Upload folder using huggingface_hub

serve.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+# Script to serve the DeepSWE-Preview-FP8 model with vLLM
+
+# Source conda
+source /home/op/miniconda3/etc/profile.d/conda.sh
+
+# Activate the vllm-model environment
+conda activate vllm-model
+
+# Note: pyparsing is required by pydot which is used by some vLLM components
+# It has been installed in the vllm-model environment to prevent warnings
+
+# Run the serving script
+python /home/op/DeepSWE-Preview-FP8/serve_model.py

serve_model.py

@@ -0,0 +1,48 @@
+#!/usr/bin/env python3
+"""
+Script to serve the DeepSWE-Preview-FP8 model using vLLM with specific configurations:
+- CUDA devices 1,2
+- Max model length 32000
+- Tensor parallel size 2
+"""
+
+import os
+import subprocess
+import sys
+
+def serve_model():
+    # Set CUDA_VISIBLE_DEVICES to use only GPUs 1 and 2
+    os.environ["CUDA_VISIBLE_DEVICES"] = "1,2"
+    
+    # Build the vLLM command
+    cmd = [
+        "python", "-m", "vllm.entrypoints.openai.api_server",
+        "--host", "0.0.0.0",
+        "--port", "8550",
+        "--model", "/home/op/DeepSWE-Preview-FP8",  # Current directory
+        "--max-model-len", "32000",
+        "--tensor-parallel-size", "2",
+        "--pipeline-parallel-size", "1",
+    ]
+    
+    print("Starting vLLM server with the following configuration:")
+    print(f"CUDA_VISIBLE_DEVICES: {os.environ['CUDA_VISIBLE_DEVICES']}")
+    print(f"Model path: /home/op/DeepSWE-Preview-FP8")
+    print(f"Max model length: 32000")
+    print(f"Tensor parallel size: 2")
+    print(f"Pipeline parallel size: 1")
+    print("\nCommand:", " ".join(cmd))
+    print("\n" + "="*50)
+    
+    # Run the command
+    try:
+        subprocess.run(cmd, check=True)
+    except subprocess.CalledProcessError as e:
+        print(f"Error running vLLM server: {e}")
+        sys.exit(1)
+    except KeyboardInterrupt:
+        print("\nServer stopped by user")
+        sys.exit(0)
+
+if __name__ == "__main__":
+    serve_model()

serve_vllm.sh

@@ -0,0 +1,23 @@
+#!/bin/bash
+# Direct vLLM serving script for DeepSWE-Preview-FP8 model
+
+# Set CUDA devices
+export CUDA_VISIBLE_DEVICES=1,2
+
+# Source conda
+source /home/op/miniconda3/etc/profile.d/conda.sh
+
+# Activate the vllm-model environment
+conda activate vllm-model
+
+# Ensure pyparsing is available (needed by pydot)
+# This was previously causing warnings about missing pyparsing module
+
+# Run vLLM with all specified parameters
+python -m vllm.entrypoints.openai.api_server \\
+    --host 0.0.0.0 \\
+    --port 8550 \\
+    --model /home/op/DeepSWE-Preview-FP8 \\
+    --max-model-len 28000 \\
+    --tensor-parallel-size 2 \\
+    --trust-remote-code

upload_to_hf.py

@@ -0,0 +1,67 @@
+import os
+from huggingface_hub import HfApi
+
+# Initialize the Hugging Face API
+api = HfApi()
+
+# Repository details - using the agentica-org namespace as requested
+repo_id = "agentica-org/DeepSWE-Preview-FP8"  # Using the organization namespace
+local_dir = "/home/op/DeepSWE-Preview-FP8"
+
+# Files to upload (excluding serving scripts)
+files_to_upload = [
+    "README.md",
+    "added_tokens.json",
+    "chat_template.jinja",
+    "config.json",
+    "generation_config.json",
+    "merges.txt",
+    "model-00001-of-00007.safetensors",
+    "model-00002-of-00007.safetensors",
+    "model-00003-of-00007.safetensors",
+    "model-00004-of-00007.safetensors",
+    "model-00005-of-00007.safetensors",
+    "model-00006-of-00007.safetensors",
+    "model-00007-of-00007.safetensors",
+    "model.safetensors.index.json",
+    "special_tokens_map.json",
+    "tokenizer_config.json",
+    "tokenizer.json",
+    "vocab.json"
+]
+
+def upload_model():
+    print(f"Creating repository {repo_id}...")
+    
+    # Create the repository if it doesn't exist
+    try:
+        api.create_repo(repo_id=repo_id, repo_type="model", private=False, exist_ok=True)
+        print(f"Repository {repo_id} created or already exists.")
+    except Exception as e:
+        print(f"Error creating repository: {e}")
+        return
+    
+    # Upload files
+    print("Uploading files...")
+    for file_name in files_to_upload:
+        file_path = os.path.join(local_dir, file_name)
+        if os.path.exists(file_path):
+            try:
+                print(f"Uploading {file_name}...")
+                api.upload_file(
+                    path_or_fileobj=file_path,
+                    path_in_repo=file_name,
+                    repo_id=repo_id,
+                    repo_type="model"
+                )
+                print(f"Uploaded {file_name}")
+            except Exception as e:
+                print(f"Error uploading {file_name}: {e}")
+        else:
+            print(f"File not found: {file_path}")
+    
+    print("Upload completed!")
+    print(f"Model uploaded to: https://huggingface.co/{repo_id}")
+
+if __name__ == "__main__":
+    upload_model()