Spaces:
No application file
No application file
File size: 8,857 Bytes
4f24301 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 | import gc
from typing import Tuple, Dict, Any, Optional, List, Generator
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers.generation.streamers import TextIteratorStreamer
from threading import Thread
from deepforest_agent.conf.config import Config
class Llama32ModelManager:
"""
Manages Llama-3.2-3B-Instruct model instances for text generation tasks.
Attributes:
model_id (str): HuggingFace model identifier
load_count (int): Number of times model has been loaded
"""
def __init__(self, model_id: str = Config.AGENT_MODELS["ecology_analysis"]):
"""
Initialize the Llama-3.2-3B model manager.
Args:
model_id (str, optional): HuggingFace model identifier.
Defaults to "meta-llama/Llama-3.2-3B-Instruct".
"""
self.model_id = model_id
self.load_count = 0
def generate_response(
self,
messages: List[Dict[str, str]],
max_new_tokens: int = Config.AGENT_CONFIGS["ecology_analysis"]["max_new_tokens"],
temperature: float = Config.AGENT_CONFIGS["ecology_analysis"]["temperature"],
top_p: float = Config.AGENT_CONFIGS["ecology_analysis"]["top_p"],
tools: Optional[List[Dict[str, Any]]] = None
) -> str:
"""
Generate text response using Llama-3.2-3B-Instruct.
Args:
messages: List of message dictionaries with 'role' and 'content'
max_new_tokens: Maximum tokens to generate
temperature: Sampling temperature
top_p: Top-p sampling
tools (Optional[List[Dict[str, Any]]]): List of tools (not used for Llama)
Returns:
str: Generated response text
Raises:
Exception: If generation fails due to model issues, memory, or other errors
"""
print(f"Loading Llama-3.2-3B for inference #{self.load_count + 1}")
model, tokenizer = self._load_model()
self.load_count += 1
try:
# Llama uses standard chat template without xml_tools
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
generated_ids = model.generate(
model_inputs.input_ids,
max_new_tokens=max_new_tokens,
temperature=temperature,
top_p=top_p,
do_sample=True,
pad_token_id=tokenizer.eos_token_id
)
generated_ids = [
output_ids[len(input_ids):]
for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
return response
except Exception as e:
print(f"Error during Llama-3.2-3B text generation: {e}")
raise e
finally:
print(f"Releasing Llama-3.2-3B GPU memory after inference")
if 'model' in locals():
if hasattr(model, 'cpu'):
model.cpu()
del model
if 'tokenizer' in locals():
del tokenizer
if 'model_inputs' in locals():
del model_inputs
if 'generated_ids' in locals():
del generated_ids
# Multiple garbage collection passes
for _ in range(3):
gc.collect()
if torch.cuda.is_available():
torch.cuda.empty_cache()
torch.cuda.ipc_collect()
torch.cuda.synchronize()
try:
torch.cuda.memory._record_memory_history(enabled=None)
except:
pass
print(f"GPU memory after aggressive cleanup: {torch.cuda.memory_allocated() / 1024**3:.2f} GB allocated, {torch.cuda.memory_reserved() / 1024**3:.2f} GB cached")
def generate_response_streaming(
self,
messages: List[Dict[str, str]],
max_new_tokens: int = Config.AGENT_CONFIGS["ecology_analysis"]["max_new_tokens"],
temperature: float = Config.AGENT_CONFIGS["ecology_analysis"]["temperature"],
top_p: float = Config.AGENT_CONFIGS["ecology_analysis"]["top_p"],
) -> Generator[Dict[str, Any], None, None]:
"""
Generate text response with streaming (token by token).
Args:
messages: List of message dictionaries with 'role' and 'content'
max_new_tokens: Maximum tokens to generate
temperature: Sampling temperature
top_p: Top-p sampling
Yields:
Dict[str, Any]: Dictionary containing:
- token: The generated token/text chunk
- is_complete: Whether generation is finished
Raises:
Exception: If generation fails due to model issues, memory, or other errors
"""
print(f"Loading Llama-3.2-3B for streaming inference #{self.load_count + 1}")
model, tokenizer = self._load_model()
self.load_count += 1
try:
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)
streamer = TextIteratorStreamer(
tokenizer,
timeout=60.0,
skip_prompt=True,
skip_special_tokens=True
)
generation_kwargs = {
"input_ids": model_inputs.input_ids,
"max_new_tokens": max_new_tokens,
"temperature": temperature,
"top_p": top_p,
"do_sample": True,
"pad_token_id": tokenizer.eos_token_id,
"streamer": streamer
}
thread = Thread(target=model.generate, kwargs=generation_kwargs)
thread.start()
for new_text in streamer:
yield {"token": new_text, "is_complete": False}
thread.join()
yield {"token": "", "is_complete": True}
except Exception as e:
print(f"Error during Llama-3.2-3B streaming generation: {e}")
yield {"token": f"[Error: {str(e)}]", "is_complete": True}
finally:
print(f"Releasing Llama-3.2-3B GPU memory after inference")
if 'model' in locals():
if hasattr(model, 'cpu'):
model.cpu()
del model
if 'tokenizer' in locals():
del tokenizer
if 'model_inputs' in locals():
del model_inputs
if 'generated_ids' in locals():
del generated_ids
# Multiple garbage collection passes
for _ in range(3):
gc.collect()
if torch.cuda.is_available():
torch.cuda.empty_cache()
torch.cuda.ipc_collect()
torch.cuda.synchronize()
try:
torch.cuda.memory._record_memory_history(enabled=None)
except:
pass
print(f"GPU memory after aggressive cleanup: {torch.cuda.memory_allocated() / 1024**3:.2f} GB allocated, {torch.cuda.memory_reserved() / 1024**3:.2f} GB cached")
def _load_model(self) -> Tuple[AutoModelForCausalLM, AutoTokenizer]:
"""
Private method for model and tokenizer loading.
Returns:
Tuple[AutoModelForCausalLM, AutoTokenizer]: Loaded model and tokenizer
Raises:
Exception: If model loading fails due to network, memory, or other issues
"""
try:
tokenizer = AutoTokenizer.from_pretrained(
self.model_id,
trust_remote_code=True
)
# Llama models may need specific configurations
model = AutoModelForCausalLM.from_pretrained(
self.model_id,
torch_dtype="auto",
device_map="auto",
trust_remote_code=True,
low_cpu_mem_usage=True
)
return model, tokenizer
except Exception as e:
print(f"Error loading Llama-3.2-3B model: {e}")
raise e |