PraisonAI 2.0.75__cp313-cp313-manylinux_2_39_x86_64.whl → 2.0.77__cp313-cp313-manylinux_2_39_x86_64.whl

praisonai/deploy.py

@@ -56,7 +56,7 @@ class CloudDeployer:
             file.write("FROM python:3.11-slim\n")
             file.write("WORKDIR /app\n")
             file.write("COPY . .\n")
-            file.write("RUN pip install flask praisonai==2.0.75 gunicorn markdown\n")
+            file.write("RUN pip install flask praisonai==2.0.77 gunicorn markdown\n")
             file.write("EXPOSE 8080\n")
             file.write('CMD ["gunicorn", "-b", "0.0.0.0:8080", "api:app"]\n')
             

praisonai/train_vision.py

@@ -12,11 +12,14 @@ import yaml
 import torch
 import shutil
 import subprocess
+import gc  # For garbage collection
 
-from datasets import load_dataset, concatenate_datasets
+from datasets import load_dataset, concatenate_datasets, Dataset
 from unsloth import FastVisionModel, is_bf16_supported
 from unsloth.trainer import UnslothVisionDataCollator
-from trl import SFTTrainer, SFTConfig
+from transformers import TrainingArguments
+from trl import SFTTrainer
+from tqdm import tqdm  # Add progress bar
 
 
 class TrainVisionModel:
@@ -62,11 +65,21 @@ class TrainVisionModel:
             use_gradient_checkpointing="unsloth"
         )
         print("DEBUG: Vision model and original tokenizer loaded.")
-        if original_tokenizer.pad_token is None:
-            original_tokenizer.pad_token = original_tokenizer.eos_token
-        original_tokenizer.model_max_length = self.config.get("max_seq_length", 2048)
+        
+        # Use the full processor that supports image inputs.
         self.hf_tokenizer = original_tokenizer
         
+        # Set pad token if needed
+        if not hasattr(self.hf_tokenizer, 'pad_token') or self.hf_tokenizer.pad_token is None:
+            if hasattr(self.hf_tokenizer, 'eos_token'):
+                self.hf_tokenizer.pad_token = self.hf_tokenizer.eos_token
+            elif hasattr(self.hf_tokenizer, 'bos_token'):
+                self.hf_tokenizer.pad_token = self.hf_tokenizer.bos_token
+        
+        # Set max length
+        if hasattr(self.hf_tokenizer, 'model_max_length'):
+            self.hf_tokenizer.model_max_length = self.config.get("max_seq_length", 2048)
+        
         # Add vision-specific LoRA adapters
         self.model = FastVisionModel.get_peft_model(
             self.model,
@@ -85,38 +98,62 @@ class TrainVisionModel:
         print("DEBUG: Vision LoRA adapters added.")
 
     def convert_sample(self, sample):
-        # Use a default instruction or one from config
-        instr = self.config.get("vision_instruction", "You are an expert radiographer. Describe accurately what you see in this image.")
+        
+        instruction = self.config.get(
+            "vision_instruction",
+            "You are an expert radiographer. Describe accurately what you see in this image."
+        )
         conversation = [
-            {"role": "user", "content": [
-                {"type": "text", "text": instr},
-                {"type": "image", "image": sample["image"]}
-            ]},
-            {"role": "assistant", "content": [
-                {"type": "text", "text": sample["caption"]}
-            ]}
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": instruction},
+                    {"type": "image", "image": sample["image"]}
+                ]
+            },
+            {
+                "role": "assistant",
+                "content": [
+                    {"type": "text", "text": sample["caption"]}
+                ]
+            },
         ]
+        
         return {"messages": conversation}
 
     def load_datasets(self):
-        datasets = []
+        all_converted = []
         for dataset_info in self.config["dataset"]:
-            print("DEBUG: Loading vision dataset:", dataset_info)
-            ds = load_dataset(dataset_info["name"], split=dataset_info.get("split_type", "train"))
-            print("DEBUG: Converting dataset to vision conversation format...")
-            ds = ds.map(self.convert_sample)
-            datasets.append(ds)
-        combined = concatenate_datasets(datasets)
-        print("DEBUG: Combined vision dataset has", len(combined), "examples.")
-        return combined
+            print("\nDEBUG: Loading vision dataset:", dataset_info)
+            ds = load_dataset(
+                dataset_info["name"], 
+                split=dataset_info.get("split_type", "train")
+            )
+            print("DEBUG: Dataset size:", len(ds))
+            print("DEBUG: First raw sample:", ds[0])
+            print("DEBUG: Dataset features:", ds.features)
+            
+            print("\nDEBUG: Converting dataset to vision conversation format...")
+            converted_ds = [self.convert_sample(sample) for sample in ds]
+            
+            # Debug first converted sample
+            print("\nDEBUG: First converted sample structure:")
+            first = converted_ds[0]
+            print("DEBUG: Message keys:", first["messages"][0]["content"][1].keys())
+            print("DEBUG: Image type in converted:", type(first["messages"][0]["content"][1].get("image")))
+            
+            all_converted.extend(converted_ds)
+        
+        print("\nDEBUG: Combined vision dataset has", len(all_converted), "examples.")
+        return all_converted
 
     def train_model(self):
         print("DEBUG: Starting vision training...")
         raw_dataset = self.load_datasets()
         
-        # Build training arguments using SFTConfig for vision tasks
-        sft_config = SFTConfig(
-            per_device_train_batch_size=self.config.get("per_device_train_batch_size", 2),
+        # Build training arguments using TrainingArguments
+        training_args = TrainingArguments(
+            per_device_train_batch_size=self.config.get("per_device_train_batch_size", 1),
             gradient_accumulation_steps=self.config.get("gradient_accumulation_steps", 4),
             warmup_steps=self.config.get("warmup_steps", 5),
             max_steps=self.config.get("max_steps", 30),
@@ -131,10 +168,9 @@ class TrainVisionModel:
             output_dir=self.config.get("output_dir", "outputs"),
             report_to="none" if not os.getenv("PRAISON_WANDB") else "wandb",
             remove_unused_columns=False,
-            dataset_text_field="",
-            dataset_kwargs={"skip_prepare_dataset": True},
-            dataset_num_proc=self.config.get("dataset_num_proc", 4),
-            max_seq_length=self.config.get("max_seq_length", 2048)
+            # Add memory optimization settings
+            gradient_checkpointing=True,
+            max_grad_norm=1.0,
         )
         
         trainer = SFTTrainer(
@@ -142,7 +178,11 @@ class TrainVisionModel:
             tokenizer=self.hf_tokenizer,
             data_collator=UnslothVisionDataCollator(self.model, self.hf_tokenizer),
             train_dataset=raw_dataset,
-            args=sft_config
+            args=training_args,
+            max_seq_length=self.config.get("max_seq_length", 2048),
+            dataset_text_field="",  # Required for vision training
+            dataset_kwargs={"skip_prepare_dataset": True},  # Required for vision training
+            packing=False  # Explicitly set packing to False
         )
         print("DEBUG: Beginning vision trainer.train() ...")
         trainer.train()
@@ -200,6 +240,32 @@ class TrainVisionModel:
             quantization_method="q4_k_m"
         )
 
+    def prepare_modelfile_content(self):
+        output_model = self.config["hf_model_name"]
+        
+        template = '''{{- range $index, $_ := .Messages }}<|start_header_id|>{{ .Role }}<|end_header_id|>
+
+{{ .Content }}
+{{- if gt (len (slice $.Messages $index)) 1 }}<|eot_id|>
+{{- else if ne .Role "assistant" }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+
+{{ end }}
+{{- end }}'''
+        
+        return f"""FROM {output_model}
+TEMPLATE {template}
+PARAMETER temperature 0.6
+PARAMETER top_p 0.9
+"""
+
+    def create_and_push_ollama_model(self):
+        modelfile_content = self.prepare_modelfile_content()
+        with open("Modelfile", "w") as file:
+            file.write(modelfile_content)
+        subprocess.run(["ollama", "serve"])
+        subprocess.run(["ollama", "create", f"{self.config['ollama_model']}:{self.config['model_parameters']}", "-f", "Modelfile"])
+        subprocess.run(["ollama", "push", f"{self.config['ollama_model']}:{self.config['model_parameters']}"])
+
     def run(self):
         self.print_system_info()
         self.check_gpu()
@@ -207,6 +273,12 @@ class TrainVisionModel:
         if self.config.get("train", "true").lower() == "true":
             self.prepare_model()
             self.train_model()
+        if self.config.get("huggingface_save", "true").lower() == "true":
+            self.save_model_merged()
+        if self.config.get("huggingface_save_gguf", "true").lower() == "true":
+            self.push_model_gguf()
+        if self.config.get("ollama_save", "true").lower() == "true":
+            self.create_and_push_ollama_model()
 
 
 def main():

praisonai/upload_vision.py

@@ -0,0 +1,140 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""
+This script handles uploading trained vision models to Hugging Face and Ollama.
+It reads configuration from config.yaml and provides options to upload in different formats.
+"""
+
+import os
+import yaml
+import torch
+import shutil
+import subprocess
+from unsloth import FastVisionModel
+
+class UploadVisionModel:
+    def __init__(self, config_path="config.yaml"):
+        self.load_config(config_path)
+        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        self.model = None
+        self.hf_tokenizer = None
+
+    def load_config(self, path):
+        """Load configuration from yaml file."""
+        with open(path, "r") as file:
+            self.config = yaml.safe_load(file)
+        print("DEBUG: Loaded config:", self.config)
+
+    def prepare_model(self):
+        """Load the trained model for uploading."""
+        print("DEBUG: Loading trained model and tokenizer...")
+        self.model, original_tokenizer = FastVisionModel.from_pretrained(
+            model_name=self.config.get("output_dir", "lora_model"),
+            load_in_4bit=self.config.get("load_in_4bit", True)
+        )
+        self.hf_tokenizer = original_tokenizer
+        print("DEBUG: Model and tokenizer loaded successfully.")
+
+    def save_model_merged(self):
+        """Save merged model to Hugging Face Hub."""
+        print(f"DEBUG: Saving merged model to Hugging Face Hub: {self.config['hf_model_name']}")
+        if os.path.exists(self.config["hf_model_name"]):
+            shutil.rmtree(self.config["hf_model_name"])
+        self.model.push_to_hub_merged(
+            self.config["hf_model_name"],
+            self.hf_tokenizer,
+            save_method="merged_16bit",
+            token=os.getenv("HF_TOKEN")
+        )
+        print("DEBUG: Model saved to Hugging Face Hub successfully.")
+
+    def push_model_gguf(self):
+        """Push model in GGUF format to Hugging Face Hub."""
+        print(f"DEBUG: Pushing GGUF model to Hugging Face Hub: {self.config['hf_model_name']}")
+        self.model.push_to_hub_gguf(
+            self.config["hf_model_name"],
+            self.hf_tokenizer,
+            quantization_method=self.config.get("quantization_method", "q4_k_m"),
+            token=os.getenv("HF_TOKEN")
+        )
+        print("DEBUG: GGUF model pushed to Hugging Face Hub successfully.")
+
+    def prepare_modelfile_content(self):
+        """Prepare Ollama modelfile content using Llama 3.2 vision template."""
+        output_model = self.config["hf_model_name"]
+        
+        # Using Llama 3.2 vision template format
+        template = """{{- range $index, $_ := .Messages }}<|start_header_id|>{{ .Role }}<|end_header_id|>
+
+{{ .Content }}
+{{- if gt (len (slice $.Messages $index)) 1 }}<|eot_id|>
+{{- else if ne .Role "assistant" }}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
+
+{{ end }}
+{{- end }}"""
+        
+        # Assemble the modelfile content with Llama 3.2 vision parameters
+        modelfile = f"FROM {output_model}\n"
+        modelfile += "TEMPLATE \"""" + template + "\"""\n"
+        modelfile += "PARAMETER temperature 0.6\n"
+        modelfile += "PARAMETER top_p 0.9\n"
+        return modelfile
+
+    def create_and_push_ollama_model(self):
+        """Create and push model to Ollama."""
+        print(f"DEBUG: Creating Ollama model: {self.config['ollama_model']}:{self.config['model_parameters']}")
+        modelfile_content = self.prepare_modelfile_content()
+        with open("Modelfile", "w") as file:
+            file.write(modelfile_content)
+        
+        print("DEBUG: Starting Ollama server...")
+        subprocess.run(["ollama", "serve"])
+        
+        print("DEBUG: Creating Ollama model...")
+        subprocess.run([
+            "ollama", "create", 
+            f"{self.config['ollama_model']}:{self.config['model_parameters']}", 
+            "-f", "Modelfile"
+        ])
+        
+        print("DEBUG: Pushing model to Ollama...")
+        subprocess.run([
+            "ollama", "push", 
+            f"{self.config['ollama_model']}:{self.config['model_parameters']}"
+        ])
+        print("DEBUG: Model pushed to Ollama successfully.")
+
+    def upload(self, target="all"):
+        """
+        Upload the model to specified targets.
+        Args:
+            target (str): One of 'all', 'huggingface', 'huggingface_gguf', or 'ollama'
+        """
+        self.prepare_model()
+        
+        if target in ["all", "huggingface"]:
+            self.save_model_merged()
+        
+        if target in ["all", "huggingface_gguf"]:
+            self.push_model_gguf()
+            
+        if target in ["all", "ollama"]:
+            self.create_and_push_ollama_model()
+
+def main():
+    import argparse
+    parser = argparse.ArgumentParser(description="Upload Vision Model to Various Platforms")
+    parser.add_argument("--config", default="config.yaml", help="Path to configuration file")
+    parser.add_argument(
+        "--target", 
+        choices=["all", "huggingface", "huggingface_gguf", "ollama"],
+        default="all",
+        help="Target platform to upload to"
+    )
+    args = parser.parse_args()
+
+    uploader = UploadVisionModel(config_path=args.config)
+    uploader.upload(target=args.target)
+
+if __name__ == "__main__":
+    main()

{praisonai-2.0.75.dist-info → praisonai-2.0.77.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: PraisonAI
-Version: 2.0.75
+Version: 2.0.77
 Summary: PraisonAI is an AI Agents Framework with Self Reflection. PraisonAI application combines PraisonAI Agents, AutoGen, and CrewAI into a low-code solution for building and managing multi-agent LLM systems, focusing on simplicity, customisation, and efficient human-agent collaboration.
 Author: Mervin Praison
 Requires-Python: >=3.10,<3.13
@@ -61,7 +61,7 @@ Requires-Dist: playwright (>=1.47.0) ; extra == "code"
 Requires-Dist: plotly (>=5.24.0) ; extra == "realtime"
 Requires-Dist: praisonai-tools (>=0.0.7) ; extra == "autogen"
 Requires-Dist: praisonai-tools (>=0.0.7) ; extra == "crewai"
-Requires-Dist: praisonaiagents (>=0.0.61)
+Requires-Dist: praisonaiagents (>=0.0.62)
 Requires-Dist: pyautogen (>=0.2.19) ; extra == "autogen"
 Requires-Dist: pydantic (<=2.10.1) ; extra == "chat"
 Requires-Dist: pydantic (<=2.10.1) ; extra == "code"

{praisonai-2.0.75.dist-info → praisonai-2.0.77.dist-info}/RECORD

@@ -5,7 +5,7 @@ praisonai/api/call.py,sha256=krOfTCZM_bdbsNuWQ1PijzCHECkDvEi9jIvvZaDQUUU,11035
 praisonai/auto.py,sha256=uLDm8CU3L_3amZsd55yzf9RdBF1uW-BGSx7nl9ctNZ4,8680
 praisonai/chainlit_ui.py,sha256=bNR7s509lp0I9JlJNvwCZRUZosC64qdvlFCt8NmFamQ,12216
 praisonai/cli.py,sha256=hxGPiX8-LZanu2jiwBXIkMPm8Kk0Tt3LwDclLkUt0iY,26051
-praisonai/deploy.py,sha256=7jpkrKSFqCmtqOE1Ha0hYGAed_Rmq20mA-VXshHtaCc,6028
+praisonai/deploy.py,sha256=2gLKcli7gR-BHJSWTkWzYgmwl5VDbf4k5J02dziIAVM,6028
 praisonai/inbuilt_tools/__init__.py,sha256=fai4ZJIKz7-iOnGZv5jJX0wmT77PKa4x2jqyaJddKFA,569
 praisonai/inbuilt_tools/autogen_tools.py,sha256=kJdEv61BTYvdHOaURNEpBcWq8Rs-oC03loNFTIjT-ak,4687
 praisonai/inc/__init__.py,sha256=sPDlYBBwdk0VlWzaaM_lG0_LD07lS2HRGvPdxXJFiYg,62
@@ -34,7 +34,7 @@ praisonai/setup/setup_conda_env.sh,sha256=_pVbrXStZua6vUJTbuGiZam-zWsDDLWP0ZaFuP
 praisonai/setup.py,sha256=0jHgKnIPCtBZiGYaYyTz3PzrJI6nBy55VXk2UctXlDo,373
 praisonai/test.py,sha256=OL-wesjA5JTohr8rtr6kWoaS4ImkJg2l0GXJ-dUUfRU,4090
 praisonai/train.py,sha256=Cjb0TKU3esNrCk2OX24Qm1S1crRC00FdiGUYJLw3iPQ,24094
-praisonai/train_vision.py,sha256=VnKqI2A-d4k9enjqnHZ6D2qo-ueYTfppCkNWqyZAIFE,9699
+praisonai/train_vision.py,sha256=OLDtr5u9rszWQ80LC5iFy37yPuYguES6AQybm_2RtM4,12514
 praisonai/ui/README.md,sha256=QG9yucvBieVjCjWFzu6hL9xNtYllkoqyJ_q1b0YYAco,1124
 praisonai/ui/agents.py,sha256=1qsWE2yCaQKhuc-1uLHdMfZJeOXzBtp4pe5q7bk2EuA,32813
 praisonai/ui/callbacks.py,sha256=V4_-GjxmjDFmugUZGfQHKtNSysx7rT6i1UblbM_8lIM,1968
@@ -82,9 +82,10 @@ praisonai/ui/realtimeclient/realtimedocs.txt,sha256=hmgd8Uwy2SkjSndyyF_-ZOaNxiyH
 praisonai/ui/realtimeclient/tools.py,sha256=IJOYwVOBW5Ocn5_iV9pFkmSKR3WU3YpX3kwF0I3jikQ,7855
 praisonai/ui/sql_alchemy.py,sha256=oekZOXlRGMJ2SuC-lmgMMIzAmvbMg2DWeGTSpOzbVBM,29674
 praisonai/ui/tools.md,sha256=Ad3YH_ZCLMWlz3mDXllQnQ_S5l55LWqLdcZSh-EXrHI,3956
+praisonai/upload_vision.py,sha256=lMpFn993UiYVJxRNZQTmcbPbEajQ5TFKCNGK1Icn_hg,5253
 praisonai/version.py,sha256=ugyuFliEqtAwQmH4sTlc16YXKYbFWDmfyk87fErB8-8,21
-praisonai-2.0.75.dist-info/LICENSE,sha256=kqvFysVlnFxYOu0HxCe2HlmZmJtdmNGOxWRRkT9TsWc,1035
-praisonai-2.0.75.dist-info/METADATA,sha256=iUUjKd5zb5QehWbDd0wh70ytL0FxhjjI1ly_wYOcm3Y,21942
-praisonai-2.0.75.dist-info/WHEEL,sha256=OiNztsphQWM3l0xJ9BHQRElMnxzHbt1M68r2N60f8T8,110
-praisonai-2.0.75.dist-info/entry_points.txt,sha256=I_xc6a6MNTTfLxYmAxe0rgey0G-_hbY07oFW-ZDnkw4,135
-praisonai-2.0.75.dist-info/RECORD,,
+praisonai-2.0.77.dist-info/LICENSE,sha256=kqvFysVlnFxYOu0HxCe2HlmZmJtdmNGOxWRRkT9TsWc,1035
+praisonai-2.0.77.dist-info/METADATA,sha256=RZFBDDFCCji6tnnu4sBE-qX6rjtAOM6hS_YV5slwQjo,21942
+praisonai-2.0.77.dist-info/WHEEL,sha256=IC-58YCXkB2pVoJBgAzESes-vAnjpjsOuhYWk8sUmtI,110
+praisonai-2.0.77.dist-info/entry_points.txt,sha256=I_xc6a6MNTTfLxYmAxe0rgey0G-_hbY07oFW-ZDnkw4,135
+praisonai-2.0.77.dist-info/RECORD,,

{praisonai-2.0.75.dist-info → praisonai-2.0.77.dist-info}/WHEEL

@@ -1,4 +1,4 @@
 Wheel-Version: 1.0
-Generator: poetry-core 2.0.1
+Generator: poetry-core 2.1.1
 Root-Is-Purelib: false
 Tag: cp313-cp313-manylinux_2_39_x86_64