update

modelscope · Nov 3, 2024 · dd0234d · dd0234d
1 parent 8ccbc59
commit dd0234d
Show file tree

Hide file tree

Showing 4 changed files with 6 additions and 5 deletions.
diff --git a/swift/llm/__init__.py b/swift/llm/__init__.py
@@ -7,7 +7,7 @@
     # Recommend using `xxx_main`
     from .infer import (VllmEngine, RequestConfig, InferStats, LmdeployEngine, PtEngine, infer_main, deploy_main,
                         PtLoRARequest, InferClient, SwiftInfer, SwiftDeploy)
-    from .export import export_main, merge_lora, quantize_model
+    from .export import export_main, merge_lora, quantize_model, export_to_ollama
     from .eval import eval_main
     from .train import sft_main, pt_main, rlhf_main
     from .argument import (EvalArguments, InferArguments, SftArguments, ExportArguments, DeployArguments, RLHFArguments,
@@ -31,7 +31,7 @@
             'deploy_main', 'VllmEngine', 'RequestConfig', 'InferStats', 'LmdeployEngine', 'PtEngine', 'infer_main',
             'PtLoRARequest', 'InferClient', 'SwiftInfer', 'SwiftDeploy'
         ],
-        'export': ['export_main', 'merge_lora', 'quantize_model'],
+        'export': ['export_main', 'merge_lora', 'quantize_model', 'export_to_ollama'],
         'eval': ['eval_main'],
         'train': ['sft_main', 'pt_main', 'rlhf_main'],
         'argument': [

diff --git a/swift/llm/export/__init__.py b/swift/llm/export/__init__.py
@@ -2,3 +2,5 @@
 from .merge_lora import merge_lora
 from .export import export_main, SwiftExport
 from .quant import quantize_model
+from .ollama import export_to_ollama
+
diff --git a/swift/llm/export/export.py b/swift/llm/export/export.py
@@ -1,7 +1,7 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 from typing import Any, Dict, List, Union
 
-from swift.llm import (ExportArguments, SwiftPipeline, merge_lora, quantize_model)
+from swift.llm import ExportArguments, SwiftPipeline, merge_lora, quantize_model, export_to_ollama
 from swift.utils import get_logger
 from swift.tuners import swift_to_peft_format
 from swift.hub import default_hub
@@ -21,7 +21,7 @@ def run(self):
         elif args.quant_method is not None:
             quantize_model(args)
         elif args.to_ollama:
-            pass
+            export_to_ollama(args)
         elif args.push_to_hub:
             ckpt_dir = args.ckpt_dir or args.model
             assert ckpt_dir is not None, 'You need to specify `ckpt_dir`.'

diff --git a/swift/llm/export/ollama.py b/swift/llm/export/ollama.py
@@ -54,7 +54,6 @@ def export_to_ollama(args: ExportArguments):
             if isinstance(stop_word, list):
                 stop_word = template.tokenizer.decode(stop_word)
             f.write(f'PARAMETER stop "{stop_word}"\n')
-        generation_config = model.generation_config
         if args.temperature:
             f.write(f'PARAMETER temperature {args.temperature}\n')
         if args.top_k: