arbor-ai 0.2__py3-none-any.whl → 0.2.1__py3-none-any.whl

arbor/server/services/grpo_manager.py

@@ -270,7 +270,6 @@ class GRPOManager:
                     print("Updating inference model...")
                     # There is a case where this status is sent multiple times
                     # We need to make sure we only update the model once
-                    self.current_model = status["output_dir"]
                     self.saving_model = False
                     print("Model update complete")
                 elif status["status"] == "checkpoint_saved":
@@ -308,6 +307,9 @@ class GRPOManager:
             print(f"Failed to send batch to training process: {e}")
             raise
 
+        self.current_model = self.train_kwargs["output_dir"]
+        inference_manager.launched_model = self.current_model
+
         return {
             "current_model": self.current_model,
             "checkpoints": self.checkpoints,

arbor/server/services/inference/vllm_client.py

@@ -1,5 +1,6 @@
 # adapted from Will Brown's verifiers library (https://github.com/willccbb/verifiers)
 
+import asyncio
 import atexit
 import logging
 import time
@@ -239,7 +240,7 @@ class VLLMClient:
                     response.raise_for_status()
                     return response.json()
 
-            except httpx.TimeoutError:
+            except httpx.TimeoutException:
                 logger.error("Request timed out")
                 raise
             except InferenceBlockedError:

{arbor_ai-0.2.dist-info → arbor_ai-0.2.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: arbor-ai
-Version: 0.2
+Version: 0.2.1
 Summary: A framework for fine-tuning and managing language models
 Author-email: Noah Ziems <nziems2@nd.edu>
 Project-URL: Homepage, https://github.com/Ziems/arbor
@@ -79,6 +79,16 @@ Follow the DSPy tutorials here to see usage examples:
 
 ---
 
+### Troubleshooting
+
+**NCCL Errors**
+Certain GPU setups, particularly with newer GPUs, seem to have issues with NCCL that cause Arbor to crash. Often times of these can be fixed with the following environment variables:
+
+```bash
+export NCCL_P2P_DISABLE=1
+export NCCL_IB_DISABLE=1
+```
+
 ## 🙏 Acknowledgements
 
 Arbor builds on the shoulders of great work. We extend our thanks to:

{arbor_ai-0.2.dist-info → arbor_ai-0.2.1.dist-info}/RECORD

@@ -17,14 +17,14 @@ arbor/server/core/logging.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,
 arbor/server/services/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 arbor/server/services/dependencies.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 arbor/server/services/file_manager.py,sha256=Z9z4A4EzvPauid_DBfpim401DDtuJy_TbX4twTWDJWI,12119
-arbor/server/services/grpo_manager.py,sha256=MDpOGN99WnNg4q_8974MkAnqcClOXy6fYcD2sFvs2Ho,18487
+arbor/server/services/grpo_manager.py,sha256=jY4kc7wlKKoi7RigjJiH1VaxX6qJCOxyEc0oYCkqPlQ,18549
 arbor/server/services/inference_manager.py,sha256=a1c5zYbjk6fPM3egX2McKv7ZWPN7c-QH_Qogu9iay90,9597
 arbor/server/services/job_manager.py,sha256=m_d4UPwN_82f7t7K443DaFpFoyv7JZSZKml8tawt1Bk,2186
 arbor/server/services/training_manager.py,sha256=oQdhpfxdgp_lCTb_lxhvjupdLrcg6HL3TEbct_q9F6I,21065
 arbor/server/services/comms/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 arbor/server/services/comms/comms.py,sha256=3KN3mzwPvfW2_L5hq02JdAk6yOMyhY0_pBz-DDr5A3o,7694
 arbor/server/services/inference/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-arbor/server/services/inference/vllm_client.py,sha256=P4etwX47VVMEaVWUOT-aP6_OONf8ZzniwXndmJujNxY,18250
+arbor/server/services/inference/vllm_client.py,sha256=06-VfdcwKqq8_ZRWaER3OnSVLtvL87bLdljSrkXfm-A,18269
 arbor/server/services/inference/vllm_serve.py,sha256=UZAGo7CyshR3-9fhXCTKhXeidqNqbY6LyU9DDNiX_Sw,109543
 arbor/server/services/scripts/dpo_training.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 arbor/server/services/scripts/grpo_training.py,sha256=6kXzMwn3rZXHdEn0xe_Kd9d7tbdYb76zE0zbi02xCm4,31314
@@ -34,9 +34,9 @@ arbor/server/services/scripts/utils/arg_parser.py,sha256=ur_iyhc_Ie00tjq63vK4Sde
 arbor/server/services/scripts/utils/dataset.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 arbor/server/utils/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 arbor/server/utils/helpers.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-arbor_ai-0.2.dist-info/licenses/LICENSE,sha256=5vFGrbOFeXXM83JV9o16w7ohH4WLeu3-57GocJSz8ow,1067
-arbor_ai-0.2.dist-info/METADATA,sha256=LieUwdo2RQBgh5ukQJh-NHUA2_CBS1Dr9YqjSbgcEnM,2504
-arbor_ai-0.2.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-arbor_ai-0.2.dist-info/entry_points.txt,sha256=PGBX-MfNwfIl8UPFgsX3gjtXLqSogRhOktKMpZUysD0,40
-arbor_ai-0.2.dist-info/top_level.txt,sha256=jzWdp3BRYqvZDMFsPajrcftvvlluzVDErkD8IMRfhYs,6
-arbor_ai-0.2.dist-info/RECORD,,
+arbor_ai-0.2.1.dist-info/licenses/LICENSE,sha256=5vFGrbOFeXXM83JV9o16w7ohH4WLeu3-57GocJSz8ow,1067
+arbor_ai-0.2.1.dist-info/METADATA,sha256=34XAZBm8OLlsSBicLmRn_hhbltn0pDNlAj5WOjn9LtE,2791
+arbor_ai-0.2.1.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+arbor_ai-0.2.1.dist-info/entry_points.txt,sha256=PGBX-MfNwfIl8UPFgsX3gjtXLqSogRhOktKMpZUysD0,40
+arbor_ai-0.2.1.dist-info/top_level.txt,sha256=jzWdp3BRYqvZDMFsPajrcftvvlluzVDErkD8IMRfhYs,6
+arbor_ai-0.2.1.dist-info/RECORD,,