vastai-sdk 0.3.3.dev4__tar.gz → 0.3.4.dev2__tar.gz

{vastai_sdk-0.3.3.dev4 → vastai_sdk-0.3.4.dev2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: vastai-sdk
-Version: 0.3.3.dev4
+Version: 0.3.4.dev2
 Summary: SDK for Vast.ai GPU Cloud Service
 License-File: LICENSE
 Author: Chris McKenzie

{vastai_sdk-0.3.3.dev4 → vastai_sdk-0.3.4.dev2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "vastai-sdk"
-version = "0.3.3-dev.4"
+version = "0.3.4-dev.2"
 description = "SDK for Vast.ai GPU Cloud Service"
 readme = "README.md"
 authors = [

{vastai_sdk-0.3.3.dev4 → vastai_sdk-0.3.4.dev2}/vastai/serverless/client/client.py

@@ -189,17 +189,6 @@ class Serverless:
                 raise RuntimeError(f"get_endpoint_workers failed: HTTP {resp.status} - {text}")
 
             data = await resp.json(content_type=None)
-
-            # If error message from authenticate_endpoint_apikey_by_id occurs, there is a possibility that
-            # the endpoint's worker instances are not ready to be queried. If an error message occurs,
-            # return an empty list and print the error message to the user. The endpoint get_endpoint_workers
-            # should normally return a list of dictionaries containing worker instance information.
-            if isinstance(data,dict):
-                if 'error_msg' in data.keys():
-                    self.logger.warning(f"Received the following error from get_endpoint_workers:{data['error_msg']}.\nEndpoint may not be ready for query. Check credentials or wait a few minutes and try again.")
-                    return []
-
-            
             if not isinstance(data, list):
                 raise RuntimeError(f"Unexpected response type (wanted list): {type(data)}")
 

{vastai_sdk-0.3.3.dev4 → vastai_sdk-0.3.4.dev2}/vastai/serverless/server/lib/backend.py

@@ -32,7 +32,7 @@ from .data_types import (
     BenchmarkResult
 )
 
-VERSION = "1.0.0"
+VERSION = "1.0.1"
 
 MSG_HISTORY_LEN = 100
 log = logging.getLogger(__file__)
@@ -344,16 +344,11 @@ class Backend:
                 return False
 
         message = {
-            key: value
-            for (key, value) in (dataclasses.asdict(auth_data).items())
-            if key != "signature" and key != "__request_id"
+            "url" : auth_data.url
         }
         if auth_data.reqnum < (self.reqnum - MSG_HISTORY_LEN):
             log.error(f"Signature error: reqnum failure, got {auth_data.reqnum}, current_reqnum: {self.reqnum}")
             return False
-        elif message in self.msg_history:
-            log.error(f"Signature error: message {message} already in message history")
-            return False
         elif verify_signature(json.dumps(message, indent=4, sort_keys=True), auth_data.signature):
             self.reqnum = max(auth_data.reqnum, self.reqnum)
             self.msg_history.append(message)

{vastai_sdk-0.3.3.dev4 → vastai_sdk-0.3.4.dev2}/vastai/serverless/server/lib/metrics.py

@@ -151,13 +151,16 @@ class Metrics:
     #######################################Private#######################################
 
     async def __send_delete_requests_and_reset(self):
-        async def post(report_addr: str, requests: list[dict]) -> bool:
+        async def post(report_addr: str, idxs: list[int], success_flag: bool) -> bool:
             data = {
                 "worker_id": self.id,
                 "mtoken": self.mtoken,
-                "requests": requests,
+                "request_idxs": idxs,
+                "success": success_flag,
             }
-            log.debug(f"Deleting requests: {[r['request_idx'] for r in requests]}")
+            log.debug(
+                f"Deleting requests that {'succeeded' if success_flag else 'failed'}: {data['request_idxs']}"
+            )
             full_path = report_addr.rstrip("/") + "/delete_requests/"
             for attempt in range(1, 4):
                 try:
@@ -174,12 +177,13 @@ class Metrics:
                 log.debug(f"Retrying delete_request, attempt: {attempt}")
             return False
 
-        requests_payload = [
-            {"request_idx": r.request_idx, "success": r.success, "status": r.status}
-            for r in self.model_metrics.requests_deleting
-        ]
+        # Take a snapshot of what we plan to send this tick.
+        # New arrivals after this snapshot will remain in the queue for the next tick.
+        snapshot = list(self.model_metrics.requests_deleting)
+        success_idxs = [r.request_idx for r in snapshot if r.success is True]
+        failed_idxs  = [r.request_idx for r in snapshot if r.success is False]
 
-        if not requests_payload:
+        if not success_idxs and not failed_idxs:
             return  # nothing to do
 
         for report_addr in self.report_addr:
@@ -187,10 +191,17 @@ class Metrics:
             if report_addr == "https://cloud.vast.ai/api/v0":
                 # Patch: ignore the Redis API report_addr
                 continue
+            sent_success = True
+            sent_failed  = True
+
+            if success_idxs:
+                sent_success = await post(report_addr, success_idxs, True)
+            if failed_idxs:
+                sent_failed = await post(report_addr, failed_idxs, False)
 
-            if await post(report_addr, requests_payload):
+            if sent_success and sent_failed:
                 # Remove only the items we actually sent from the live queue.
-                sent_set = {r["request_idx"] for r in requests_payload}
+                sent_set = set(success_idxs) | set(failed_idxs)
                 self.model_metrics.requests_deleting[:] = [
                     r for r in self.model_metrics.requests_deleting
                     if r.request_idx not in sent_set