Add support for both protocol versions on worker daemon

skynet/config.py

@@ -26,6 +26,7 @@ class DgpuConfig(msgspec.Struct):
     poll_time: float = 0.5  # wait time for polling updates from contract
     log_level: str = 'info'
     log_file: str = 'dgpu.log'  # log file path (only used when tui = true)
+    proto_version: int = 0
 
 
 class FrontendConfig(msgspec.Struct):

skynet/contract.py

@@ -5,10 +5,10 @@ from leap import CLEOS
 from leap.protocol import Name
 
 from skynet.types import (
-    ConfigV1,
+    Config, ConfigV0, ConfigV1,
-    AccountV1,
+    Account, AccountV0, AccountV1,
     WorkerV0,
-    RequestV1,
+    Request, RequestV0, RequestV1,
     BodyV0,
     WorkerStatusV0,
     ResultV0
@@ -33,37 +33,39 @@ class WorkerStatusNotFound(BaseException):
 
 class GPUContractAPI:
 
-    def __init__(self, cleos: CLEOS):
+    def __init__(self, cleos: CLEOS, proto_version: int = 0):
         self.receiver = 'gpu.scd'
         self._cleos = cleos
+        self.proto_version = proto_version
 
     # views into data
 
-    async def get_config(self) -> ConfigV1:
+    async def get_config(self) -> Config:
         rows = await self._cleos.aget_table(
             self.receiver, self.receiver, 'config',
-            resp_cls=ConfigV1
+            resp_cls=ConfigV1 if self.proto_version > 1 else ConfigV0
         )
         if len(rows) == 0:
             raise ConfigNotFound()
 
         return rows[0]
 
-    async def get_user(self, user: str) -> AccountV1:
+    async def get_user(self, user: str) -> Account:
         rows = await self._cleos.aget_table(
             self.receiver, self.receiver, 'users',
             key_type='name',
             lower_bound=user,
             upper_bound=user,
-            resp_cls=AccountV1
+            resp_cls=AccountV1 if self.proto_version > 1 else AccountV0
         )
         if len(rows) == 0:
             raise AccountNotFound(user)
 
         return rows[0]
 
-    async def get_users(self) -> list[AccountV1]:
+    async def get_users(self) -> list[Account]:
-        return await self._cleos.aget_table(self.receiver, self.receiver, 'users', resp_cls=AccountV1)
+        return await self._cleos.aget_table(
+            self.receiver, self.receiver, 'users', resp_cls=AccountV1 if self.proto_version > 0 else AccountV0)
 
     async def get_worker(self, worker: str) -> WorkerV0:
         rows = await self._cleos.aget_table(
@@ -78,31 +80,32 @@ class GPUContractAPI:
 
         return rows[0]
 
-    async def get_workers(self) -> list[AccountV1]:
+    async def get_workers(self) -> list[WorkerV0]:
         return await self._cleos.aget_table(self.receiver, self.receiver, 'workers', resp_cls=WorkerV0)
 
-    async def get_queue(self) -> RequestV1:
+    async def get_queue(self) -> Request:
-        return await self._cleos.aget_table(self.receiver, self.receiver, 'queue', resp_cls=RequestV1)
+        return await self._cleos.aget_table(
+            self.receiver, self.receiver, 'queue', resp_cls=RequestV1 if self.proto_version > 0 else RequestV0)
 
-    async def get_request(self, request_id: int) -> RequestV1:
+    async def get_request(self, request_id: int) -> Request:
         rows = await self._cleos.aget_table(
             self.receiver, self.receiver, 'queue',
             lower_bound=request_id,
             upper_bound=request_id,
-            resp_cls=RequestV1
+            resp_cls=RequestV1 if self.proto_version > 0  else RequestV0
         )
         if len(rows) == 0:
             raise RequestNotFound(request_id)
 
         return rows[0]
 
-    async def get_requests_since(self, seconds: int) -> list[RequestV1]:
+    async def get_requests_since(self, seconds: int) -> list[Request]:
         return await self._cleos.aget_table(
             self.receiver, self.receiver, 'queue',
             index_position=2,
             key_type='i64',
             lower_bound=int(time.time()) - seconds,
-            resp_cls=RequestV1
+            resp_cls=RequestV1 if self.proto_version > 0  else RequestV0
         )
 
     async def get_statuses_for_request(self, request_id: int) -> list[WorkerStatusV0]:
@@ -242,12 +245,17 @@ class GPUContractAPI:
         worker: str,
         request_id: int,
         result_hash: str,
-        ipfs_hash: str
+        ipfs_hash: str,
+        request_hash: str | None = None
     ):
+        args = [worker, request_id, result_hash, ipfs_hash]
+        if request_hash:
+            args.insert(2, request_hash)
+
         return await self._cleos.a_push_action(
             self.receiver,
             'submit',
-            [worker, request_id, result_hash, ipfs_hash],
+            args,
             worker,
             key=self._cleos.private_keys[worker]
         )

skynet/dgpu/daemon.py

@@ -1,4 +1,6 @@
+import os
 import logging
+import contextlib
 from functools import partial
 
 import trio
@@ -22,7 +24,7 @@ from skynet.dgpu.network import (
 async def maybe_update_tui_balance(contract: GPUContractAPI):
     async def _fn(tui):
         # update balance
-        balance = await contract.get_user(tui.config.account).balance
+        balance = (await contract.get_user(tui.config.account)).balance
         tui.set_header_text(new_balance=f'balance: {balance}')
 
     await maybe_update_tui_async(_fn)
@@ -126,16 +128,18 @@ async def maybe_serve_one(
                     used by torch each step of the inference, it will use a
                     trio.from_thread to unblock the main thread and pump the event loop
                     '''
-                    output_hash, output = await trio.to_thread.run_sync(
+                    with open(os.devnull, 'w') as devnull:
-                        partial(
+                        with contextlib.redirect_stdout(devnull):
-                            compute_one,
+                            output_hash, output = await trio.to_thread.run_sync(
-                            model,
+                                partial(
-                            req.id,
+                                    compute_one,
-                            mode, body.params,
+                                    model,
-                            inputs=inputs,
+                                    req.id,
-                            should_cancel=state_mngr.should_cancel_work,
+                                    mode, body.params,
-                        )
+                                    inputs=inputs,
-                    )
+                                    should_cancel=state_mngr.should_cancel_work,
+                                )
+                            )
 
                 case _:
                     raise DGPUComputeError(
@@ -146,7 +150,12 @@ async def maybe_serve_one(
 
             ipfs_hash = await ipfs_api.publish(output, type=output_type)
 
-            await contract.submit_work(config.account, req.id, output_hash, ipfs_hash)
+            maybe_request_hash = None
+            if config.proto_version == 0:
+                maybe_request_hash = req.hash_v0()
+
+            await contract.submit_work(
+                config.account, req.id, output_hash, ipfs_hash, request_hash=maybe_request_hash)
 
             await state_mngr.update_state()
 

skynet/dgpu/network.py

@@ -16,7 +16,7 @@ from skynet.config import load_skynet_toml
 from skynet.contract import GPUContractAPI
 from skynet.types import (
     BodyV0,
-    RequestV1,
+    Request,
     WorkerStatusV0,
     ResultV0
 )
@@ -63,7 +63,7 @@ class ContractState:
         self._config = load_skynet_toml().dgpu
         self._poll_index = 0
 
-        self._queue: list[RequestV1] = []
+        self._queue: list[Request] = []
         self._status_by_rid: dict[int, list[WorkerStatusV0]] = {}
         self._results: list[ResultV0] = []
 
@@ -139,7 +139,7 @@ class ContractState:
         return len(self._queue)
 
     @property
-    def first(self) -> RequestV1 | None:
+    def first(self) -> Request | None:
         if len(self._queue) > 0:
             return self._queue[0]
 

skynet/dgpu/tui.py

@@ -8,7 +8,7 @@ from skynet.config import DgpuConfig as Config
 
 
 class WorkerMonitor:
-    def __init__(self):
+    def __init__(self, config: Config):
         self.requests = []
         self.header_info = {}
 
@@ -63,6 +63,7 @@ class WorkerMonitor:
             event_loop=self.event_loop,
             unhandled_input=self._exit_on_q
         )
+        self.config = config
 
     def _create_listbox_body(self, requests):
         """
@@ -197,7 +198,7 @@ def init_tui(config: Config):
     global _tui
     assert not _tui
     setup_logging_for_tui(config)
-    _tui = WorkerMonitor()
+    _tui = WorkerMonitor(config)
     return _tui
 
 

skynet/types.py

@@ -1,4 +1,5 @@
 from enum import StrEnum
+from hashlib import sha256
 
 from msgspec import Struct
 
@@ -58,6 +59,8 @@ class ConfigV1(Struct):
     token_symbol: str
     global_nonce: int
 
+type Config = ConfigV0 | ConfigV1
+
 '''
 RequestV0
 
@@ -122,6 +125,16 @@ class RequestV0(Struct):
     binary_data: str
     timestamp: str
 
+    def hash_v0(self) -> str:
+        hash_str = (
+            str(self.nonce)
+            +
+            self.body
+            +
+            self.binary_data
+        )
+        return sha256(hash_str.encode('utf-8')).hexdigest()
+
 '''
 RequestV1
 
@@ -154,6 +167,8 @@ class RequestV1(Struct):
     binary_data: str
     timestamp: str
 
+type Request = RequestV0 | RequestV1
+
 
 '''
 AccountV0
@@ -200,6 +215,7 @@ class AccountV1(Struct):
     user: str
     balance: str
 
+type Account = AccountV0 | AccountV1
 
 '''
 WorkerV0