vibehash 0.4.0__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- vibehash-0.4.0/PKG-INFO +367 -0
- vibehash-0.4.0/README.md +352 -0
- vibehash-0.4.0/pyproject.toml +28 -0
- vibehash-0.4.0/setup.cfg +4 -0
- vibehash-0.4.0/tests/test_consensus.py +329 -0
- vibehash-0.4.0/tests/test_core.py +547 -0
- vibehash-0.4.0/tests/test_gpu.py +133 -0
- vibehash-0.4.0/tests/test_hardening.py +745 -0
- vibehash-0.4.0/tests/test_minerui.py +78 -0
- vibehash-0.4.0/tests/test_network.py +380 -0
- vibehash-0.4.0/tests/test_setup.py +216 -0
- vibehash-0.4.0/tests/test_validators.py +229 -0
- vibehash-0.4.0/vibehash/__init__.py +1 -0
- vibehash-0.4.0/vibehash/__main__.py +3 -0
- vibehash-0.4.0/vibehash/blobs.py +157 -0
- vibehash-0.4.0/vibehash/cli.py +1256 -0
- vibehash-0.4.0/vibehash/client.py +476 -0
- vibehash-0.4.0/vibehash/compute.py +208 -0
- vibehash-0.4.0/vibehash/consensus.py +432 -0
- vibehash-0.4.0/vibehash/coordinator.py +791 -0
- vibehash-0.4.0/vibehash/crypto.py +248 -0
- vibehash-0.4.0/vibehash/data/bip39-english.txt +2048 -0
- vibehash-0.4.0/vibehash/data/fruit-new.csv +13 -0
- vibehash-0.4.0/vibehash/data/fruit.csv +321 -0
- vibehash-0.4.0/vibehash/economics.py +159 -0
- vibehash-0.4.0/vibehash/fees.py +71 -0
- vibehash-0.4.0/vibehash/gpusetup.py +222 -0
- vibehash-0.4.0/vibehash/hardware.py +52 -0
- vibehash-0.4.0/vibehash/ledger.py +551 -0
- vibehash-0.4.0/vibehash/minerui.py +400 -0
- vibehash-0.4.0/vibehash/mnemonic.py +137 -0
- vibehash-0.4.0/vibehash/model.py +181 -0
- vibehash-0.4.0/vibehash/node.py +764 -0
- vibehash-0.4.0/vibehash/p2p.py +69 -0
- vibehash-0.4.0/vibehash/protocol.py +139 -0
- vibehash-0.4.0/vibehash/ratelimit.py +295 -0
- vibehash-0.4.0/vibehash/receipts.py +68 -0
- vibehash-0.4.0/vibehash/replay.py +79 -0
- vibehash-0.4.0/vibehash/settings.py +63 -0
- vibehash-0.4.0/vibehash/storage.py +176 -0
- vibehash-0.4.0/vibehash/tabular.py +166 -0
- vibehash-0.4.0/vibehash/tasks.py +215 -0
- vibehash-0.4.0/vibehash/tls.py +194 -0
- vibehash-0.4.0/vibehash/verification.py +114 -0
- vibehash-0.4.0/vibehash/wallet.py +88 -0
- vibehash-0.4.0/vibehash/wire.py +115 -0
- vibehash-0.4.0/vibehash/worker.py +127 -0
- vibehash-0.4.0/vibehash.egg-info/PKG-INFO +367 -0
- vibehash-0.4.0/vibehash.egg-info/SOURCES.txt +51 -0
- vibehash-0.4.0/vibehash.egg-info/dependency_links.txt +1 -0
- vibehash-0.4.0/vibehash.egg-info/entry_points.txt +2 -0
- vibehash-0.4.0/vibehash.egg-info/requires.txt +2 -0
- vibehash-0.4.0/vibehash.egg-info/top_level.txt +1 -0
vibehash-0.4.0/PKG-INFO
ADDED
|
@@ -0,0 +1,367 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: vibehash
|
|
3
|
+
Version: 0.4.0
|
|
4
|
+
Summary: Vibe Hash: earn VIBE by training AI models on everyday computers (testnet)
|
|
5
|
+
Keywords: proof-of-useful-work,distributed-training,cryptocurrency,testnet
|
|
6
|
+
Classifier: Development Status :: 3 - Alpha
|
|
7
|
+
Classifier: Environment :: Console
|
|
8
|
+
Classifier: Operating System :: OS Independent
|
|
9
|
+
Classifier: Programming Language :: Python :: 3
|
|
10
|
+
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
|
|
11
|
+
Requires-Python: >=3.10
|
|
12
|
+
Description-Content-Type: text/markdown
|
|
13
|
+
Requires-Dist: numpy>=1.26
|
|
14
|
+
Requires-Dist: cryptography>=42
|
|
15
|
+
|
|
16
|
+
# Vibe Hash (VIBE)
|
|
17
|
+
|
|
18
|
+
A proof-of-useful-work cryptocurrency. Instead of hashing, "miners" train AI
|
|
19
|
+
models in small, verifiable chunks sized for mid- and low-end hardware. Every
|
|
20
|
+
verified chunk is paid pro rata. There's no lottery, so an old laptop earns the
|
|
21
|
+
same rate per FLOP as a gaming rig.
|
|
22
|
+
|
|
23
|
+
This is a **testnet implementation**: validator nodes that take turns leading
|
|
24
|
+
rounds and vote on every block (BFT consensus, so no single node is in charge),
|
|
25
|
+
real worker clients that talk to them over HTTP, wallets, and a persistent
|
|
26
|
+
chain that anyone can audit.
|
|
27
|
+
See [DESIGN.md](DESIGN.md) for the protocol, economics, threat model, and
|
|
28
|
+
what's still needed before a mainnet (§9).
|
|
29
|
+
|
|
30
|
+
## Install
|
|
31
|
+
|
|
32
|
+
Python 3.10+. Vibe Hash **isn't on PyPI yet**, so `pip install vibehash`
|
|
33
|
+
says "No matching distribution found" until a release is published (see
|
|
34
|
+
"Publishing a release" below). Until then, build a wheel with
|
|
35
|
+
`python -m build` in the project folder and install the file it writes to
|
|
36
|
+
`dist/`, which you can send to anyone. Run this in the folder where the file
|
|
37
|
+
is:
|
|
38
|
+
|
|
39
|
+
```bash
|
|
40
|
+
python -m pip install --upgrade vibehash-0.4.0-py3-none-any.whl
|
|
41
|
+
```
|
|
42
|
+
|
|
43
|
+
Or, to work on the source, install the project folder in editable mode so
|
|
44
|
+
your changes apply without reinstalling. Run this in the project folder:
|
|
45
|
+
|
|
46
|
+
```bash
|
|
47
|
+
python -m pip install -e .
|
|
48
|
+
```
|
|
49
|
+
|
|
50
|
+
Once it's published, anyone can install it with:
|
|
51
|
+
|
|
52
|
+
```bash
|
|
53
|
+
python -m pip install --upgrade vibehash
|
|
54
|
+
```
|
|
55
|
+
|
|
56
|
+
Each adds a `vibehash` command (also available as `python -m vibehash`) and
|
|
57
|
+
pulls in NumPy and `cryptography`.
|
|
58
|
+
|
|
59
|
+
## Quick start
|
|
60
|
+
|
|
61
|
+
Try a whole network on one computer: a node, two workers and a training job.
|
|
62
|
+
It needs no wallets, files or flags, prints what the workers earn each round,
|
|
63
|
+
and carries on where it stopped the next time you run it:
|
|
64
|
+
|
|
65
|
+
```bash
|
|
66
|
+
python -m vibehash testnet
|
|
67
|
+
```
|
|
68
|
+
|
|
69
|
+
To watch validators run it instead of one node, give it more validators. Each
|
|
70
|
+
gets its own port, they take turns leading rounds, and every block needs votes
|
|
71
|
+
from validators holding more than two thirds of the stake. Add `--reset` if
|
|
72
|
+
you've run a testnet before, since the number of validators is fixed when the
|
|
73
|
+
testnet is first made:
|
|
74
|
+
|
|
75
|
+
```bash
|
|
76
|
+
python -m vibehash testnet --validators 4
|
|
77
|
+
```
|
|
78
|
+
|
|
79
|
+
To join someone else's network as a worker, run the guided setup with their
|
|
80
|
+
node's address, then start working:
|
|
81
|
+
|
|
82
|
+
```bash
|
|
83
|
+
python -m vibehash setup --node 192.168.1.20
|
|
84
|
+
```
|
|
85
|
+
```bash
|
|
86
|
+
python -m vibehash worker
|
|
87
|
+
```
|
|
88
|
+
|
|
89
|
+
To train a model, save your examples as a CSV file (a header row, numeric
|
|
90
|
+
input columns, and the column to predict last), then:
|
|
91
|
+
|
|
92
|
+
```bash
|
|
93
|
+
python -m vibehash train my-data.csv --budget 100
|
|
94
|
+
```
|
|
95
|
+
```bash
|
|
96
|
+
python -m vibehash job status
|
|
97
|
+
```
|
|
98
|
+
```bash
|
|
99
|
+
python -m vibehash job download
|
|
100
|
+
```
|
|
101
|
+
```bash
|
|
102
|
+
python -m vibehash predict new-rows.csv
|
|
103
|
+
```
|
|
104
|
+
|
|
105
|
+
`train` sizes a model to the data, scales the inputs, holds back a fifth of
|
|
106
|
+
the rows to measure accuracy, and prints what it will train and what it
|
|
107
|
+
costs before posting. `job status` shows the accuracy climbing round by round,
|
|
108
|
+
`job download` saves the model as one file (`my-data-model.npz`), and
|
|
109
|
+
`predict` labels new rows (`--out predictions.csv` saves them all).
|
|
110
|
+
`python -m vibehash example` writes a sample, `fruit.csv`, to try first. On
|
|
111
|
+
your own testnet, pay with its requester wallet by adding
|
|
112
|
+
`--node http://127.0.0.1:8545 --wallet ~/.vibehash/testnet/requester.json` to
|
|
113
|
+
`train`. Jobs take turns a round at a time, so a new one starts at the next
|
|
114
|
+
round.
|
|
115
|
+
|
|
116
|
+
In a terminal, `worker` and `testnet` show a live mining dashboard: a drill
|
|
117
|
+
boring through rock while you train, the round and phase, your speed, units
|
|
118
|
+
mined, and VIBE earned as it lands. Add `--plain` for the classic log lines
|
|
119
|
+
(piped output and services get those automatically; `NO_COLOR` turns colours
|
|
120
|
+
off).
|
|
121
|
+
|
|
122
|
+
`setup` saves the node address, creates your wallet (encrypted with a password
|
|
123
|
+
you choose, in `~/.vibehash/wallet.json`), finds your GPU and offers to install
|
|
124
|
+
the PyTorch build for it, and offers to stake once you hold VIBE. Every
|
|
125
|
+
command after that uses the saved node and wallet, so `--node` and `--wallet`
|
|
126
|
+
are only needed to override them. Re-running `setup` keeps what's done.
|
|
127
|
+
|
|
128
|
+
The new wallet comes with **24 recovery words**, shown once. Write them down
|
|
129
|
+
and keep them offline: they bring the wallet back if the file or its password
|
|
130
|
+
is lost, and anyone who sees them can take its coins:
|
|
131
|
+
|
|
132
|
+
```bash
|
|
133
|
+
python -m vibehash wallet restore
|
|
134
|
+
```
|
|
135
|
+
|
|
136
|
+
`wallet restore` asks for the words (the first four letters of each are
|
|
137
|
+
enough) and a new password. Add `--passphrase` if you gave the words one, and
|
|
138
|
+
`--out FILE` to restore somewhere other than the default wallet.
|
|
139
|
+
|
|
140
|
+
**GPUs.** `vibehash install-gpu` does the GPU step on its own. It reads the
|
|
141
|
+
card and driver (NVIDIA through `nvidia-smi`, Intel Arc, Apple silicon, AMD
|
|
142
|
+
with ROCm on Linux), checks which PyTorch builds are currently published for
|
|
143
|
+
your driver and Python version, installs the newest one that fits, and
|
|
144
|
+
self-tests it. `vibehash devices` shows what's usable. Workers use the first
|
|
145
|
+
GPU they find; `--device cpu|cuda|xpu|mps` overrides it.
|
|
146
|
+
|
|
147
|
+
## Publishing a release
|
|
148
|
+
|
|
149
|
+
Releases publish themselves from GitHub. `.github/workflows/publish.yml` runs
|
|
150
|
+
when a version tag is pushed: it checks the tag matches `pyproject.toml`, runs
|
|
151
|
+
the fast tests, builds the wheel and source archive, and uploads them to PyPI
|
|
152
|
+
through a **trusted publisher**, so no API token is stored anywhere. PyPI's
|
|
153
|
+
trusted publisher for `vibehash` names owner `AlphaChad5000`, repository
|
|
154
|
+
`vibehash`, workflow `publish.yml` and environment `pypi`.
|
|
155
|
+
|
|
156
|
+
1. Bump `version` in `pyproject.toml` (PyPI never accepts the same version
|
|
157
|
+
twice), commit it, and push.
|
|
158
|
+
2. Tag the commit with that version and push the tag. This publishes it:
|
|
159
|
+
|
|
160
|
+
```bash
|
|
161
|
+
git tag v0.4.0
|
|
162
|
+
```
|
|
163
|
+
```bash
|
|
164
|
+
git push origin v0.4.0
|
|
165
|
+
```
|
|
166
|
+
|
|
167
|
+
Follow the run under the repository's Actions tab. To try a build locally
|
|
168
|
+
first, `python -m build` writes the packages to `dist/`.
|
|
169
|
+
|
|
170
|
+
## Run a network for other people
|
|
171
|
+
|
|
172
|
+
`vibehash testnet --host 0.0.0.0` is the quick way: other computers on your
|
|
173
|
+
network can join it with `setup --node YOUR-IP`, and it prints how to send
|
|
174
|
+
them VIBE to stake. Its wallets are unencrypted test keys, so for anything
|
|
175
|
+
longer-lived, set the network up by hand:
|
|
176
|
+
|
|
177
|
+
The commands below use bash syntax; on PowerShell set environment variables
|
|
178
|
+
with `$env:VIBEHASH_PASSWORD = "..."` instead.
|
|
179
|
+
|
|
180
|
+
1. Make wallets: one for the node operator, one per worker, one for whoever posts jobs. Each prints its 24 recovery words once.
|
|
181
|
+
|
|
182
|
+
```bash
|
|
183
|
+
python -m vibehash wallet create --out node.json
|
|
184
|
+
```
|
|
185
|
+
```bash
|
|
186
|
+
python -m vibehash wallet create --out alice.json
|
|
187
|
+
```
|
|
188
|
+
```bash
|
|
189
|
+
python -m vibehash wallet create --out requester.json
|
|
190
|
+
```
|
|
191
|
+
|
|
192
|
+
2. Write a genesis file that funds and stakes them (use the addresses printed above). `--validator` names the node as a validator, with the address other machines reach it at. `--economics mainnet` picks the long issuance schedule (3.125 VIBE a round, halving about every 4 years, 21M VIBE at most) instead of the fast testnet one. The genesis hash commits to it.
|
|
193
|
+
|
|
194
|
+
```bash
|
|
195
|
+
python -m vibehash genesis create --chain-id my-testnet-1 --out genesis.json --stake NODE_ADDR=10 --validator NODE_ADDR=http://192.168.1.20:8545 --alloc ALICE_ADDR=20 --alloc REQUESTER_ADDR=500
|
|
196
|
+
```
|
|
197
|
+
|
|
198
|
+
3. Start the node. Use `--host 0.0.0.0` to accept workers from other machines.
|
|
199
|
+
|
|
200
|
+
```bash
|
|
201
|
+
python -m vibehash node --genesis genesis.json --wallet node.json --data-dir node-data --host 0.0.0.0
|
|
202
|
+
```
|
|
203
|
+
|
|
204
|
+
For more validators, give each its own staked wallet and its own `--validator ADDR=URL` in the genesis, and start one node per wallet on its own machine, all with the same `genesis.json`. They find each other from the addresses in it. A staked account can also join a running network as a validator. Its node announces the address where other machines can reach it:
|
|
205
|
+
|
|
206
|
+
```bash
|
|
207
|
+
python -m vibehash validator join --url http://203.0.113.7:8545 --wallet bob.json
|
|
208
|
+
```
|
|
209
|
+
```bash
|
|
210
|
+
python -m vibehash node --genesis genesis.json --wallet bob.json --data-dir bob-data --host 0.0.0.0 --announce http://203.0.113.7:8545 --peer http://192.168.1.20:8545
|
|
211
|
+
```
|
|
212
|
+
|
|
213
|
+
A node whose wallet isn't a validator, started with `--peer`, follows the chain: it checks and serves every block, and points workers at whichever validator is leading. `vibehash validator list` shows the validators and their share of the votes.
|
|
214
|
+
|
|
215
|
+
4. Each worker runs setup with the node's address and their wallet, staking from the VIBE genesis gave them, then starts contributing.
|
|
216
|
+
|
|
217
|
+
```bash
|
|
218
|
+
python -m vibehash setup --node http://127.0.0.1:8545 --wallet alice.json --stake 10
|
|
219
|
+
```
|
|
220
|
+
```bash
|
|
221
|
+
python -m vibehash worker
|
|
222
|
+
```
|
|
223
|
+
|
|
224
|
+
5. Post a training job: from a CSV with `train`, or with full control over every setting from a job file and a prepared `.npz` dataset.
|
|
225
|
+
|
|
226
|
+
```bash
|
|
227
|
+
python -m vibehash train my-data.csv --node http://127.0.0.1:8545 --wallet requester.json --budget 150
|
|
228
|
+
```
|
|
229
|
+
```bash
|
|
230
|
+
python -m vibehash post-job --node http://127.0.0.1:8545 --wallet requester.json --spec examples/job.json --budget 150 --dataset examples/demo-dataset.npz
|
|
231
|
+
```
|
|
232
|
+
|
|
233
|
+
6. Check balances and independently audit the whole chain. `chain verify` re-checks every block from genesis, checks each block's commit certificate (the validators' signatures that made it final), and replays every round from the record its block commits to, so it catches a node that favoured a worker, invented a registration or forged a verdict, not just bad arithmetic. It also checks the signed receipts workers on this computer received for their messages (or any receipts file with `--receipts FILE`), which prove a node that accepted a message and then left it out of the round. For the most recent rounds (the node keeps their checkpoints for `--keep-rounds`, 10 by default) it downloads the checkpoints and re-runs every check that needs the model weights, so it catches a node that misreported a sanity check or sided with a false accusation.
|
|
234
|
+
|
|
235
|
+
```bash
|
|
236
|
+
python -m vibehash balance --node http://127.0.0.1:8545 --wallet alice.json
|
|
237
|
+
```
|
|
238
|
+
```bash
|
|
239
|
+
python -m vibehash chain verify --node http://127.0.0.1:8545 --genesis genesis.json
|
|
240
|
+
```
|
|
241
|
+
|
|
242
|
+
A dataset is an `.npz` with `x_train`, `y_train`, `x_val`, `y_val`. It's never
|
|
243
|
+
unpickled, and workers check its sha256 against the job before using it.
|
|
244
|
+
|
|
245
|
+
### Fees, HTTPS and rate limits
|
|
246
|
+
|
|
247
|
+
**Fees.** Every transaction pays a small fee to the validator that proposes
|
|
248
|
+
its block. A node relays transactions paying at least `--min-fee` (0.0001 VIBE
|
|
249
|
+
by default; `--min-fee 0` relays free ones). `send`, `stake`, `unstake`,
|
|
250
|
+
`post-job` and `cancel-job` pay the node's minimum unless you give `--fee`,
|
|
251
|
+
and `balance` shows it. The fee comes out of your spendable balance, so keep
|
|
252
|
+
a little unstaked VIBE to unstake with. Worker messages are free. When the
|
|
253
|
+
mempool is full, higher fees get in first.
|
|
254
|
+
|
|
255
|
+
**HTTPS.** A node can serve HTTPS itself. For a testnet, make a self-signed
|
|
256
|
+
certificate for the names and addresses workers will use, and start the node
|
|
257
|
+
with it:
|
|
258
|
+
|
|
259
|
+
```bash
|
|
260
|
+
python -m vibehash tls cert --host 192.168.1.20
|
|
261
|
+
```
|
|
262
|
+
```bash
|
|
263
|
+
python -m vibehash node --genesis genesis.json --wallet node.json --data-dir node-data --host 0.0.0.0 --tls-cert node-cert.pem --tls-key node-key.pem
|
|
264
|
+
```
|
|
265
|
+
|
|
266
|
+
Both print the certificate's fingerprint (`vibehash tls fingerprint
|
|
267
|
+
node-cert.pem` shows it again). Workers pin it in setup, and every later
|
|
268
|
+
command then trusts that exact certificate from that node and nothing else:
|
|
269
|
+
|
|
270
|
+
```bash
|
|
271
|
+
python -m vibehash setup --node 192.168.1.20 --tls-fingerprint FINGERPRINT
|
|
272
|
+
```
|
|
273
|
+
|
|
274
|
+
With a certificate from a real CA, plain `setup --node https://...` works.
|
|
275
|
+
`--tls-ca FILE` trusts a private CA instead, and `$VIBEHASH_TLS_FINGERPRINT`
|
|
276
|
+
or `$VIBEHASH_TLS_CA` set the trust without setup. Without the `cryptography`
|
|
277
|
+
package, `tls cert` prints the openssl command to run instead.
|
|
278
|
+
|
|
279
|
+
**Rate limits.** Each client address gets its own budget of reads, worker
|
|
280
|
+
messages, transactions and uploads, at most 32 connections (256 in all), and
|
|
281
|
+
socket timeouts that cut off slow-loris clients. Over budget, a node answers
|
|
282
|
+
429, and clients wait and retry. Clients on the node's own computer are
|
|
283
|
+
never limited, so a local testnet is never throttled. Behind a reverse proxy
|
|
284
|
+
every client looks local, so limit at the proxy or add `--rate-limit-local`.
|
|
285
|
+
`--rate-limit off` turns the budgets off, `--rate-limit tx=1/10,read=100`
|
|
286
|
+
changes some (requests a second and burst; the buckets are read, worker, tx,
|
|
287
|
+
upload and other), and `--max-connections N` sets the overall cap. DESIGN.md
|
|
288
|
+
§10 has the numbers.
|
|
289
|
+
|
|
290
|
+
## Simulate the economics
|
|
291
|
+
|
|
292
|
+
Runs the same round logic as a real node, with 14 simulated rigs (7 of them
|
|
293
|
+
cheating), in a few seconds:
|
|
294
|
+
|
|
295
|
+
```bash
|
|
296
|
+
python simulate.py --rounds 20
|
|
297
|
+
```
|
|
298
|
+
|
|
299
|
+
## Tests
|
|
300
|
+
|
|
301
|
+
```bash
|
|
302
|
+
python -m unittest discover -s tests -t .
|
|
303
|
+
```
|
|
304
|
+
|
|
305
|
+
`tests/test_network.py` starts a real node on localhost with four worker
|
|
306
|
+
clients (one cheating), uploads a dataset, runs three rounds, audits the chain
|
|
307
|
+
from the outside, and restarts the node from disk.
|
|
308
|
+
`tests/test_validators.py` runs four validators on localhost:
|
|
309
|
+
- they take turns leading rounds and outvote a leader that pays itself;
|
|
310
|
+
- they carry on without one that goes offline, and it catches up when it's back;
|
|
311
|
+
- a new node syncs the whole chain and refuses a block whose certificate is
|
|
312
|
+
short of two thirds of the stake.
|
|
313
|
+
|
|
314
|
+
`tests/test_consensus.py` puts the voting itself through crashed, lying and
|
|
315
|
+
double-voting validators, lost messages and a split network, on a simulated
|
|
316
|
+
clock. `tests/test_hardening.py` covers fees, issuance schedules, HTTPS
|
|
317
|
+
nodes, rate limits and recovery words, including the official BIP39 and
|
|
318
|
+
SLIP-0010 test vectors.
|
|
319
|
+
|
|
320
|
+
## Layout
|
|
321
|
+
|
|
322
|
+
| File | What it does |
|
|
323
|
+
|---|---|
|
|
324
|
+
| `vibehash/crypto.py` | Ed25519 (OpenSSL, or RFC 8032 pure-Python fallback), hashing, Merkle trees |
|
|
325
|
+
| `vibehash/model.py` | Reference workload: multi-layer MLP with deterministic, replayable SGD (float32 or float64); dataset loading |
|
|
326
|
+
| `vibehash/compute.py` | Compute backends: NumPy on CPU, PyTorch on NVIDIA / Intel / Apple / AMD GPUs; device detection |
|
|
327
|
+
| `vibehash/tasks.py` | Job specs (validated, stored on chain), work units, commitments, reveals, verdicts |
|
|
328
|
+
| `vibehash/verification.py` | Commit–reveal, Merkle openings, sanity checks, replay spot-checks |
|
|
329
|
+
| `vibehash/economics.py` | Issuance schedules (testnet, mainnet), reward split, reputation, slashing |
|
|
330
|
+
| `vibehash/ledger.py` | Transactions, blocks, state, unbonding; derives rewards/slashes from round summaries |
|
|
331
|
+
| `vibehash/fees.py` | Transaction fees: the relay minimum, the fee-ordered mempool, eviction |
|
|
332
|
+
| `vibehash/coordinator.py` | The round state machine, shared by the node and the simulator |
|
|
333
|
+
| `vibehash/worker.py` | Training, committing, revealing, verifying (plus cheat modes for tests) |
|
|
334
|
+
| `vibehash/node.py` | HTTP node: API, wall-clock round driver, leading rounds and committing blocks validators agree on |
|
|
335
|
+
| `vibehash/consensus.py` | Tendermint-style BFT voting among validators; commit certificates |
|
|
336
|
+
| `vibehash/p2p.py` | Peers: gossiping votes and transactions between nodes |
|
|
337
|
+
| `vibehash/replay.py` | Re-runs a block's round from its record: how validators check a proposal |
|
|
338
|
+
| `vibehash/receipts.py` | Signed receipts that prove a node dropped a message it accepted |
|
|
339
|
+
| `vibehash/ratelimit.py` | Per-client rate limits, connection caps and timeouts for the node's server |
|
|
340
|
+
| `vibehash/tls.py` | HTTPS: server contexts, self-signed certificates, certificate pinning for clients |
|
|
341
|
+
| `vibehash/client.py` | HTTP client and the worker loop, with live `WorkerStats` |
|
|
342
|
+
| `vibehash/minerui.py` | The terminal mining dashboard (plain ANSI, ASCII fallback) |
|
|
343
|
+
| `vibehash/tabular.py` | CSV files to training datasets, and downloaded models to predictions |
|
|
344
|
+
| `vibehash/data/` | The `fruit.csv` sample `vibehash example` writes, and the BIP39 word list |
|
|
345
|
+
| `vibehash/storage.py` | SQLite chain and weights, content-addressed datasets |
|
|
346
|
+
| `vibehash/blobs.py` | Content-addressed, chunked, resumable binary storage for checkpoints and datasets |
|
|
347
|
+
| `vibehash/wire.py` | JSON wire format and signed message envelopes |
|
|
348
|
+
| `vibehash/wallet.py` | Key files: scrypt + AES-256-GCM |
|
|
349
|
+
| `vibehash/mnemonic.py` | Recovery words: BIP39 mnemonics and SLIP-0010 Ed25519 key derivation |
|
|
350
|
+
| `vibehash/hardware.py` | RAM detection and a workload benchmark |
|
|
351
|
+
| `vibehash/cli.py` | The `vibehash` command, including `setup` and `testnet` |
|
|
352
|
+
| `vibehash/settings.py` | `~/.vibehash`: the default wallet, and the node address and its TLS trust that `setup` saves |
|
|
353
|
+
| `vibehash/gpusetup.py` | GPU detection before PyTorch is installed; picks and installs the PyTorch build for the card |
|
|
354
|
+
| `vibehash/protocol.py` | In-process driver for simulations |
|
|
355
|
+
|
|
356
|
+
## Status
|
|
357
|
+
|
|
358
|
+
Unaudited testnet software. Don't attach real value to it. DESIGN.md §9 lists
|
|
359
|
+
the remaining work:
|
|
360
|
+
- validators rechecking weight-based results before voting, with on-chain
|
|
361
|
+
disputes;
|
|
362
|
+
- slashing double-signing validators;
|
|
363
|
+
- replicating fraud evidence;
|
|
364
|
+
- model types beyond MLPs.
|
|
365
|
+
|
|
366
|
+
§10 covers what's been hardened: fees, the issuance schedule, TLS, rate limits
|
|
367
|
+
and recovery words.
|