
||||
" ezkl.get_srs(commitment=ezkl.PyCommitments.KZG)\n",
|
||||
" await ezkl.get_srs(commitment=ezkl.PyCommitments.KZG)\n",
|
||||
" ezkl.compile_circuit()\n",
|
||||
" res = await ezkl.gen_witness()\n",
|
||||
" print(res)\n",
|
||||
|
||||
@@ -1,279 +1,284 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "cf69bb3f-94e6-4dba-92cd-ce08df117d67",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Linear Regression\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"Sklearn based models are slightly finicky to get into a suitable onnx format. \n",
|
||||
"This notebook showcases how to do so using the `hummingbird-ml` python package ! "
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "95613ee9",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# check if notebook is in colab\n",
|
||||
"try:\n",
|
||||
" # install ezkl\n",
|
||||
" import google.colab\n",
|
||||
" import subprocess\n",
|
||||
" import sys\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"ezkl\"])\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"onnx\"])\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"hummingbird-ml\"])\n",
|
||||
"\n",
|
||||
"# rely on local installation of ezkl if the notebook is not in colab\n",
|
||||
"except:\n",
|
||||
" pass\n",
|
||||
"\n",
|
||||
"import os\n",
|
||||
"import torch\n",
|
||||
"import ezkl\n",
|
||||
"import json\n",
|
||||
"from hummingbird.ml import convert\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# here we create and (potentially train a model)\n",
|
||||
"\n",
|
||||
"# make sure you have the dependencies required here already installed\n",
|
||||
"import numpy as np\n",
|
||||
"from sklearn.linear_model import LinearRegression\n",
|
||||
"X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])\n",
|
||||
"# y = 1 * x_0 + 2 * x_1 + 3\n",
|
||||
"y = np.dot(X, np.array([1, 2])) + 3\n",
|
||||
"reg = LinearRegression().fit(X, y)\n",
|
||||
"reg.score(X, y)\n",
|
||||
"\n",
|
||||
"circuit = convert(reg, \"torch\", X[:1]).model\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "b37637c4",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"model_path = os.path.join('network.onnx')\n",
|
||||
"compiled_model_path = os.path.join('network.compiled')\n",
|
||||
"pk_path = os.path.join('test.pk')\n",
|
||||
"vk_path = os.path.join('test.vk')\n",
|
||||
"settings_path = os.path.join('settings.json')\n",
|
||||
"\n",
|
||||
"witness_path = os.path.join('witness.json')\n",
|
||||
"data_path = os.path.join('input.json')"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "82db373a",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"\n",
|
||||
"\n",
|
||||
"# export to onnx format\n",
|
||||
"# !!!!!!!!!!!!!!!!! This will flash a warning but it is fine !!!!!!!!!!!!!!!!!!!!!\n",
|
||||
"\n",
|
||||
"# Input to the model\n",
|
||||
"shape = X.shape[1:]\n",
|
||||
"x = torch.rand(1, *shape, requires_grad=True)\n",
|
||||
"torch_out = circuit(x)\n",
|
||||
"# Export the model\n",
|
||||
"torch.onnx.export(circuit, # model being run\n",
|
||||
" # model input (or a tuple for multiple inputs)\n",
|
||||
" x,\n",
|
||||
" # where to save the model (can be a file or file-like object)\n",
|
||||
" \"network.onnx\",\n",
|
||||
" export_params=True, # store the trained parameter weights inside the model file\n",
|
||||
" opset_version=10, # the ONNX version to export the model to\n",
|
||||
" do_constant_folding=True, # whether to execute constant folding for optimization\n",
|
||||
" input_names=['input'], # the model's input names\n",
|
||||
" output_names=['output'], # the model's output names\n",
|
||||
" dynamic_axes={'input': {0: 'batch_size'}, # variable length axes\n",
|
||||
" 'output': {0: 'batch_size'}})\n",
|
||||
"\n",
|
||||
"d = ((x).detach().numpy()).reshape([-1]).tolist()\n",
|
||||
"\n",
|
||||
"data = dict(input_shapes=[shape],\n",
|
||||
" input_data=[d],\n",
|
||||
" output_data=[((o).detach().numpy()).reshape([-1]).tolist() for o in torch_out])\n",
|
||||
"\n",
|
||||
"# Serialize data into file:\n",
|
||||
"json.dump(data, open(\"input.json\", 'w'))\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "d5e374a2",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"!RUST_LOG=trace\n",
|
||||
"# TODO: Dictionary outputs\n",
|
||||
"res = ezkl.gen_settings(model_path, settings_path)\n",
|
||||
"assert res == True\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"cal_path = os.path.join(\"calibration.json\")\n",
|
||||
"\n",
|
||||
"data_array = (torch.randn(20, *shape).detach().numpy()).reshape([-1]).tolist()\n",
|
||||
"\n",
|
||||
"data = dict(input_data = [data_array])\n",
|
||||
"\n",
|
||||
"# Serialize data into file:\n",
|
||||
"json.dump(data, open(cal_path, 'w'))\n",
|
||||
"\n",
|
||||
"res = await ezkl.calibrate_settings(data_path, model_path, settings_path, \"resources\")\n",
|
||||
"assert res == True\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "3aa4f090",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"res = ezkl.compile_circuit(model_path, compiled_model_path, settings_path)\n",
|
||||
"assert res == True"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "8b74dcee",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# srs path\n",
|
||||
"res = await ezkl.get_srs( settings_path)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "18c8b7c7",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# now generate the witness file \n",
|
||||
"\n",
|
||||
"res = await ezkl.gen_witness(data_path, compiled_model_path, witness_path)\n",
|
||||
"assert os.path.isfile(witness_path)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "b1c561a8",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"\n",
|
||||
"# HERE WE SETUP THE CIRCUIT PARAMS\n",
|
||||
"# WE GOT KEYS\n",
|
||||
"# WE GOT CIRCUIT PARAMETERS\n",
|
||||
"# EVERYTHING ANYONE HAS EVER NEEDED FOR ZK\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"res = ezkl.setup(\n",
|
||||
" compiled_model_path,\n",
|
||||
" vk_path,\n",
|
||||
" pk_path,\n",
|
||||
" \n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"assert os.path.isfile(vk_path)\n",
|
||||
"assert os.path.isfile(pk_path)\n",
|
||||
"assert os.path.isfile(settings_path)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "c384cbc8",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# GENERATE A PROOF\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"proof_path = os.path.join('test.pf')\n",
|
||||
"\n",
|
||||
"res = ezkl.prove(\n",
|
||||
" witness_path,\n",
|
||||
" compiled_model_path,\n",
|
||||
" pk_path,\n",
|
||||
" proof_path,\n",
|
||||
" \n",
|
||||
" \"single\",\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"print(res)\n",
|
||||
"assert os.path.isfile(proof_path)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "76f00d41",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# VERIFY IT\n",
|
||||
"\n",
|
||||
"res = ezkl.verify(\n",
|
||||
" proof_path,\n",
|
||||
" settings_path,\n",
|
||||
" vk_path,\n",
|
||||
" \n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"print(\"verified\")"
|
||||
]
|
||||
}
|
||||
],
|
||||
"metadata": {
|
||||
"kernelspec": {
|
||||
"display_name": "Python 3 (ipykernel)",
|
||||
"language": "python",
|
||||
"name": "python3"
|
||||
},
|
||||
"language_info": {
|
||||
"codemirror_mode": {
|
||||
"name": "ipython",
|
||||
"version": 3
|
||||
},
|
||||
"file_extension": ".py",
|
||||
"mimetype": "text/x-python",
|
||||
"name": "python",
|
||||
"nbconvert_exporter": "python",
|
||||
"pygments_lexer": "ipython3",
|
||||
"version": "3.9.15"
|
||||
}
|
||||
},
|
||||
"nbformat": 4,
|
||||
"nbformat_minor": 5
|
||||
}
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "cf69bb3f-94e6-4dba-92cd-ce08df117d67",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Linear Regression\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"Sklearn based models are slightly finicky to get into a suitable onnx format. \n",
|
||||
"This notebook showcases how to do so using the `hummingbird-ml` python package ! "
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "95613ee9",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# check if notebook is in colab\n",
|
||||
"try:\n",
|
||||
" # install ezkl\n",
|
||||
" import google.colab\n",
|
||||
" import subprocess\n",
|
||||
" import sys\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"ezkl\"])\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"onnx\"])\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"hummingbird-ml\"])\n",
|
||||
"\n",
|
||||
"# rely on local installation of ezkl if the notebook is not in colab\n",
|
||||
"except:\n",
|
||||
" pass\n",
|
||||
"\n",
|
||||
"import os\n",
|
||||
"import torch\n",
|
||||
"import ezkl\n",
|
||||
"import json\n",
|
||||
"from hummingbird.ml import convert\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# here we create and (potentially train a model)\n",
|
||||
"\n",
|
||||
"# make sure you have the dependencies required here already installed\n",
|
||||
"import numpy as np\n",
|
||||
"from sklearn.linear_model import LinearRegression\n",
|
||||
"X = np.array([[1, 1], [1, 2], [2, 2], [2, 3]])\n",
|
||||
"# y = 1 * x_0 + 2 * x_1 + 3\n",
|
||||
"y = np.dot(X, np.array([1, 2])) + 3\n",
|
||||
"reg = LinearRegression().fit(X, y)\n",
|
||||
"reg.score(X, y)\n",
|
||||
"\n",
|
||||
"circuit = convert(reg, \"torch\", X[:1]).model\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "b37637c4",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"model_path = os.path.join('network.onnx')\n",
|
||||
"compiled_model_path = os.path.join('network.compiled')\n",
|
||||
"pk_path = os.path.join('test.pk')\n",
|
||||
"vk_path = os.path.join('test.vk')\n",
|
||||
"settings_path = os.path.join('settings.json')\n",
|
||||
"\n",
|
||||
"witness_path = os.path.join('witness.json')\n",
|
||||
"data_path = os.path.join('input.json')"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "82db373a",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"\n",
|
||||
"\n",
|
||||
"# export to onnx format\n",
|
||||
"# !!!!!!!!!!!!!!!!! This will flash a warning but it is fine !!!!!!!!!!!!!!!!!!!!!\n",
|
||||
"\n",
|
||||
"# Input to the model\n",
|
||||
"shape = X.shape[1:]\n",
|
||||
"x = torch.rand(1, *shape, requires_grad=True)\n",
|
||||
"torch_out = circuit(x)\n",
|
||||
"# Export the model\n",
|
||||
"torch.onnx.export(circuit, # model being run\n",
|
||||
" # model input (or a tuple for multiple inputs)\n",
|
||||
" x,\n",
|
||||
" # where to save the model (can be a file or file-like object)\n",
|
||||
" \"network.onnx\",\n",
|
||||
" export_params=True, # store the trained parameter weights inside the model file\n",
|
||||
" opset_version=10, # the ONNX version to export the model to\n",
|
||||
" do_constant_folding=True, # whether to execute constant folding for optimization\n",
|
||||
" input_names=['input'], # the model's input names\n",
|
||||
" output_names=['output'], # the model's output names\n",
|
||||
" dynamic_axes={'input': {0: 'batch_size'}, # variable length axes\n",
|
||||
" 'output': {0: 'batch_size'}})\n",
|
||||
"\n",
|
||||
"d = ((x).detach().numpy()).reshape([-1]).tolist()\n",
|
||||
"\n",
|
||||
"data = dict(input_shapes=[shape],\n",
|
||||
" input_data=[d],\n",
|
||||
" output_data=[((o).detach().numpy()).reshape([-1]).tolist() for o in torch_out])\n",
|
||||
"\n",
|
||||
"# Serialize data into file:\n",
|
||||
"json.dump(data, open(\"input.json\", 'w'))\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# note that you can also call the following function to generate random data for the model\n",
|
||||
"# it is functionally equivalent to the code above\n",
|
||||
"ezkl.gen_random_data()\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "d5e374a2",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"!RUST_LOG=trace\n",
|
||||
"# TODO: Dictionary outputs\n",
|
||||
"res = ezkl.gen_settings(model_path, settings_path)\n",
|
||||
"assert res == True\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"cal_path = os.path.join(\"calibration.json\")\n",
|
||||
"\n",
|
||||
"data_array = (torch.randn(20, *shape).detach().numpy()).reshape([-1]).tolist()\n",
|
||||
"\n",
|
||||
"data = dict(input_data = [data_array])\n",
|
||||
"\n",
|
||||
"# Serialize data into file:\n",
|
||||
"json.dump(data, open(cal_path, 'w'))\n",
|
||||
"\n",
|
||||
"res = await ezkl.calibrate_settings(data_path, model_path, settings_path, \"resources\")\n",
|
||||
"assert res == True\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "3aa4f090",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"res = ezkl.compile_circuit(model_path, compiled_model_path, settings_path)\n",
|
||||
"assert res == True"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "8b74dcee",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# srs path\n",
|
||||
"res = await ezkl.get_srs( settings_path)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "18c8b7c7",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# now generate the witness file \n",
|
||||
"\n",
|
||||
"res = await ezkl.gen_witness(data_path, compiled_model_path, witness_path)\n",
|
||||
"assert os.path.isfile(witness_path)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "b1c561a8",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"\n",
|
||||
"# HERE WE SETUP THE CIRCUIT PARAMS\n",
|
||||
"# WE GOT KEYS\n",
|
||||
"# WE GOT CIRCUIT PARAMETERS\n",
|
||||
"# EVERYTHING ANYONE HAS EVER NEEDED FOR ZK\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"res = ezkl.setup(\n",
|
||||
" compiled_model_path,\n",
|
||||
" vk_path,\n",
|
||||
" pk_path,\n",
|
||||
" \n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"assert os.path.isfile(vk_path)\n",
|
||||
"assert os.path.isfile(pk_path)\n",
|
||||
"assert os.path.isfile(settings_path)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "c384cbc8",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# GENERATE A PROOF\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"proof_path = os.path.join('test.pf')\n",
|
||||
"\n",
|
||||
"res = ezkl.prove(\n",
|
||||
" witness_path,\n",
|
||||
" compiled_model_path,\n",
|
||||
" pk_path,\n",
|
||||
" proof_path,\n",
|
||||
" \n",
|
||||
" \"single\",\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"print(res)\n",
|
||||
"assert os.path.isfile(proof_path)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"id": "76f00d41",
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# VERIFY IT\n",
|
||||
"\n",
|
||||
"res = ezkl.verify(\n",
|
||||
" proof_path,\n",
|
||||
" settings_path,\n",
|
||||
" vk_path,\n",
|
||||
" \n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"print(\"verified\")"
|
||||
]
|
||||
}
|
||||
],
|
||||
"metadata": {
|
||||
"kernelspec": {
|
||||
"display_name": "Python 3 (ipykernel)",
|
||||
"language": "python",
|
||||
"name": "python3"
|
||||
},
|
||||
"language_info": {
|
||||
"codemirror_mode": {
|
||||
"name": "ipython",
|
||||
"version": 3
|
||||
},
|
||||
"file_extension": ".py",
|
||||
"mimetype": "text/x-python",
|
||||
"name": "python",
|
||||
"nbconvert_exporter": "python",
|
||||
"pygments_lexer": "ipython3",
|
||||
"version": "3.9.15"
|
||||
}
|
||||
},
|
||||
"nbformat": 4,
|
||||
"nbformat_minor": 5
|
||||
}
|
||||
|
||||
@@ -1,456 +1,459 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Mean of ERC20 transfer amounts\n",
|
||||
"\n",
|
||||
"This notebook shows how to calculate the mean of ERC20 transfer amounts, pulling data in from a Postgres database. First we install and get the necessary libraries running. \n",
|
||||
"The first of which is [shovel](https://indexsupply.com/shovel/docs/#getting-started), which is a library that allows us to pull data from the Ethereum blockchain into a Postgres database.\n",
|
||||
"\n",
|
||||
"Make sure you install postgres if needed https://indexsupply.com/shovel/docs/#getting-started. \n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import os\n",
|
||||
"import getpass\n",
|
||||
"import json\n",
|
||||
"import time\n",
|
||||
"import subprocess\n",
|
||||
"\n",
|
||||
"# swap out for the relevant linux/amd64, darwin/arm64, darwin/amd64, windows/amd64\n",
|
||||
"os.system(\"curl -LO https://indexsupply.net/bin/1.0/linux/amd64/shovel\")\n",
|
||||
"os.system(\"chmod +x shovel\")\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"os.environ[\"PG_URL\"] = \"postgres://\" + getpass.getuser() + \":@localhost:5432/shovel\"\n",
|
||||
"\n",
|
||||
"# create a config.json file with the following contents\n",
|
||||
"config = {\n",
|
||||
" \"pg_url\": \"$PG_URL\",\n",
|
||||
" \"eth_sources\": [\n",
|
||||
" {\"name\": \"mainnet\", \"chain_id\": 1, \"url\": \"https://ethereum-rpc.publicnode.com\"},\n",
|
||||
" {\"name\": \"base\", \"chain_id\": 8453, \"url\": \"https://base-rpc.publicnode.com\"}\n",
|
||||
" ],\n",
|
||||
" \"integrations\": [{\n",
|
||||
" \"name\": \"usdc_transfer\",\n",
|
||||
" \"enabled\": True,\n",
|
||||
" \"sources\": [{\"name\": \"mainnet\"}, {\"name\": \"base\"}],\n",
|
||||
" \"table\": {\n",
|
||||
" \"name\": \"usdc\",\n",
|
||||
" \"columns\": [\n",
|
||||
" {\"name\": \"log_addr\", \"type\": \"bytea\"},\n",
|
||||
" {\"name\": \"block_num\", \"type\": \"numeric\"},\n",
|
||||
" {\"name\": \"f\", \"type\": \"bytea\"},\n",
|
||||
" {\"name\": \"t\", \"type\": \"bytea\"},\n",
|
||||
" {\"name\": \"v\", \"type\": \"numeric\"}\n",
|
||||
" ]\n",
|
||||
" },\n",
|
||||
" \"block\": [\n",
|
||||
" {\"name\": \"block_num\", \"column\": \"block_num\"},\n",
|
||||
" {\n",
|
||||
" \"name\": \"log_addr\",\n",
|
||||
" \"column\": \"log_addr\",\n",
|
||||
" \"filter_op\": \"contains\",\n",
|
||||
" \"filter_arg\": [\n",
|
||||
" \"a0b86991c6218b36c1d19d4a2e9eb0ce3606eb48\",\n",
|
||||
" \"833589fCD6eDb6E08f4c7C32D4f71b54bdA02913\"\n",
|
||||
" ]\n",
|
||||
" }\n",
|
||||
" ],\n",
|
||||
" \"event\": {\n",
|
||||
" \"name\": \"Transfer\",\n",
|
||||
" \"type\": \"event\",\n",
|
||||
" \"anonymous\": False,\n",
|
||||
" \"inputs\": [\n",
|
||||
" {\"indexed\": True, \"name\": \"from\", \"type\": \"address\", \"column\": \"f\"},\n",
|
||||
" {\"indexed\": True, \"name\": \"to\", \"type\": \"address\", \"column\": \"t\"},\n",
|
||||
" {\"indexed\": False, \"name\": \"value\", \"type\": \"uint256\", \"column\": \"v\"}\n",
|
||||
" ]\n",
|
||||
" }\n",
|
||||
" }]\n",
|
||||
"}\n",
|
||||
"\n",
|
||||
"# write the config to a file\n",
|
||||
"with open(\"config.json\", \"w\") as f:\n",
|
||||
" f.write(json.dumps(config))\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# print the two env variables\n",
|
||||
"os.system(\"echo $PG_URL\")\n",
|
||||
"\n",
|
||||
"os.system(\"createdb -h localhost -p 5432 shovel\")\n",
|
||||
"\n",
|
||||
"os.system(\"echo shovel is now installed. starting:\")\n",
|
||||
"\n",
|
||||
"command = [\"./shovel\", \"-config\", \"config.json\"]\n",
|
||||
"proc = subprocess.Popen(command)\n",
|
||||
"\n",
|
||||
"os.system(\"echo shovel started.\")\n",
|
||||
"\n",
|
||||
"time.sleep(10)\n",
|
||||
"\n",
|
||||
"# after we've fetched some data -- kill the process\n",
|
||||
"proc.terminate()\n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "2wIAHwqH2_mo"
|
||||
},
|
||||
"source": [
|
||||
"**Import Dependencies**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "9Byiv2Nc2MsK"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# check if notebook is in colab\n",
|
||||
"try:\n",
|
||||
" # install ezkl\n",
|
||||
" import google.colab\n",
|
||||
" import subprocess\n",
|
||||
" import sys\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"ezkl\"])\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"onnx\"])\n",
|
||||
"\n",
|
||||
"# rely on local installation of ezkl if the notebook is not in colab\n",
|
||||
"except:\n",
|
||||
" pass\n",
|
||||
"\n",
|
||||
"import ezkl\n",
|
||||
"import torch\n",
|
||||
"import datetime\n",
|
||||
"import pandas as pd\n",
|
||||
"import requests\n",
|
||||
"import json\n",
|
||||
"import os\n",
|
||||
"\n",
|
||||
"import logging\n",
|
||||
"# # uncomment for more descriptive logging \n",
|
||||
"FORMAT = '%(levelname)s %(name)s %(asctime)-15s %(filename)s:%(lineno)d %(message)s'\n",
|
||||
"logging.basicConfig(format=FORMAT)\n",
|
||||
"logging.getLogger().setLevel(logging.DEBUG)\n",
|
||||
"\n",
|
||||
"print(\"ezkl version: \", ezkl.__version__)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "osjj-0Ta3E8O"
|
||||
},
|
||||
"source": [
|
||||
"**Create Computational Graph**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"base_uri": "https://localhost:8080/"
|
||||
},
|
||||
"id": "x1vl9ZXF3EEW",
|
||||
"outputId": "bda21d02-fe5f-4fb2-8106-f51a8e2e67aa"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from torch import nn\n",
|
||||
"import torch\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"class Model(nn.Module):\n",
|
||||
" def __init__(self):\n",
|
||||
" super(Model, self).__init__()\n",
|
||||
"\n",
|
||||
" # x is a time series \n",
|
||||
" def forward(self, x):\n",
|
||||
" return [torch.mean(x)]\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"circuit = Model()\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"x = 0.1*torch.rand(1,*[1,5], requires_grad=True)\n",
|
||||
"\n",
|
||||
"# # print(torch.__version__)\n",
|
||||
"device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
|
||||
"\n",
|
||||
"print(device)\n",
|
||||
"\n",
|
||||
"circuit.to(device)\n",
|
||||
"\n",
|
||||
"# Flips the neural net into inference mode\n",
|
||||
"circuit.eval()\n",
|
||||
"\n",
|
||||
"# Export the model\n",
|
||||
"torch.onnx.export(circuit, # model being run\n",
|
||||
" x, # model input (or a tuple for multiple inputs)\n",
|
||||
" \"lol.onnx\", # where to save the model (can be a file or file-like object)\n",
|
||||
" export_params=True, # store the trained parameter weights inside the model file\n",
|
||||
" opset_version=11, # the ONNX version to export the model to\n",
|
||||
" do_constant_folding=True, # whether to execute constant folding for optimization\n",
|
||||
" input_names = ['input'], # the model's input names\n",
|
||||
" output_names = ['output'], # the model's output names\n",
|
||||
" dynamic_axes={'input' : {0 : 'batch_size'}, # variable length axes\n",
|
||||
" 'output' : {0 : 'batch_size'}})\n",
|
||||
"\n",
|
||||
"# export(circuit, input_shape=[1, 20])\n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "E3qCeX-X5xqd"
|
||||
},
|
||||
"source": [
|
||||
"**Set Data Source and Get Data**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"base_uri": "https://localhost:8080/"
|
||||
},
|
||||
"id": "6RAMplxk5xPk",
|
||||
"outputId": "bd2158fe-0c00-44fd-e632-6a3f70cdb7c9"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import getpass\n",
|
||||
"# make an input.json file from the df above\n",
|
||||
"input_filename = os.path.join('input.json')\n",
|
||||
"\n",
|
||||
"pg_input_file = dict(input_data = {\n",
|
||||
" \"host\": \"localhost\",\n",
|
||||
" # make sure you replace this with your own username\n",
|
||||
" \"user\": getpass.getuser(),\n",
|
||||
" \"dbname\": \"shovel\",\n",
|
||||
" \"password\": \"\",\n",
|
||||
" \"query\": \"SELECT v FROM usdc ORDER BY block_num DESC LIMIT 5\",\n",
|
||||
" \"port\": \"5432\",\n",
|
||||
"})\n",
|
||||
"\n",
|
||||
"json_formatted_str = json.dumps(pg_input_file, indent=2)\n",
|
||||
"print(json_formatted_str)\n",
|
||||
"\n",
|
||||
"\n",
|
||||
" # Serialize data into file:\n",
|
||||
"json.dump(pg_input_file, open(input_filename, 'w' ))\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# this corresponds to 4 batches\n",
|
||||
"calibration_filename = os.path.join('calibration.json')\n",
|
||||
"\n",
|
||||
"pg_cal_file = dict(input_data = {\n",
|
||||
" \"host\": \"localhost\",\n",
|
||||
" # make sure you replace this with your own username\n",
|
||||
" \"user\": getpass.getuser(),\n",
|
||||
" \"dbname\": \"shovel\",\n",
|
||||
" \"password\": \"\",\n",
|
||||
" \"query\": \"SELECT v FROM usdc ORDER BY block_num DESC LIMIT 20\",\n",
|
||||
" \"port\": \"5432\",\n",
|
||||
"})\n",
|
||||
"\n",
|
||||
" # Serialize data into file:\n",
|
||||
"json.dump( pg_cal_file, open(calibration_filename, 'w' ))"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "eLJ7oirQ_HQR"
|
||||
},
|
||||
"source": [
|
||||
"**EZKL Workflow**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "rNw0C9QL6W88"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import subprocess\n",
|
||||
"import os\n",
|
||||
"\n",
|
||||
"onnx_filename = os.path.join('lol.onnx')\n",
|
||||
"compiled_filename = os.path.join('lol.compiled')\n",
|
||||
"settings_filename = os.path.join('settings.json')\n",
|
||||
"\n",
|
||||
"# Generate settings using ezkl\n",
|
||||
"res = ezkl.gen_settings(onnx_filename, settings_filename)\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"\n",
|
||||
"res = await ezkl.calibrate_settings(input_filename, onnx_filename, settings_filename, \"resources\")\n",
|
||||
"\n",
|
||||
"assert res == True"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"\n",
|
||||
"ezkl.compile_circuit(onnx_filename, compiled_filename, settings_filename)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"base_uri": "https://localhost:8080/"
|
||||
},
|
||||
"id": "4MmE9SX66_Il",
|
||||
"outputId": "16403639-66a4-4280-ac7f-6966b75de5a3"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# generate settings\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# show the settings.json\n",
|
||||
"with open(\"settings.json\") as f:\n",
|
||||
" data = json.load(f)\n",
|
||||
" json_formatted_str = json.dumps(data, indent=2)\n",
|
||||
"\n",
|
||||
" print(json_formatted_str)\n",
|
||||
"\n",
|
||||
"assert os.path.exists(\"settings.json\")\n",
|
||||
"assert os.path.exists(\"input.json\")\n",
|
||||
"assert os.path.exists(\"lol.onnx\")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "fULvvnK7_CMb"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"pk_path = os.path.join('test.pk')\n",
|
||||
"vk_path = os.path.join('test.vk')\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# setup the proof\n",
|
||||
"res = ezkl.setup(\n",
|
||||
" compiled_filename,\n",
|
||||
" vk_path,\n",
|
||||
" pk_path\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"assert os.path.isfile(vk_path)\n",
|
||||
"assert os.path.isfile(pk_path)\n",
|
||||
"assert os.path.isfile(settings_filename)\n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"witness_path = \"witness.json\"\n",
|
||||
"\n",
|
||||
"# generate the witness\n",
|
||||
"res = await ezkl.gen_witness(\n",
|
||||
" input_filename,\n",
|
||||
" compiled_filename,\n",
|
||||
" witness_path\n",
|
||||
" )\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"base_uri": "https://localhost:8080/"
|
||||
},
|
||||
"id": "Oog3j6Kd-Wed",
|
||||
"outputId": "5839d0c1-5b43-476e-c2f8-6707de562260"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# prove the zk circuit\n",
|
||||
"# GENERATE A PROOF\n",
|
||||
"proof_path = os.path.join('test.pf')\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"proof = ezkl.prove(\n",
|
||||
" witness_path,\n",
|
||||
" compiled_filename,\n",
|
||||
" pk_path,\n",
|
||||
" proof_path,\n",
|
||||
" \"single\"\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"print(\"proved\")\n",
|
||||
"\n",
|
||||
"assert os.path.isfile(proof_path)\n",
|
||||
"\n"
|
||||
]
|
||||
}
|
||||
],
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"provenance": []
|
||||
},
|
||||
"kernelspec": {
|
||||
"display_name": "Python 3",
|
||||
"name": "python3"
|
||||
},
|
||||
"language_info": {
|
||||
"codemirror_mode": {
|
||||
"name": "ipython",
|
||||
"version": 3
|
||||
},
|
||||
"file_extension": ".py",
|
||||
"mimetype": "text/x-python",
|
||||
"name": "python",
|
||||
"nbconvert_exporter": "python",
|
||||
"pygments_lexer": "ipython3",
|
||||
"version": "3.12.2"
|
||||
}
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Mean of ERC20 transfer amounts\n",
|
||||
"\n",
|
||||
"This notebook shows how to calculate the mean of ERC20 transfer amounts, pulling data in from a Postgres database. First we install and get the necessary libraries running. \n",
|
||||
"The first of which is [shovel](https://indexsupply.com/shovel/docs/#getting-started), which is a library that allows us to pull data from the Ethereum blockchain into a Postgres database.\n",
|
||||
"\n",
|
||||
"Make sure you install postgres if needed https://indexsupply.com/shovel/docs/#getting-started. \n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
"nbformat": 4,
|
||||
"nbformat_minor": 0
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import os\n",
|
||||
"import getpass\n",
|
||||
"import json\n",
|
||||
"import time\n",
|
||||
"import subprocess\n",
|
||||
"\n",
|
||||
"# swap out for the relevant linux/amd64, darwin/arm64, darwin/amd64, windows/amd64\n",
|
||||
"os.system(\"curl -LO https://indexsupply.net/bin/1.0/linux/amd64/shovel\")\n",
|
||||
"os.system(\"chmod +x shovel\")\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"os.environ[\"PG_URL\"] = \"postgres://\" + getpass.getuser() + \":@localhost:5432/shovel\"\n",
|
||||
"\n",
|
||||
"# create a config.json file with the following contents\n",
|
||||
"config = {\n",
|
||||
" \"pg_url\": \"$PG_URL\",\n",
|
||||
" \"eth_sources\": [\n",
|
||||
" {\"name\": \"mainnet\", \"chain_id\": 1, \"url\": \"https://ethereum-rpc.publicnode.com\"},\n",
|
||||
" {\"name\": \"base\", \"chain_id\": 8453, \"url\": \"https://base-rpc.publicnode.com\"}\n",
|
||||
" ],\n",
|
||||
" \"integrations\": [{\n",
|
||||
" \"name\": \"usdc_transfer\",\n",
|
||||
" \"enabled\": True,\n",
|
||||
" \"sources\": [{\"name\": \"mainnet\"}, {\"name\": \"base\"}],\n",
|
||||
" \"table\": {\n",
|
||||
" \"name\": \"usdc\",\n",
|
||||
" \"columns\": [\n",
|
||||
" {\"name\": \"log_addr\", \"type\": \"bytea\"},\n",
|
||||
" {\"name\": \"block_num\", \"type\": \"numeric\"},\n",
|
||||
" {\"name\": \"f\", \"type\": \"bytea\"},\n",
|
||||
" {\"name\": \"t\", \"type\": \"bytea\"},\n",
|
||||
" {\"name\": \"v\", \"type\": \"numeric\"}\n",
|
||||
" ]\n",
|
||||
" },\n",
|
||||
" \"block\": [\n",
|
||||
" {\"name\": \"block_num\", \"column\": \"block_num\"},\n",
|
||||
" {\n",
|
||||
" \"name\": \"log_addr\",\n",
|
||||
" \"column\": \"log_addr\",\n",
|
||||
" \"filter_op\": \"contains\",\n",
|
||||
" \"filter_arg\": [\n",
|
||||
" \"a0b86991c6218b36c1d19d4a2e9eb0ce3606eb48\",\n",
|
||||
" \"833589fCD6eDb6E08f4c7C32D4f71b54bdA02913\"\n",
|
||||
" ]\n",
|
||||
" }\n",
|
||||
" ],\n",
|
||||
" \"event\": {\n",
|
||||
" \"name\": \"Transfer\",\n",
|
||||
" \"type\": \"event\",\n",
|
||||
" \"anonymous\": False,\n",
|
||||
" \"inputs\": [\n",
|
||||
" {\"indexed\": True, \"name\": \"from\", \"type\": \"address\", \"column\": \"f\"},\n",
|
||||
" {\"indexed\": True, \"name\": \"to\", \"type\": \"address\", \"column\": \"t\"},\n",
|
||||
" {\"indexed\": False, \"name\": \"value\", \"type\": \"uint256\", \"column\": \"v\"}\n",
|
||||
" ]\n",
|
||||
" }\n",
|
||||
" }]\n",
|
||||
"}\n",
|
||||
"\n",
|
||||
"# write the config to a file\n",
|
||||
"with open(\"config.json\", \"w\") as f:\n",
|
||||
" f.write(json.dumps(config))\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# print the two env variables\n",
|
||||
"os.system(\"echo $PG_URL\")\n",
|
||||
"\n",
|
||||
"os.system(\"createdb -h localhost -p 5432 shovel\")\n",
|
||||
"\n",
|
||||
"os.system(\"echo shovel is now installed. starting:\")\n",
|
||||
"\n",
|
||||
"command = [\"./shovel\", \"-config\", \"config.json\"]\n",
|
||||
"proc = subprocess.Popen(command)\n",
|
||||
"\n",
|
||||
"os.system(\"echo shovel started.\")\n",
|
||||
"\n",
|
||||
"time.sleep(10)\n",
|
||||
"\n",
|
||||
"# after we've fetched some data -- kill the process\n",
|
||||
"proc.terminate()\n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "2wIAHwqH2_mo"
|
||||
},
|
||||
"source": [
|
||||
"**Import Dependencies**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "9Byiv2Nc2MsK"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# check if notebook is in colab\n",
|
||||
"try:\n",
|
||||
" # install ezkl\n",
|
||||
" import google.colab\n",
|
||||
" import subprocess\n",
|
||||
" import sys\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"ezkl\"])\n",
|
||||
" subprocess.check_call([sys.executable, \"-m\", \"pip\", \"install\", \"onnx\"])\n",
|
||||
"\n",
|
||||
"# rely on local installation of ezkl if the notebook is not in colab\n",
|
||||
"except:\n",
|
||||
" pass\n",
|
||||
"\n",
|
||||
"import ezkl\n",
|
||||
"import torch\n",
|
||||
"import datetime\n",
|
||||
"import pandas as pd\n",
|
||||
"import requests\n",
|
||||
"import json\n",
|
||||
"import os\n",
|
||||
"\n",
|
||||
"import logging\n",
|
||||
"# # uncomment for more descriptive logging \n",
|
||||
"FORMAT = '%(levelname)s %(name)s %(asctime)-15s %(filename)s:%(lineno)d %(message)s'\n",
|
||||
"logging.basicConfig(format=FORMAT)\n",
|
||||
"logging.getLogger().setLevel(logging.DEBUG)\n",
|
||||
"\n",
|
||||
"print(\"ezkl version: \", ezkl.__version__)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "osjj-0Ta3E8O"
|
||||
},
|
||||
"source": [
|
||||
"**Create Computational Graph**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"base_uri": "https://localhost:8080/"
|
||||
},
|
||||
"id": "x1vl9ZXF3EEW",
|
||||
"outputId": "bda21d02-fe5f-4fb2-8106-f51a8e2e67aa"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"from torch import nn\n",
|
||||
"import torch\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"class Model(nn.Module):\n",
|
||||
" def __init__(self):\n",
|
||||
" super(Model, self).__init__()\n",
|
||||
"\n",
|
||||
" # x is a time series \n",
|
||||
" def forward(self, x):\n",
|
||||
" return [torch.mean(x)]\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"circuit = Model()\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"x = 0.1*torch.rand(1,*[1,5], requires_grad=True)\n",
|
||||
"\n",
|
||||
"# # print(torch.__version__)\n",
|
||||
"device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
|
||||
"\n",
|
||||
"print(device)\n",
|
||||
"\n",
|
||||
"circuit.to(device)\n",
|
||||
"\n",
|
||||
"# Flips the neural net into inference mode\n",
|
||||
"circuit.eval()\n",
|
||||
"\n",
|
||||
"# Export the model\n",
|
||||
"torch.onnx.export(circuit, # model being run\n",
|
||||
" x, # model input (or a tuple for multiple inputs)\n",
|
||||
" \"lol.onnx\", # where to save the model (can be a file or file-like object)\n",
|
||||
" export_params=True, # store the trained parameter weights inside the model file\n",
|
||||
" opset_version=11, # the ONNX version to export the model to\n",
|
||||
" do_constant_folding=True, # whether to execute constant folding for optimization\n",
|
||||
" input_names = ['input'], # the model's input names\n",
|
||||
" output_names = ['output'], # the model's output names\n",
|
||||
" dynamic_axes={'input' : {0 : 'batch_size'}, # variable length axes\n",
|
||||
" 'output' : {0 : 'batch_size'}})\n",
|
||||
"\n",
|
||||
"# export(circuit, input_shape=[1, 20])\n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "E3qCeX-X5xqd"
|
||||
},
|
||||
"source": [
|
||||
"**Set Data Source and Get Data**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"base_uri": "https://localhost:8080/"
|
||||
},
|
||||
"id": "6RAMplxk5xPk",
|
||||
"outputId": "bd2158fe-0c00-44fd-e632-6a3f70cdb7c9"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import getpass\n",
|
||||
"# make an input.json file from the df above\n",
|
||||
"input_filename = os.path.join('input.json')\n",
|
||||
"\n",
|
||||
"pg_input_file = dict(input_data = {\n",
|
||||
" \"host\": \"localhost\",\n",
|
||||
" # make sure you replace this with your own username\n",
|
||||
" \"user\": getpass.getuser(),\n",
|
||||
" \"dbname\": \"shovel\",\n",
|
||||
" \"password\": \"\",\n",
|
||||
" \"query\": \"SELECT v FROM usdc ORDER BY block_num DESC LIMIT 5\",\n",
|
||||
" \"port\": \"5432\",\n",
|
||||
"})\n",
|
||||
"\n",
|
||||
"json_formatted_str = json.dumps(pg_input_file, indent=2)\n",
|
||||
"print(json_formatted_str)\n",
|
||||
"\n",
|
||||
"\n",
|
||||
" # Serialize data into file:\n",
|
||||
"json.dump(pg_input_file, open(input_filename, 'w' ))\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# this corresponds to 4 batches\n",
|
||||
"calibration_filename = os.path.join('calibration.json')\n",
|
||||
"\n",
|
||||
"pg_cal_file = dict(input_data = {\n",
|
||||
" \"host\": \"localhost\",\n",
|
||||
" # make sure you replace this with your own username\n",
|
||||
" \"user\": getpass.getuser(),\n",
|
||||
" \"dbname\": \"shovel\",\n",
|
||||
" \"password\": \"\",\n",
|
||||
" \"query\": \"SELECT v FROM usdc ORDER BY block_num DESC LIMIT 20\",\n",
|
||||
" \"port\": \"5432\",\n",
|
||||
"})\n",
|
||||
"\n",
|
||||
" # Serialize data into file:\n",
|
||||
"json.dump( pg_cal_file, open(calibration_filename, 'w' ))"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {
|
||||
"id": "eLJ7oirQ_HQR"
|
||||
},
|
||||
"source": [
|
||||
"**EZKL Workflow**"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "rNw0C9QL6W88"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"import subprocess\n",
|
||||
"import os\n",
|
||||
"\n",
|
||||
"onnx_filename = os.path.join('lol.onnx')\n",
|
||||
"compiled_filename = os.path.join('lol.compiled')\n",
|
||||
"settings_filename = os.path.join('settings.json')\n",
|
||||
"\n",
|
||||
"# Generate settings using ezkl\n",
|
||||
"res = ezkl.gen_settings(onnx_filename, settings_filename)\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"\n",
|
||||
"res = await ezkl.calibrate_settings(input_filename, onnx_filename, settings_filename, \"resources\")\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"\n",
|
||||
"await ezkl.get_srs(settings_filename)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"\n",
|
||||
"ezkl.compile_circuit(onnx_filename, compiled_filename, settings_filename)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"base_uri": "https://localhost:8080/"
|
||||
},
|
||||
"id": "4MmE9SX66_Il",
|
||||
"outputId": "16403639-66a4-4280-ac7f-6966b75de5a3"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# generate settings\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# show the settings.json\n",
|
||||
"with open(\"settings.json\") as f:\n",
|
||||
" data = json.load(f)\n",
|
||||
" json_formatted_str = json.dumps(data, indent=2)\n",
|
||||
"\n",
|
||||
" print(json_formatted_str)\n",
|
||||
"\n",
|
||||
"assert os.path.exists(\"settings.json\")\n",
|
||||
"assert os.path.exists(\"input.json\")\n",
|
||||
"assert os.path.exists(\"lol.onnx\")"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"id": "fULvvnK7_CMb"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"pk_path = os.path.join('test.pk')\n",
|
||||
"vk_path = os.path.join('test.vk')\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"# setup the proof\n",
|
||||
"res = ezkl.setup(\n",
|
||||
" compiled_filename,\n",
|
||||
" vk_path,\n",
|
||||
" pk_path\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"assert res == True\n",
|
||||
"assert os.path.isfile(vk_path)\n",
|
||||
"assert os.path.isfile(pk_path)\n",
|
||||
"assert os.path.isfile(settings_filename)\n",
|
||||
"\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"witness_path = \"witness.json\"\n",
|
||||
"\n",
|
||||
"# generate the witness\n",
|
||||
"res = await ezkl.gen_witness(\n",
|
||||
" input_filename,\n",
|
||||
" compiled_filename,\n",
|
||||
" witness_path\n",
|
||||
" )\n"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"base_uri": "https://localhost:8080/"
|
||||
},
|
||||
"id": "Oog3j6Kd-Wed",
|
||||
"outputId": "5839d0c1-5b43-476e-c2f8-6707de562260"
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# prove the zk circuit\n",
|
||||
"# GENERATE A PROOF\n",
|
||||
"proof_path = os.path.join('test.pf')\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"proof = ezkl.prove(\n",
|
||||
" witness_path,\n",
|
||||
" compiled_filename,\n",
|
||||
" pk_path,\n",
|
||||
" proof_path,\n",
|
||||
" \"single\"\n",
|
||||
" )\n",
|
||||
"\n",
|
||||
"\n",
|
||||
"print(\"proved\")\n",
|
||||
"\n",
|
||||
"assert os.path.isfile(proof_path)\n",
|
||||
"\n"
|
||||
]
|
||||
}
|
||||
],
|
||||
"metadata": {
|
||||
"colab": {
|
||||
"provenance": []
|
||||
},
|
||||
"kernelspec": {
|
||||
"display_name": ".env",
|
||||
"language": "python",
|
||||
"name": "python3"
|
||||
},
|
||||
"language_info": {
|
||||
"codemirror_mode": {
|
||||
"name": "ipython",
|
||||
"version": 3
|
||||
},
|
||||
"file_extension": ".py",
|
||||
"mimetype": "text/x-python",
|
||||
"name": "python",
|
||||
"nbconvert_exporter": "python",
|
||||
"pygments_lexer": "ipython3",
|
||||
"version": "3.12.7"
|
||||
}
|
||||
},
|
||||
"nbformat": 4,
|
||||
"nbformat_minor": 0
|
||||
}
|
||||
|
||||
@@ -12,6 +12,7 @@ asyncio_mode = "auto"
|
||||
|
||||
[project]
|
||||
name = "ezkl"
|
||||
version = "0.0.0"
|
||||
requires-python = ">=3.7"
|
||||
classifiers = [
|
||||
"Programming Language :: Rust",
|
||||
|
||||
@@ -938,6 +938,45 @@ fn gen_settings(
|
||||
Ok(true)
|
||||
}
|
||||
|
||||
/// Generates random data for the model
|
||||
///
|
||||
/// Arguments
|
||||
/// ---------
|
||||
/// model: str
|
||||
/// Path to the onnx file
|
||||
///
|
||||
/// output: str
|
||||
/// Path to create the data file
|
||||
///
|
||||
/// seed: int
|
||||
/// Random seed to use for generated data
|
||||
///
|
||||
/// variables
|
||||
/// Returns
|
||||
/// -------
|
||||
/// bool
|
||||
///
|
||||
#[pyfunction(signature = (
|
||||
model=PathBuf::from(DEFAULT_MODEL),
|
||||
output=PathBuf::from(DEFAULT_SETTINGS),
|
||||
variables=Vec::from([("batch_size".to_string(), 1)]),
|
||||
seed=DEFAULT_SEED.parse().unwrap(),
|
||||
))]
|
||||
#[gen_stub_pyfunction]
|
||||
fn gen_random_data(
|
||||
model: PathBuf,
|
||||
output: PathBuf,
|
||||
variables: Vec<(String, usize)>,
|
||||
seed: u64,
|
||||
) -> Result<bool, PyErr> {
|
||||
crate::execute::gen_random_data(model, output, variables, seed).map_err(|e| {
|
||||
let err_str = format!("Failed to generate settings: {}", e);
|
||||
PyRuntimeError::new_err(err_str)
|
||||
})?;
|
||||
|
||||
Ok(true)
|
||||
}
|
||||
|
||||
/// Calibrates the circuit settings
|
||||
///
|
||||
/// Arguments
|
||||
@@ -2055,6 +2094,7 @@ fn ezkl(m: &Bound<'_, PyModule>) -> PyResult<()> {
|
||||
m.add_function(wrap_pyfunction!(get_srs, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(gen_witness, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(gen_settings, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(gen_random_data, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(calibrate_settings, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(aggregate, m)?)?;
|
||||
m.add_function(wrap_pyfunction!(mock_aggregate, m)?)?;
|
||||
|
||||
@@ -30,6 +30,8 @@ use crate::{
|
||||
use super::*;
|
||||
use crate::circuit::ops::lookup::LookupOp;
|
||||
|
||||
const ASCII_ALPHABET: &str = "abcdefghijklmnopqrstuvwxyz";
|
||||
|
||||
/// Calculate the L1 distance between two tensors.
|
||||
/// ```
|
||||
/// use ezkl::tensor::Tensor;
|
||||
@@ -418,10 +420,6 @@ pub fn dot<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
values[0].remove_indices(&mut removal_indices, true)?;
|
||||
values[1].remove_indices(&mut removal_indices, true)?;
|
||||
|
||||
let elapsed = global_start.elapsed();
|
||||
trace!("filtering const zero indices took: {:?}", elapsed);
|
||||
|
||||
let start = instant::Instant::now();
|
||||
let mut inputs = vec![];
|
||||
let block_width = config.custom_gates.output.num_inner_cols();
|
||||
|
||||
@@ -429,37 +427,22 @@ pub fn dot<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
for (i, input) in values.iter_mut().enumerate() {
|
||||
input.pad_to_zero_rem(block_width, ValType::Constant(F::ZERO))?;
|
||||
let inp = {
|
||||
let (res, len) = region.assign_with_duplication(
|
||||
&config.custom_gates.inputs[i],
|
||||
input,
|
||||
&config.check_mode,
|
||||
false,
|
||||
)?;
|
||||
let (res, len) = region
|
||||
.assign_with_duplication_unconstrained(&config.custom_gates.inputs[i], input)?;
|
||||
assigned_len = len;
|
||||
res.get_inner()?
|
||||
};
|
||||
inputs.push(inp);
|
||||
}
|
||||
|
||||
let elapsed = start.elapsed();
|
||||
trace!("assigning inputs took: {:?}", elapsed);
|
||||
|
||||
// Now we can assign the dot product
|
||||
// time this step
|
||||
let start = instant::Instant::now();
|
||||
let accumulated_dot = accumulated::dot(&[inputs[0].clone(), inputs[1].clone()], block_width)?;
|
||||
let elapsed = start.elapsed();
|
||||
trace!("calculating accumulated dot took: {:?}", elapsed);
|
||||
|
||||
let start = instant::Instant::now();
|
||||
let (output, output_assigned_len) = region.assign_with_duplication(
|
||||
let (output, output_assigned_len) = region.assign_with_duplication_constrained(
|
||||
&config.custom_gates.output,
|
||||
&accumulated_dot.into(),
|
||||
&config.check_mode,
|
||||
true,
|
||||
)?;
|
||||
let elapsed = start.elapsed();
|
||||
trace!("assigning output took: {:?}", elapsed);
|
||||
|
||||
// enable the selectors
|
||||
if !region.is_dummy() {
|
||||
@@ -1000,7 +983,6 @@ fn select<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
region: &mut RegionCtx<F>,
|
||||
values: &[ValTensor<F>; 2],
|
||||
) -> Result<ValTensor<F>, CircuitError> {
|
||||
let start = instant::Instant::now();
|
||||
let (mut input, index) = (values[0].clone(), values[1].clone());
|
||||
input.flatten();
|
||||
|
||||
@@ -1028,9 +1010,6 @@ fn select<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
let (_, assigned_output) =
|
||||
dynamic_lookup(config, region, &[index, output], &[dim_indices, input])?;
|
||||
|
||||
let end = start.elapsed();
|
||||
trace!("select took: {:?}", end);
|
||||
|
||||
Ok(assigned_output)
|
||||
}
|
||||
|
||||
@@ -1092,7 +1071,6 @@ pub(crate) fn dynamic_lookup<F: PrimeField + TensorType + PartialOrd + std::hash
|
||||
lookups: &[ValTensor<F>; 2],
|
||||
tables: &[ValTensor<F>; 2],
|
||||
) -> Result<(ValTensor<F>, ValTensor<F>), CircuitError> {
|
||||
let start = instant::Instant::now();
|
||||
// if not all lookups same length err
|
||||
if lookups[0].len() != lookups[1].len() {
|
||||
return Err(CircuitError::MismatchedLookupLength(
|
||||
@@ -1126,28 +1104,20 @@ pub(crate) fn dynamic_lookup<F: PrimeField + TensorType + PartialOrd + std::hash
|
||||
}
|
||||
let table_len = table_0.len();
|
||||
|
||||
trace!("assigning tables took: {:?}", start.elapsed());
|
||||
|
||||
// now create a vartensor of constants for the dynamic lookup index
|
||||
let table_index = create_constant_tensor(F::from(dynamic_lookup_index as u64), table_len);
|
||||
let _table_index =
|
||||
region.assign_dynamic_lookup(&config.dynamic_lookups.tables[2], &table_index)?;
|
||||
|
||||
trace!("assigning table index took: {:?}", start.elapsed());
|
||||
|
||||
let lookup_0 = region.assign(&config.dynamic_lookups.inputs[0], &lookup_0)?;
|
||||
let lookup_1 = region.assign(&config.dynamic_lookups.inputs[1], &lookup_1)?;
|
||||
let lookup_len = lookup_0.len();
|
||||
|
||||
trace!("assigning lookups took: {:?}", start.elapsed());
|
||||
|
||||
// now set the lookup index
|
||||
let lookup_index = create_constant_tensor(F::from(dynamic_lookup_index as u64), lookup_len);
|
||||
|
||||
let _lookup_index = region.assign(&config.dynamic_lookups.inputs[2], &lookup_index)?;
|
||||
|
||||
trace!("assigning lookup index took: {:?}", start.elapsed());
|
||||
|
||||
let mut lookup_block = 0;
|
||||
|
||||
if !region.is_dummy() {
|
||||
@@ -1194,9 +1164,6 @@ pub(crate) fn dynamic_lookup<F: PrimeField + TensorType + PartialOrd + std::hash
|
||||
region.increment_dynamic_lookup_index(1);
|
||||
region.increment(lookup_len);
|
||||
|
||||
let end = start.elapsed();
|
||||
trace!("dynamic lookup took: {:?}", end);
|
||||
|
||||
Ok((lookup_0, lookup_1))
|
||||
}
|
||||
|
||||
@@ -1441,7 +1408,6 @@ pub(crate) fn linearize_element_index<F: PrimeField + TensorType + PartialOrd +
|
||||
dim: usize,
|
||||
is_flat_index: bool,
|
||||
) -> Result<ValTensor<F>, CircuitError> {
|
||||
let start_time = instant::Instant::now();
|
||||
let index = values[0].clone();
|
||||
if !is_flat_index {
|
||||
assert_eq!(index.dims().len(), dims.len());
|
||||
@@ -1515,9 +1481,6 @@ pub(crate) fn linearize_element_index<F: PrimeField + TensorType + PartialOrd +
|
||||
|
||||
region.apply_in_loop(&mut output, inner_loop_function)?;
|
||||
|
||||
let elapsed = start_time.elapsed();
|
||||
trace!("linearize_element_index took: {:?}", elapsed);
|
||||
|
||||
Ok(output.into())
|
||||
}
|
||||
|
||||
@@ -1949,16 +1912,11 @@ pub fn sum<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
|
||||
region.flush()?;
|
||||
// time this entire function run
|
||||
let global_start = instant::Instant::now();
|
||||
|
||||
let mut values = values.clone();
|
||||
|
||||
// this section has been optimized to death, don't mess with it
|
||||
values[0].remove_const_zero_values();
|
||||
|
||||
let elapsed = global_start.elapsed();
|
||||
trace!("filtering const zero indices took: {:?}", elapsed);
|
||||
|
||||
// if empty return a const
|
||||
if values[0].is_empty() {
|
||||
return Ok(create_zero_tensor(1));
|
||||
@@ -1970,12 +1928,8 @@ pub fn sum<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
let input = {
|
||||
let mut input = values[0].clone();
|
||||
input.pad_to_zero_rem(block_width, ValType::Constant(F::ZERO))?;
|
||||
let (res, len) = region.assign_with_duplication(
|
||||
&config.custom_gates.inputs[1],
|
||||
&input,
|
||||
&config.check_mode,
|
||||
false,
|
||||
)?;
|
||||
let (res, len) =
|
||||
region.assign_with_duplication_unconstrained(&config.custom_gates.inputs[1], &input)?;
|
||||
assigned_len = len;
|
||||
res.get_inner()?
|
||||
};
|
||||
@@ -1983,11 +1937,10 @@ pub fn sum<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
// Now we can assign the dot product
|
||||
let accumulated_sum = accumulated::sum(&input, block_width)?;
|
||||
|
||||
let (output, output_assigned_len) = region.assign_with_duplication(
|
||||
let (output, output_assigned_len) = region.assign_with_duplication_constrained(
|
||||
&config.custom_gates.output,
|
||||
&accumulated_sum.into(),
|
||||
&config.check_mode,
|
||||
true,
|
||||
)?;
|
||||
|
||||
// enable the selectors
|
||||
@@ -2053,13 +2006,10 @@ pub fn prod<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
) -> Result<ValTensor<F>, CircuitError> {
|
||||
region.flush()?;
|
||||
// time this entire function run
|
||||
let global_start = instant::Instant::now();
|
||||
|
||||
// this section has been optimized to death, don't mess with it
|
||||
let removal_indices = values[0].get_const_zero_indices();
|
||||
|
||||
let elapsed = global_start.elapsed();
|
||||
trace!("finding const zero indices took: {:?}", elapsed);
|
||||
// if empty return a const
|
||||
if !removal_indices.is_empty() {
|
||||
return Ok(create_zero_tensor(1));
|
||||
@@ -2070,12 +2020,8 @@ pub fn prod<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
let input = {
|
||||
let mut input = values[0].clone();
|
||||
input.pad_to_zero_rem(block_width, ValType::Constant(F::ONE))?;
|
||||
let (res, len) = region.assign_with_duplication(
|
||||
&config.custom_gates.inputs[1],
|
||||
&input,
|
||||
&config.check_mode,
|
||||
false,
|
||||
)?;
|
||||
let (res, len) =
|
||||
region.assign_with_duplication_unconstrained(&config.custom_gates.inputs[1], &input)?;
|
||||
assigned_len = len;
|
||||
res.get_inner()?
|
||||
};
|
||||
@@ -2083,11 +2029,10 @@ pub fn prod<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
// Now we can assign the dot product
|
||||
let accumulated_prod = accumulated::prod(&input, block_width)?;
|
||||
|
||||
let (output, output_assigned_len) = region.assign_with_duplication(
|
||||
let (output, output_assigned_len) = region.assign_with_duplication_constrained(
|
||||
&config.custom_gates.output,
|
||||
&accumulated_prod.into(),
|
||||
&config.check_mode,
|
||||
true,
|
||||
)?;
|
||||
|
||||
// enable the selectors
|
||||
@@ -2440,7 +2385,6 @@ pub(crate) fn pairwise<F: PrimeField + TensorType + PartialOrd + std::hash::Hash
|
||||
let orig_lhs = lhs.clone();
|
||||
let orig_rhs = rhs.clone();
|
||||
|
||||
let start = instant::Instant::now();
|
||||
let first_zero_indices = HashSet::from_iter(lhs.get_const_zero_indices());
|
||||
let second_zero_indices = HashSet::from_iter(rhs.get_const_zero_indices());
|
||||
|
||||
@@ -2455,7 +2399,6 @@ pub(crate) fn pairwise<F: PrimeField + TensorType + PartialOrd + std::hash::Hash
|
||||
BaseOp::Sub => second_zero_indices.clone(),
|
||||
_ => return Err(CircuitError::UnsupportedOp),
|
||||
};
|
||||
trace!("setting up indices took {:?}", start.elapsed());
|
||||
|
||||
if lhs.len() != rhs.len() {
|
||||
return Err(CircuitError::DimMismatch(format!(
|
||||
@@ -2480,7 +2423,6 @@ pub(crate) fn pairwise<F: PrimeField + TensorType + PartialOrd + std::hash::Hash
|
||||
|
||||
// Now we can assign the dot product
|
||||
// time the calc
|
||||
let start = instant::Instant::now();
|
||||
let op_result = match op {
|
||||
BaseOp::Add => add(&inputs),
|
||||
BaseOp::Sub => sub(&inputs),
|
||||
@@ -2491,20 +2433,13 @@ pub(crate) fn pairwise<F: PrimeField + TensorType + PartialOrd + std::hash::Hash
|
||||
error!("{}", e);
|
||||
halo2_proofs::plonk::Error::Synthesis
|
||||
})?;
|
||||
trace!("pairwise {} calc took {:?}", op.as_str(), start.elapsed());
|
||||
|
||||
let start = instant::Instant::now();
|
||||
let assigned_len = op_result.len() - removal_indices.len();
|
||||
let mut output = region.assign_with_omissions(
|
||||
&config.custom_gates.output,
|
||||
&op_result.into(),
|
||||
&removal_indices,
|
||||
)?;
|
||||
trace!(
|
||||
"pairwise {} input assign took {:?}",
|
||||
op.as_str(),
|
||||
start.elapsed()
|
||||
);
|
||||
|
||||
// Enable the selectors
|
||||
if !region.is_dummy() {
|
||||
@@ -2671,9 +2606,7 @@ pub fn greater<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
rhs.expand(&broadcasted_shape)?;
|
||||
|
||||
let diff = pairwise(config, region, &[lhs, rhs], BaseOp::Sub)?;
|
||||
|
||||
let sign = sign(config, region, &[diff])?;
|
||||
|
||||
equals(config, region, &[sign, create_unit_tensor(1)])
|
||||
}
|
||||
|
||||
@@ -5286,75 +5219,72 @@ pub(crate) fn decompose<F: PrimeField + TensorType + PartialOrd + std::hash::Has
|
||||
base: &usize,
|
||||
n: &usize,
|
||||
) -> Result<ValTensor<F>, CircuitError> {
|
||||
let input = values[0].clone();
|
||||
let mut input = values[0].clone();
|
||||
|
||||
let is_assigned = !input.all_prev_assigned();
|
||||
|
||||
let bases: ValTensor<F> = Tensor::from(
|
||||
(0..*n)
|
||||
.rev()
|
||||
.map(|x| ValType::Constant(integer_rep_to_felt(base.pow(x as u32) as IntegerRep))),
|
||||
if !is_assigned {
|
||||
input = region.assign(&config.custom_gates.inputs[0], &input)?;
|
||||
}
|
||||
|
||||
let mut bases: ValTensor<F> = Tensor::from(
|
||||
// repeat it input.len() times
|
||||
(0..input.len()).flat_map(|_| {
|
||||
(0..*n)
|
||||
.rev()
|
||||
.map(|x| ValType::Constant(integer_rep_to_felt(base.pow(x as u32) as IntegerRep)))
|
||||
}),
|
||||
)
|
||||
.into();
|
||||
let mut bases_dims = input.dims().to_vec();
|
||||
bases_dims.push(*n);
|
||||
bases.reshape(&bases_dims)?;
|
||||
|
||||
let cartesian_coord = input
|
||||
.dims()
|
||||
.iter()
|
||||
.map(|x| 0..*x)
|
||||
.multi_cartesian_product()
|
||||
.collect::<Vec<_>>();
|
||||
let mut decomposed_dims = input.dims().to_vec();
|
||||
decomposed_dims.push(*n + 1);
|
||||
|
||||
let mut output: Tensor<Tensor<ValType<F>>> = Tensor::new(None, input.dims())?;
|
||||
let claimed_output = if region.witness_gen() {
|
||||
input.decompose(*base, *n)?
|
||||
} else {
|
||||
let decomposed_len = decomposed_dims.iter().product();
|
||||
let claimed_output = Tensor::new(
|
||||
Some(&vec![ValType::Value(Value::unknown()); decomposed_len]),
|
||||
&decomposed_dims,
|
||||
)?;
|
||||
|
||||
let inner_loop_function =
|
||||
|i: usize, region: &mut RegionCtx<F>| -> Result<Tensor<ValType<F>>, CircuitError> {
|
||||
let coord = cartesian_coord[i].clone();
|
||||
let slice = coord.iter().map(|x| *x..*x + 1).collect::<Vec<_>>();
|
||||
let mut sliced_input = input.get_slice(&slice)?;
|
||||
sliced_input.flatten();
|
||||
claimed_output.into()
|
||||
};
|
||||
region.assign(&config.custom_gates.output, &claimed_output)?;
|
||||
region.increment(claimed_output.len());
|
||||
|
||||
if !is_assigned {
|
||||
sliced_input = region.assign(&config.custom_gates.inputs[0], &sliced_input)?;
|
||||
}
|
||||
let input_slice = input.dims().iter().map(|x| 0..*x).collect::<Vec<_>>();
|
||||
let mut sign_slice = input_slice.clone();
|
||||
sign_slice.push(0..1);
|
||||
let mut rest_slice = input_slice.clone();
|
||||
rest_slice.push(1..n + 1);
|
||||
|
||||
let mut claimed_output_slice = if region.witness_gen() {
|
||||
sliced_input.decompose(*base, *n)?
|
||||
} else {
|
||||
Tensor::from(vec![ValType::Value(Value::unknown()); *n + 1].into_iter()).into()
|
||||
};
|
||||
let sign = claimed_output.get_slice(&sign_slice)?;
|
||||
let rest = claimed_output.get_slice(&rest_slice)?;
|
||||
|
||||
claimed_output_slice =
|
||||
region.assign(&config.custom_gates.inputs[1], &claimed_output_slice)?;
|
||||
claimed_output_slice.flatten();
|
||||
let sign = range_check(config, region, &[sign], &(-1, 1))?;
|
||||
let rest = range_check(config, region, &[rest], &(0, (*base - 1) as i128))?;
|
||||
|
||||
region.increment(claimed_output_slice.len());
|
||||
// equation needs to be constructed as ij,ij->i but for arbitrary n dims we need to construct this dynamically
|
||||
// indices should map in order of the alphabet
|
||||
// start with lhs
|
||||
let lhs = ASCII_ALPHABET.chars().take(rest.dims().len()).join("");
|
||||
let rhs = ASCII_ALPHABET.chars().take(rest.dims().len() - 1).join("");
|
||||
let equation = format!("{},{}->{}", lhs, lhs, rhs);
|
||||
|
||||
// get the sign bit and make sure it is valid
|
||||
let sign = claimed_output_slice.first()?;
|
||||
let sign = range_check(config, region, &[sign], &(-1, 1))?;
|
||||
// now add the rhs
|
||||
|
||||
// get the rest of the thing and make sure it is in the correct range
|
||||
let rest = claimed_output_slice.get_slice(&[1..claimed_output_slice.len()])?;
|
||||
let prod_decomp = einsum(config, region, &[rest.clone(), bases], &equation)?;
|
||||
|
||||
let rest = range_check(config, region, &[rest], &(0, (base - 1) as i128))?;
|
||||
let signed_decomp = pairwise(config, region, &[prod_decomp, sign], BaseOp::Mult)?;
|
||||
|
||||
let prod_decomp = dot(config, region, &[rest, bases.clone()])?;
|
||||
enforce_equality(config, region, &[input, signed_decomp])?;
|
||||
|
||||
let signed_decomp = pairwise(config, region, &[prod_decomp, sign], BaseOp::Mult)?;
|
||||
|
||||
enforce_equality(config, region, &[sliced_input, signed_decomp])?;
|
||||
|
||||
Ok(claimed_output_slice.get_inner_tensor()?.clone())
|
||||
};
|
||||
|
||||
region.apply_in_loop(&mut output, inner_loop_function)?;
|
||||
|
||||
let mut combined_output = output.combine()?;
|
||||
let mut output_dims = input.dims().to_vec();
|
||||
output_dims.push(*n + 1);
|
||||
combined_output.reshape(&output_dims)?;
|
||||
|
||||
Ok(combined_output.into())
|
||||
Ok(claimed_output)
|
||||
}
|
||||
|
||||
pub(crate) fn sign<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
|
||||
@@ -671,22 +671,17 @@ impl<'a, F: PrimeField + TensorType + PartialOrd + std::hash::Hash> RegionCtx<'a
|
||||
}
|
||||
|
||||
/// Assign a valtensor to a vartensor with duplication
|
||||
pub fn assign_with_duplication(
|
||||
pub fn assign_with_duplication_unconstrained(
|
||||
&mut self,
|
||||
var: &VarTensor,
|
||||
values: &ValTensor<F>,
|
||||
check_mode: &crate::circuit::CheckMode,
|
||||
single_inner_col: bool,
|
||||
) -> Result<(ValTensor<F>, usize), Error> {
|
||||
if let Some(region) = &self.region {
|
||||
// duplicates every nth element to adjust for column overflow
|
||||
let (res, len) = var.assign_with_duplication(
|
||||
let (res, len) = var.assign_with_duplication_unconstrained(
|
||||
&mut region.borrow_mut(),
|
||||
self.row,
|
||||
self.linear_coord,
|
||||
values,
|
||||
check_mode,
|
||||
single_inner_col,
|
||||
&mut self.assigned_constants,
|
||||
)?;
|
||||
Ok((res, len))
|
||||
@@ -695,7 +690,37 @@ impl<'a, F: PrimeField + TensorType + PartialOrd + std::hash::Hash> RegionCtx<'a
|
||||
self.row,
|
||||
self.linear_coord,
|
||||
values,
|
||||
single_inner_col,
|
||||
false,
|
||||
&mut self.assigned_constants,
|
||||
)?;
|
||||
Ok((values.clone(), len))
|
||||
}
|
||||
}
|
||||
|
||||
/// Assign a valtensor to a vartensor with duplication
|
||||
pub fn assign_with_duplication_constrained(
|
||||
&mut self,
|
||||
var: &VarTensor,
|
||||
values: &ValTensor<F>,
|
||||
check_mode: &crate::circuit::CheckMode,
|
||||
) -> Result<(ValTensor<F>, usize), Error> {
|
||||
if let Some(region) = &self.region {
|
||||
// duplicates every nth element to adjust for column overflow
|
||||
let (res, len) = var.assign_with_duplication_constrained(
|
||||
&mut region.borrow_mut(),
|
||||
self.row,
|
||||
self.linear_coord,
|
||||
values,
|
||||
check_mode,
|
||||
&mut self.assigned_constants,
|
||||
)?;
|
||||
Ok((res, len))
|
||||
} else {
|
||||
let (_, len) = var.dummy_assign_with_duplication(
|
||||
self.row,
|
||||
self.linear_coord,
|
||||
values,
|
||||
true,
|
||||
&mut self.assigned_constants,
|
||||
)?;
|
||||
Ok((values.clone(), len))
|
||||
|
||||
@@ -90,6 +90,8 @@ pub const DEFAULT_USE_REDUCED_SRS_FOR_VERIFICATION: &str = "false";
|
||||
pub const DEFAULT_ONLY_RANGE_CHECK_REBASE: &str = "false";
|
||||
/// Default commitment
|
||||
pub const DEFAULT_COMMITMENT: &str = "kzg";
|
||||
/// Default seed used to generate random data
|
||||
pub const DEFAULT_SEED: &str = "21242";
|
||||
|
||||
#[cfg(feature = "python-bindings")]
|
||||
/// Converts TranscriptType into a PyObject (Required for TranscriptType to be compatible with Python)
|
||||
@@ -422,7 +424,21 @@ pub enum Commands {
|
||||
#[clap(flatten)]
|
||||
args: RunArgs,
|
||||
},
|
||||
|
||||
/// Generate random data for a model
|
||||
GenRandomData {
|
||||
/// The path to the .onnx model file
|
||||
#[arg(short = 'M', long, default_value = DEFAULT_MODEL, value_hint = clap::ValueHint::FilePath)]
|
||||
model: Option<PathBuf>,
|
||||
/// The path to the .json data file
|
||||
#[arg(short = 'D', long, default_value = DEFAULT_DATA, value_hint = clap::ValueHint::FilePath)]
|
||||
data: Option<PathBuf>,
|
||||
/// Hand-written parser for graph variables, eg. batch_size=1
|
||||
#[cfg_attr(all(feature = "ezkl", not(target_arch = "wasm32")), arg(short = 'V', long, value_parser = crate::parse_key_val::<String, usize>, default_value = "batch_size->1", value_delimiter = ',', value_hint = clap::ValueHint::Other))]
|
||||
variables: Vec<(String, usize)>,
|
||||
/// random seed for reproducibility (optional)
|
||||
#[arg(long, value_hint = clap::ValueHint::Other, default_value = DEFAULT_SEED)]
|
||||
seed: u64,
|
||||
},
|
||||
/// Calibrates the proving scale, lookup bits and logrows from a circuit settings file.
|
||||
CalibrateSettings {
|
||||
/// The path to the .json calibration data file.
|
||||
|
||||
@@ -488,7 +488,6 @@ pub async fn deploy_da_verifier_via_solidity(
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
match call_to_account {
|
||||
Some(call) => {
|
||||
deploy_single_da_contract(
|
||||
|
||||
@@ -65,6 +65,8 @@ use std::str::FromStr;
|
||||
use std::time::Duration;
|
||||
use tabled::Tabled;
|
||||
use thiserror::Error;
|
||||
use tract_onnx::prelude::IntoTensor;
|
||||
use tract_onnx::prelude::Tensor as TractTensor;
|
||||
|
||||
use lazy_static::lazy_static;
|
||||
|
||||
@@ -134,6 +136,17 @@ pub async fn run(command: Commands) -> Result<String, EZKLError> {
|
||||
settings_path.unwrap_or(DEFAULT_SETTINGS.into()),
|
||||
args,
|
||||
),
|
||||
Commands::GenRandomData {
|
||||
model,
|
||||
data,
|
||||
variables,
|
||||
seed,
|
||||
} => gen_random_data(
|
||||
model.unwrap_or(DEFAULT_MODEL.into()),
|
||||
data.unwrap_or(DEFAULT_DATA.into()),
|
||||
variables,
|
||||
seed,
|
||||
),
|
||||
Commands::CalibrateSettings {
|
||||
model,
|
||||
settings_path,
|
||||
@@ -828,6 +841,71 @@ pub(crate) fn gen_circuit_settings(
|
||||
Ok(String::new())
|
||||
}
|
||||
|
||||
/// Generate a circuit settings file
|
||||
pub(crate) fn gen_random_data(
|
||||
model_path: PathBuf,
|
||||
data_path: PathBuf,
|
||||
variables: Vec<(String, usize)>,
|
||||
seed: u64,
|
||||
) -> Result<String, EZKLError> {
|
||||
let mut file = std::fs::File::open(&model_path).map_err(|e| {
|
||||
crate::graph::errors::GraphError::ReadWriteFileError(
|
||||
model_path.display().to_string(),
|
||||
e.to_string(),
|
||||
)
|
||||
})?;
|
||||
|
||||
let (tract_model, _symbol_values) = Model::load_onnx_using_tract(&mut file, &variables)?;
|
||||
|
||||
let input_facts = tract_model
|
||||
.input_outlets()
|
||||
.map_err(|e| EZKLError::from(e.to_string()))?
|
||||
.iter()
|
||||
.map(|&i| tract_model.outlet_fact(i))
|
||||
.collect::<tract_onnx::prelude::TractResult<Vec<_>>>()
|
||||
.map_err(|e| EZKLError::from(e.to_string()))?;
|
||||
|
||||
/// Generates a random tensor of a given size and type.
|
||||
fn random(
|
||||
sizes: &[usize],
|
||||
datum_type: tract_onnx::prelude::DatumType,
|
||||
seed: u64,
|
||||
) -> TractTensor {
|
||||
use rand::{Rng, SeedableRng};
|
||||
let mut rng = rand::rngs::StdRng::seed_from_u64(seed);
|
||||
|
||||
let mut tensor = TractTensor::zero::<f32>(sizes).unwrap();
|
||||
let slice = tensor.as_slice_mut::<f32>().unwrap();
|
||||
slice.iter_mut().for_each(|x| *x = rng.gen());
|
||||
tensor.cast_to_dt(datum_type).unwrap().into_owned()
|
||||
}
|
||||
|
||||
fn tensor_for_fact(fact: &tract_onnx::prelude::TypedFact, seed: u64) -> TractTensor {
|
||||
if let Some(value) = &fact.konst {
|
||||
return value.clone().into_tensor();
|
||||
}
|
||||
|
||||
random(
|
||||
fact.shape
|
||||
.as_concrete()
|
||||
.expect("Expected concrete shape, found: {fact:?}"),
|
||||
fact.datum_type,
|
||||
seed,
|
||||
)
|
||||
}
|
||||
|
||||
let generated = input_facts
|
||||
.iter()
|
||||
.map(|v| tensor_for_fact(v, seed))
|
||||
.collect_vec();
|
||||
|
||||
let data = GraphData::from_tract_data(&generated)?;
|
||||
|
||||
data.save(data_path)?;
|
||||
|
||||
Ok(String::new())
|
||||
}
|
||||
|
||||
// not for wasm targets
|
||||
pub(crate) fn init_spinner() -> ProgressBar {
|
||||
let pb = indicatif::ProgressBar::new_spinner();
|
||||
|
||||
@@ -557,6 +557,34 @@ impl GraphData {
|
||||
Ok(inputs)
|
||||
}
|
||||
|
||||
// not wasm
|
||||
#[cfg(all(feature = "ezkl", not(target_arch = "wasm32")))]
|
||||
/// Convert the tract data to tract data
|
||||
pub fn from_tract_data(tensors: &[TractTensor]) -> Result<Self, GraphError> {
|
||||
use tract_onnx::prelude::DatumType;
|
||||
|
||||
let mut input_data = vec![];
|
||||
for tensor in tensors {
|
||||
match tensor.datum_type() {
|
||||
tract_onnx::prelude::DatumType::Bool => {
|
||||
let tensor = tensor.to_array_view::<bool>()?;
|
||||
let tensor = tensor.iter().map(|e| FileSourceInner::Bool(*e)).collect();
|
||||
input_data.push(tensor);
|
||||
}
|
||||
_ => {
|
||||
let cast_tensor = tensor.cast_to_dt(DatumType::F64)?;
|
||||
let tensor = cast_tensor.to_array_view::<f64>()?;
|
||||
let tensor = tensor.iter().map(|e| FileSourceInner::Float(*e)).collect();
|
||||
input_data.push(tensor);
|
||||
}
|
||||
}
|
||||
}
|
||||
Ok(GraphData {
|
||||
input_data: DataSource::File(input_data),
|
||||
output_data: None,
|
||||
})
|
||||
}
|
||||
|
||||
///
|
||||
pub fn new(input_data: DataSource) -> Self {
|
||||
GraphData {
|
||||
|
||||
@@ -280,7 +280,13 @@ impl GraphWitness {
|
||||
})?;
|
||||
|
||||
let reader = std::io::BufReader::with_capacity(*EZKL_BUF_CAPACITY, file);
|
||||
serde_json::from_reader(reader).map_err(|e| e.into())
|
||||
let witness: GraphWitness =
|
||||
serde_json::from_reader(reader).map_err(|e| Into::<GraphError>::into(e))?;
|
||||
|
||||
// check versions match
|
||||
crate::check_version_string_matches(witness.version.as_deref().unwrap_or(""));
|
||||
|
||||
Ok(witness)
|
||||
}
|
||||
|
||||
/// Save the model input to a file
|
||||
@@ -572,10 +578,14 @@ impl GraphSettings {
|
||||
// buf reader
|
||||
let reader =
|
||||
std::io::BufReader::with_capacity(*EZKL_BUF_CAPACITY, std::fs::File::open(path)?);
|
||||
serde_json::from_reader(reader).map_err(|e| {
|
||||
let settings: GraphSettings = serde_json::from_reader(reader).map_err(|e| {
|
||||
error!("failed to load settings file at {}", e);
|
||||
std::io::Error::new(std::io::ErrorKind::Other, e)
|
||||
})
|
||||
})?;
|
||||
|
||||
crate::check_version_string_matches(&settings.version);
|
||||
|
||||
Ok(settings)
|
||||
}
|
||||
|
||||
/// Export the ezkl configuration as json
|
||||
@@ -697,6 +707,9 @@ impl GraphCircuit {
|
||||
let reader = std::io::BufReader::with_capacity(*EZKL_BUF_CAPACITY, f);
|
||||
let result: GraphCircuit = bincode::deserialize_from(reader)?;
|
||||
|
||||
// check the versions matche
|
||||
crate::check_version_string_matches(&result.core.settings.version);
|
||||
|
||||
Ok(result)
|
||||
}
|
||||
}
|
||||
|
||||
@@ -621,16 +621,18 @@ impl Model {
|
||||
/// * `scale` - The scale to use for quantization.
|
||||
/// * `public_params` - Whether to make the params public.
|
||||
#[cfg(all(feature = "ezkl", not(target_arch = "wasm32")))]
|
||||
fn load_onnx_using_tract(
|
||||
pub(crate) fn load_onnx_using_tract(
|
||||
reader: &mut dyn std::io::Read,
|
||||
run_args: &RunArgs,
|
||||
variables: &[(String, usize)],
|
||||
) -> Result<TractResult, GraphError> {
|
||||
use tract_onnx::tract_hir::internal::GenericFactoid;
|
||||
|
||||
let mut model = tract_onnx::onnx().model_for_read(reader)?;
|
||||
|
||||
let variables: std::collections::HashMap<String, usize> =
|
||||
std::collections::HashMap::from_iter(run_args.variables.clone());
|
||||
std::collections::HashMap::from_iter(
|
||||
variables.into_iter().map(|(k, v)| (k.clone(), *v)),
|
||||
);
|
||||
|
||||
for (i, id) in model.clone().inputs.iter().enumerate() {
|
||||
let input = model.node_mut(id.node);
|
||||
@@ -655,7 +657,7 @@ impl Model {
|
||||
}
|
||||
|
||||
let mut symbol_values = SymbolValues::default();
|
||||
for (symbol, value) in run_args.variables.iter() {
|
||||
for (symbol, value) in variables.iter() {
|
||||
let symbol = model.symbols.sym(symbol);
|
||||
symbol_values = symbol_values.with(&symbol, *value as i64);
|
||||
debug!("set {} to {}", symbol, value);
|
||||
@@ -683,7 +685,7 @@ impl Model {
|
||||
) -> Result<ParsedNodes, GraphError> {
|
||||
let start_time = instant::Instant::now();
|
||||
|
||||
let (model, symbol_values) = Self::load_onnx_using_tract(reader, run_args)?;
|
||||
let (model, symbol_values) = Self::load_onnx_using_tract(reader, &run_args.variables)?;
|
||||
|
||||
let scales = VarScales::from_args(run_args);
|
||||
let nodes = Self::nodes_from_graph(
|
||||
@@ -964,7 +966,7 @@ impl Model {
|
||||
GraphError::ReadWriteFileError(model_path.display().to_string(), e.to_string())
|
||||
})?;
|
||||
|
||||
let (model, _) = Model::load_onnx_using_tract(&mut file, run_args)?;
|
||||
let (model, _) = Model::load_onnx_using_tract(&mut file, &run_args.variables)?;
|
||||
|
||||
let datum_types: Vec<DatumType> = model
|
||||
.input_outlets()?
|
||||
@@ -1226,6 +1228,7 @@ impl Model {
|
||||
values.iter().map(|v| v.dims()).collect_vec()
|
||||
);
|
||||
|
||||
let start = instant::Instant::now();
|
||||
match &node {
|
||||
NodeType::Node(n) => {
|
||||
let res = if node.is_constant() && node.num_uses() == 1 {
|
||||
@@ -1363,6 +1366,7 @@ impl Model {
|
||||
results.insert(*idx, full_results);
|
||||
}
|
||||
}
|
||||
debug!("------------ layout of {} took {:?}", idx, start.elapsed());
|
||||
}
|
||||
|
||||
// we do this so we can support multiple passes of the same model and have deterministic results (Non-assigned inputs etc... etc...)
|
||||
|
||||
@@ -142,8 +142,6 @@ use tract_onnx::prelude::SymbolValues;
|
||||
pub fn extract_tensor_value(
|
||||
input: Arc<tract_onnx::prelude::Tensor>,
|
||||
) -> Result<Tensor<f32>, GraphError> {
|
||||
use maybe_rayon::prelude::{IntoParallelRefIterator, ParallelIterator};
|
||||
|
||||
let dt = input.datum_type();
|
||||
let dims = input.shape().to_vec();
|
||||
|
||||
@@ -156,7 +154,7 @@ pub fn extract_tensor_value(
|
||||
match dt {
|
||||
DatumType::F16 => {
|
||||
let vec = input.as_slice::<tract_onnx::prelude::f16>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| (*x).into()).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| (*x).into()).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::F32 => {
|
||||
@@ -165,61 +163,61 @@ pub fn extract_tensor_value(
|
||||
}
|
||||
DatumType::F64 => {
|
||||
let vec = input.as_slice::<f64>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::I64 => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<i64>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::I32 => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<i32>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::I16 => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<i16>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::I8 => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<i8>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::U8 => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<u8>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::U16 => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<u16>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::U32 => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<u32>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::U64 => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<u64>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::Bool => {
|
||||
// Generally a shape or hyperparam
|
||||
let vec = input.as_slice::<bool>()?.to_vec();
|
||||
let cast: Vec<f32> = vec.par_iter().map(|x| *x as usize as f32).collect();
|
||||
let cast: Vec<f32> = vec.iter().map(|x| *x as usize as f32).collect();
|
||||
const_value = Tensor::<f32>::new(Some(&cast), &dims)?;
|
||||
}
|
||||
DatumType::TDim => {
|
||||
@@ -227,7 +225,7 @@ pub fn extract_tensor_value(
|
||||
let vec = input.as_slice::<tract_onnx::prelude::TDim>()?.to_vec();
|
||||
|
||||
let cast: Result<Vec<f32>, GraphError> = vec
|
||||
.par_iter()
|
||||
.iter()
|
||||
.map(|x| match x.to_i64() {
|
||||
Ok(v) => Ok(v as f32),
|
||||
Err(_) => match x.to_i64() {
|
||||
|
||||
27
src/lib.rs
27
src/lib.rs
@@ -420,3 +420,30 @@ where
|
||||
let b = s[pos + 2..].parse()?;
|
||||
Ok((a, b))
|
||||
}
|
||||
|
||||
/// Check if the version string matches the artifact version
|
||||
/// If the version string does not match the artifact version, log a warning
|
||||
pub fn check_version_string_matches(artifact_version: &str) {
|
||||
if artifact_version == "0.0.0"
|
||||
|| artifact_version == "source - no compatibility guaranteed"
|
||||
|| artifact_version.is_empty()
|
||||
{
|
||||
log::warn!("Artifact version is 0.0.0, skipping version check");
|
||||
return;
|
||||
}
|
||||
|
||||
let version = crate::version();
|
||||
|
||||
if version == "source - no compatibility guaranteed" {
|
||||
log::warn!("Compiled source version is not guaranteed to match artifact version");
|
||||
return;
|
||||
}
|
||||
|
||||
if version != artifact_version {
|
||||
log::warn!(
|
||||
"Version mismatch: CLI version is {} but artifact version is {}",
|
||||
version,
|
||||
artifact_version
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
@@ -822,6 +822,7 @@ where
|
||||
Scheme::Scalar: PrimeField + SerdeObject + FromUniformBytes<64>,
|
||||
{
|
||||
debug!("loading proving key from {:?}", path);
|
||||
let start = instant::Instant::now();
|
||||
let f = File::open(path.clone()).map_err(|e| PfsysError::LoadPk(format!("{}", e)))?;
|
||||
let mut reader = BufReader::with_capacity(*EZKL_BUF_CAPACITY, f);
|
||||
let pk = ProvingKey::<Scheme::Curve>::read::<_, C>(
|
||||
@@ -830,7 +831,8 @@ where
|
||||
params,
|
||||
)
|
||||
.map_err(|e| PfsysError::LoadPk(format!("{}", e)))?;
|
||||
info!("loaded proving key ✅");
|
||||
let elapsed = start.elapsed();
|
||||
info!("loaded proving key in {:?}", elapsed);
|
||||
Ok(pk)
|
||||
}
|
||||
|
||||
|
||||
@@ -638,42 +638,44 @@ impl<T: Clone + TensorType> Tensor<T> {
|
||||
where
|
||||
T: Send + Sync,
|
||||
{
|
||||
if indices.is_empty() {
|
||||
// Fast path: empty indices or full tensor slice
|
||||
if indices.is_empty()
|
||||
|| indices.iter().map(|x| x.end - x.start).collect::<Vec<_>>() == self.dims
|
||||
{
|
||||
return Ok(self.clone());
|
||||
}
|
||||
|
||||
// Validate dimensions
|
||||
if self.dims.len() < indices.len() {
|
||||
return Err(TensorError::DimError(format!(
|
||||
"The dimensionality of the slice {:?} is greater than the tensor's {:?}",
|
||||
indices, self.dims
|
||||
)));
|
||||
} else if indices.iter().map(|x| x.end - x.start).collect::<Vec<_>>() == self.dims {
|
||||
// else if slice is the same as dims, return self
|
||||
return Ok(self.clone());
|
||||
}
|
||||
|
||||
// if indices weren't specified we fill them in as required
|
||||
let mut full_indices = indices.to_vec();
|
||||
// Pre-allocate the full indices vector with capacity
|
||||
let mut full_indices = Vec::with_capacity(self.dims.len());
|
||||
full_indices.extend_from_slice(indices);
|
||||
|
||||
for i in 0..(self.dims.len() - indices.len()) {
|
||||
full_indices.push(0..self.dims()[indices.len() + i])
|
||||
}
|
||||
// Fill remaining dimensions
|
||||
full_indices.extend((indices.len()..self.dims.len()).map(|i| 0..self.dims[i]));
|
||||
|
||||
let cartesian_coord: Vec<Vec<usize>> = full_indices
|
||||
// Pre-calculate total size and allocate result vector
|
||||
let total_size: usize = full_indices
|
||||
.iter()
|
||||
.cloned()
|
||||
.multi_cartesian_product()
|
||||
.collect();
|
||||
|
||||
let res: Vec<T> = cartesian_coord
|
||||
.par_iter()
|
||||
.map(|e| {
|
||||
let index = self.get_index(e);
|
||||
self[index].clone()
|
||||
})
|
||||
.collect();
|
||||
.map(|range| range.end - range.start)
|
||||
.product();
|
||||
let mut res = Vec::with_capacity(total_size);
|
||||
|
||||
// Calculate new dimensions once
|
||||
let dims: Vec<usize> = full_indices.iter().map(|e| e.end - e.start).collect();
|
||||
|
||||
// Use iterator directly without collecting into intermediate Vec
|
||||
for coord in full_indices.iter().cloned().multi_cartesian_product() {
|
||||
let index = self.get_index(&coord);
|
||||
res.push(self[index].clone());
|
||||
}
|
||||
|
||||
Tensor::new(Some(&res), &dims)
|
||||
}
|
||||
|
||||
@@ -831,7 +833,7 @@ impl<T: Clone + TensorType> Tensor<T> {
|
||||
num_repeats: usize,
|
||||
initial_offset: usize,
|
||||
) -> Result<Tensor<T>, TensorError> {
|
||||
let mut inner: Vec<T> = vec![];
|
||||
let mut inner: Vec<T> = Vec::with_capacity(self.inner.len());
|
||||
let mut offset = initial_offset;
|
||||
for (i, elem) in self.inner.clone().into_iter().enumerate() {
|
||||
if (i + offset + 1) % n == 0 {
|
||||
@@ -860,20 +862,22 @@ impl<T: Clone + TensorType> Tensor<T> {
|
||||
num_repeats: usize,
|
||||
initial_offset: usize,
|
||||
) -> Result<Tensor<T>, TensorError> {
|
||||
let mut inner: Vec<T> = vec![];
|
||||
let mut indices_to_remove = std::collections::HashSet::new();
|
||||
for i in 0..self.inner.len() {
|
||||
if (i + initial_offset + 1) % n == 0 {
|
||||
for j in 1..(1 + num_repeats) {
|
||||
indices_to_remove.insert(i + j);
|
||||
}
|
||||
}
|
||||
}
|
||||
// Pre-calculate capacity to avoid reallocations
|
||||
let estimated_size = self.inner.len() - (self.inner.len() / n) * num_repeats;
|
||||
let mut inner = Vec::with_capacity(estimated_size);
|
||||
|
||||
let old_inner = self.inner.clone();
|
||||
for (i, elem) in old_inner.into_iter().enumerate() {
|
||||
if !indices_to_remove.contains(&i) {
|
||||
inner.push(elem.clone());
|
||||
// Use iterator directly instead of creating intermediate collections
|
||||
let mut i = 0;
|
||||
while i < self.inner.len() {
|
||||
// Add the current element
|
||||
inner.push(self.inner[i].clone());
|
||||
|
||||
// If this is an nth position (accounting for offset)
|
||||
if (i + initial_offset + 1) % n == 0 {
|
||||
// Skip the next num_repeats elements
|
||||
i += num_repeats + 1;
|
||||
} else {
|
||||
i += 1;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -494,16 +494,56 @@ impl VarTensor {
|
||||
}
|
||||
}
|
||||
|
||||
/// Assigns specific values (`ValTensor`) to the columns of the inner tensor but allows for column wrapping for accumulated operations.
|
||||
pub fn assign_with_duplication_unconstrained<
|
||||
F: PrimeField + TensorType + PartialOrd + std::hash::Hash,
|
||||
>(
|
||||
&self,
|
||||
region: &mut Region<F>,
|
||||
offset: usize,
|
||||
values: &ValTensor<F>,
|
||||
constants: &mut ConstantsMap<F>,
|
||||
) -> Result<(ValTensor<F>, usize), halo2_proofs::plonk::Error> {
|
||||
match values {
|
||||
ValTensor::Instance { .. } => unimplemented!("duplication is not supported on instance columns. increase K if you require more rows."),
|
||||
ValTensor::Value { inner: v, dims , ..} => {
|
||||
|
||||
let duplication_freq = self.block_size();
|
||||
|
||||
let num_repeats = self.num_inner_cols();
|
||||
|
||||
let duplication_offset = offset;
|
||||
|
||||
// duplicates every nth element to adjust for column overflow
|
||||
let v = v.duplicate_every_n(duplication_freq, num_repeats, duplication_offset).unwrap();
|
||||
let mut res: ValTensor<F> = {
|
||||
v.enum_map(|coord, k| {
|
||||
let cell = self.assign_value(region, offset, k.clone(), coord, constants)?;
|
||||
Ok::<_, halo2_proofs::plonk::Error>(cell)
|
||||
|
||||
})?.into()};
|
||||
let total_used_len = res.len();
|
||||
res.remove_every_n(duplication_freq, num_repeats, duplication_offset).unwrap();
|
||||
|
||||
res.reshape(dims).unwrap();
|
||||
res.set_scale(values.scale());
|
||||
|
||||
Ok((res, total_used_len))
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// Assigns specific values (`ValTensor`) to the columns of the inner tensor but allows for column wrapping for accumulated operations.
|
||||
/// Duplication occurs by copying the last cell of the column to the first cell next column and creating a copy constraint between the two.
|
||||
pub fn assign_with_duplication<F: PrimeField + TensorType + PartialOrd + std::hash::Hash>(
|
||||
pub fn assign_with_duplication_constrained<
|
||||
F: PrimeField + TensorType + PartialOrd + std::hash::Hash,
|
||||
>(
|
||||
&self,
|
||||
region: &mut Region<F>,
|
||||
row: usize,
|
||||
offset: usize,
|
||||
values: &ValTensor<F>,
|
||||
check_mode: &CheckMode,
|
||||
single_inner_col: bool,
|
||||
constants: &mut ConstantsMap<F>,
|
||||
) -> Result<(ValTensor<F>, usize), halo2_proofs::plonk::Error> {
|
||||
let mut prev_cell = None;
|
||||
@@ -512,34 +552,16 @@ impl VarTensor {
|
||||
ValTensor::Instance { .. } => unimplemented!("duplication is not supported on instance columns. increase K if you require more rows."),
|
||||
ValTensor::Value { inner: v, dims , ..} => {
|
||||
|
||||
let duplication_freq = if single_inner_col {
|
||||
self.col_size()
|
||||
} else {
|
||||
self.block_size()
|
||||
};
|
||||
|
||||
let num_repeats = if single_inner_col {
|
||||
1
|
||||
} else {
|
||||
self.num_inner_cols()
|
||||
};
|
||||
|
||||
let duplication_offset = if single_inner_col {
|
||||
row
|
||||
} else {
|
||||
offset
|
||||
};
|
||||
let duplication_freq = self.col_size();
|
||||
let num_repeats = 1;
|
||||
let duplication_offset = row;
|
||||
|
||||
// duplicates every nth element to adjust for column overflow
|
||||
let v = v.duplicate_every_n(duplication_freq, num_repeats, duplication_offset).unwrap();
|
||||
let mut res: ValTensor<F> = {
|
||||
v.enum_map(|coord, k| {
|
||||
|
||||
let step = if !single_inner_col {
|
||||
1
|
||||
} else {
|
||||
self.num_inner_cols()
|
||||
};
|
||||
let step = self.num_inner_cols();
|
||||
|
||||
let (x, y, z) = self.cartesian_coord(offset + coord * step);
|
||||
if matches!(check_mode, CheckMode::SAFE) && coord > 0 && z == 0 && y == 0 {
|
||||
@@ -549,11 +571,13 @@ impl VarTensor {
|
||||
|
||||
let cell = self.assign_value(region, offset, k.clone(), coord * step, constants)?;
|
||||
|
||||
if single_inner_col {
|
||||
if z == 0 {
|
||||
let at_end_of_column = z == duplication_freq - 1;
|
||||
let at_beginning_of_column = z == 0;
|
||||
|
||||
if at_end_of_column {
|
||||
// if we are at the end of the column, we need to copy the cell to the next column
|
||||
prev_cell = Some(cell.clone());
|
||||
} else if coord > 0 && z == 0 && single_inner_col {
|
||||
} else if coord > 0 && at_beginning_of_column {
|
||||
if let Some(prev_cell) = prev_cell.as_ref() {
|
||||
let cell = cell.cell().ok_or({
|
||||
error!("Error getting cell: {:?}", (x,y));
|
||||
@@ -563,10 +587,10 @@ impl VarTensor {
|
||||
halo2_proofs::plonk::Error::Synthesis})?;
|
||||
region.constrain_equal(prev_cell,cell)?;
|
||||
} else {
|
||||
error!("Error copy-constraining previous value: {:?}", (x,y));
|
||||
error!("Previous cell was not set");
|
||||
return Err(halo2_proofs::plonk::Error::Synthesis);
|
||||
}
|
||||
}}
|
||||
}
|
||||
|
||||
Ok(cell)
|
||||
|
||||
@@ -577,20 +601,6 @@ impl VarTensor {
|
||||
res.reshape(dims).unwrap();
|
||||
res.set_scale(values.scale());
|
||||
|
||||
if matches!(check_mode, CheckMode::SAFE) {
|
||||
// during key generation this will be 0 so we use this as a flag to check
|
||||
// TODO: this isn't very safe and would be better to get the phase directly
|
||||
let res_evals = res.int_evals().unwrap();
|
||||
let is_assigned = res_evals
|
||||
.iter()
|
||||
.all(|&x| x == 0);
|
||||
if !is_assigned {
|
||||
assert_eq!(
|
||||
values.int_evals().unwrap(),
|
||||
res_evals
|
||||
)};
|
||||
}
|
||||
|
||||
Ok((res, total_used_len))
|
||||
}
|
||||
}
|
||||
|
||||
Binary file not shown.
Binary file not shown.
Reference in New Issue
Block a user