Spaces:
Running
on
Zero
Running
on
Zero
Create load_for_inference.py
Browse files- load_for_inference.py +241 -0
load_for_inference.py
ADDED
@@ -0,0 +1,241 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
"""
|
2 |
+
Rose Beeper Model - Inference Example
|
3 |
+
Simple script showing how to load and use the model for text generation
|
4 |
+
"""
|
5 |
+
|
6 |
+
import torch
|
7 |
+
from tokenizers import Tokenizer
|
8 |
+
from huggingface_hub import hf_hub_download
|
9 |
+
|
10 |
+
# Import the extracted components (assuming they're in a module called 'beeper_inference')
|
11 |
+
# from beeper_inference import BeeperRoseGPT, BeeperIO, generate, get_default_config
|
12 |
+
|
13 |
+
def load_model_for_inference(
|
14 |
+
checkpoint_path: str = None,
|
15 |
+
tokenizer_path: str = "beeper.tokenizer.json",
|
16 |
+
hf_repo: str = "AbstractPhil/beeper-rose-v5",
|
17 |
+
device: str = "cuda"
|
18 |
+
):
|
19 |
+
"""
|
20 |
+
Load the Rose Beeper model for inference.
|
21 |
+
|
22 |
+
Args:
|
23 |
+
checkpoint_path: Path to local checkpoint file (.pt or .safetensors)
|
24 |
+
tokenizer_path: Path to tokenizer file
|
25 |
+
hf_repo: HuggingFace repository to download from if no local checkpoint
|
26 |
+
device: Device to load model on ("cuda" or "cpu")
|
27 |
+
|
28 |
+
Returns:
|
29 |
+
Tuple of (model, tokenizer, config)
|
30 |
+
"""
|
31 |
+
# Get default configuration
|
32 |
+
config = get_default_config()
|
33 |
+
|
34 |
+
# Set device
|
35 |
+
device = torch.device(device if torch.cuda.is_available() else "cpu")
|
36 |
+
|
37 |
+
# Initialize model
|
38 |
+
model = BeeperRoseGPT(config).to(device)
|
39 |
+
|
40 |
+
# Initialize pentachora banks
|
41 |
+
# These are the default sizes from the training configuration
|
42 |
+
cap_cfg = config.get("capoera", {})
|
43 |
+
coarse_C = 20 # Approximate number of alive datasets
|
44 |
+
model.ensure_pentachora(
|
45 |
+
coarse_C=coarse_C,
|
46 |
+
medium_C=int(cap_cfg.get("topic_bins", 512)),
|
47 |
+
fine_C=int(cap_cfg.get("mood_bins", 7)),
|
48 |
+
dim=config["dim"],
|
49 |
+
device=device
|
50 |
+
)
|
51 |
+
|
52 |
+
# Load checkpoint
|
53 |
+
loaded = False
|
54 |
+
|
55 |
+
# Try loading from local path
|
56 |
+
if checkpoint_path and os.path.exists(checkpoint_path):
|
57 |
+
print(f"Loading model from: {checkpoint_path}")
|
58 |
+
missing, unexpected = BeeperIO.load_into_model(
|
59 |
+
model, checkpoint_path, map_location="cpu", strict=False
|
60 |
+
)
|
61 |
+
print(f"Loaded | missing={len(missing)} unexpected={len(unexpected)}")
|
62 |
+
loaded = True
|
63 |
+
|
64 |
+
# Try downloading from HuggingFace
|
65 |
+
if not loaded and hf_repo:
|
66 |
+
try:
|
67 |
+
print(f"Downloading model from HuggingFace: {hf_repo}")
|
68 |
+
path = hf_hub_download(repo_id=hf_repo, filename="beeper_final.safetensors")
|
69 |
+
missing, unexpected = BeeperIO.load_into_model(
|
70 |
+
model, path, map_location="cpu", strict=False
|
71 |
+
)
|
72 |
+
print(f"Loaded | missing={len(missing)} unexpected={len(unexpected)}")
|
73 |
+
loaded = True
|
74 |
+
except Exception as e:
|
75 |
+
print(f"Failed to download from HuggingFace: {e}")
|
76 |
+
|
77 |
+
if not loaded:
|
78 |
+
print("WARNING: No weights loaded, using random initialization!")
|
79 |
+
|
80 |
+
# Load tokenizer
|
81 |
+
if os.path.exists(tokenizer_path):
|
82 |
+
tok = Tokenizer.from_file(tokenizer_path)
|
83 |
+
print(f"Loaded tokenizer from: {tokenizer_path}")
|
84 |
+
else:
|
85 |
+
# Try downloading tokenizer from HF
|
86 |
+
try:
|
87 |
+
tok_path = hf_hub_download(repo_id=hf_repo, filename="tokenizer.json")
|
88 |
+
tok = Tokenizer.from_file(tok_path)
|
89 |
+
print(f"Downloaded tokenizer from HuggingFace")
|
90 |
+
except Exception as e:
|
91 |
+
raise RuntimeError(f"Could not load tokenizer: {e}")
|
92 |
+
|
93 |
+
# Set model to eval mode
|
94 |
+
model.eval()
|
95 |
+
|
96 |
+
return model, tok, config
|
97 |
+
|
98 |
+
|
99 |
+
def interactive_generation(model, tokenizer, config, device="cuda"):
|
100 |
+
"""
|
101 |
+
Interactive text generation loop.
|
102 |
+
|
103 |
+
Args:
|
104 |
+
model: The loaded BeeperRoseGPT model
|
105 |
+
tokenizer: The tokenizer
|
106 |
+
config: Model configuration
|
107 |
+
device: Device to run on
|
108 |
+
"""
|
109 |
+
device = torch.device(device if torch.cuda.is_available() else "cpu")
|
110 |
+
model = model.to(device)
|
111 |
+
|
112 |
+
print("\n=== Rose Beeper Interactive Generation ===")
|
113 |
+
print("Enter your prompt (or 'quit' to exit)")
|
114 |
+
print("Commands: /temp <value>, /top_k <value>, /top_p <value>, /max <tokens>")
|
115 |
+
print("-" * 50)
|
116 |
+
|
117 |
+
# Generation settings (can be modified)
|
118 |
+
settings = {
|
119 |
+
"max_new_tokens": 100,
|
120 |
+
"temperature": config["temperature"],
|
121 |
+
"top_k": config["top_k"],
|
122 |
+
"top_p": config["top_p"],
|
123 |
+
"repetition_penalty": config["repetition_penalty"],
|
124 |
+
"presence_penalty": config["presence_penalty"],
|
125 |
+
"frequency_penalty": config["frequency_penalty"],
|
126 |
+
}
|
127 |
+
|
128 |
+
while True:
|
129 |
+
prompt = input("\nPrompt: ").strip()
|
130 |
+
|
131 |
+
if prompt.lower() == 'quit':
|
132 |
+
break
|
133 |
+
|
134 |
+
# Handle commands
|
135 |
+
if prompt.startswith('/'):
|
136 |
+
parts = prompt.split()
|
137 |
+
cmd = parts[0].lower()
|
138 |
+
|
139 |
+
if cmd == '/temp' and len(parts) > 1:
|
140 |
+
settings["temperature"] = float(parts[1])
|
141 |
+
print(f"Temperature set to {settings['temperature']}")
|
142 |
+
continue
|
143 |
+
elif cmd == '/top_k' and len(parts) > 1:
|
144 |
+
settings["top_k"] = int(parts[1])
|
145 |
+
print(f"Top-k set to {settings['top_k']}")
|
146 |
+
continue
|
147 |
+
elif cmd == '/top_p' and len(parts) > 1:
|
148 |
+
settings["top_p"] = float(parts[1])
|
149 |
+
print(f"Top-p set to {settings['top_p']}")
|
150 |
+
continue
|
151 |
+
elif cmd == '/max' and len(parts) > 1:
|
152 |
+
settings["max_new_tokens"] = int(parts[1])
|
153 |
+
print(f"Max tokens set to {settings['max_new_tokens']}")
|
154 |
+
continue
|
155 |
+
else:
|
156 |
+
print("Unknown command")
|
157 |
+
continue
|
158 |
+
|
159 |
+
if not prompt:
|
160 |
+
continue
|
161 |
+
|
162 |
+
# Generate text
|
163 |
+
print("\nGenerating...")
|
164 |
+
output = generate(
|
165 |
+
model=model,
|
166 |
+
tok=tokenizer,
|
167 |
+
cfg=config,
|
168 |
+
prompt=prompt,
|
169 |
+
device=device,
|
170 |
+
**settings
|
171 |
+
)
|
172 |
+
|
173 |
+
print("\nOutput:", output)
|
174 |
+
print("-" * 50)
|
175 |
+
|
176 |
+
|
177 |
+
def batch_generation_example(model, tokenizer, config, device="cuda"):
|
178 |
+
"""
|
179 |
+
Example of batch generation with different settings.
|
180 |
+
"""
|
181 |
+
device = torch.device(device if torch.cuda.is_available() else "cpu")
|
182 |
+
model = model.to(device)
|
183 |
+
|
184 |
+
prompts = [
|
185 |
+
"The robot went to school and",
|
186 |
+
"Once upon a time in a magical forest",
|
187 |
+
"The scientist discovered that",
|
188 |
+
"In the year 2050, humanity",
|
189 |
+
"The philosophy of mind suggests",
|
190 |
+
]
|
191 |
+
|
192 |
+
print("\n=== Batch Generation Examples ===\n")
|
193 |
+
|
194 |
+
for prompt in prompts:
|
195 |
+
print(f"Prompt: {prompt}")
|
196 |
+
|
197 |
+
# Generate with different temperatures
|
198 |
+
for temp in [0.5, 0.9, 1.2]:
|
199 |
+
output = generate(
|
200 |
+
model=model,
|
201 |
+
tok=tokenizer,
|
202 |
+
cfg=config,
|
203 |
+
prompt=prompt,
|
204 |
+
max_new_tokens=50,
|
205 |
+
temperature=temp,
|
206 |
+
device=device
|
207 |
+
)
|
208 |
+
print(f" Temp {temp}: {output}")
|
209 |
+
|
210 |
+
print("-" * 50)
|
211 |
+
|
212 |
+
|
213 |
+
# Main execution example
|
214 |
+
if __name__ == "__main__":
|
215 |
+
import os
|
216 |
+
|
217 |
+
# Load model
|
218 |
+
model, tokenizer, config = load_model_for_inference(
|
219 |
+
checkpoint_path=None, # Will download from HF
|
220 |
+
hf_repo="AbstractPhil/beeper-rose-v5",
|
221 |
+
device="cuda"
|
222 |
+
)
|
223 |
+
|
224 |
+
# Example: Single generation
|
225 |
+
print("\n=== Single Generation Example ===")
|
226 |
+
output = generate(
|
227 |
+
model=model,
|
228 |
+
tok=tokenizer,
|
229 |
+
cfg=config,
|
230 |
+
prompt="The meaning of life is",
|
231 |
+
max_new_tokens=100,
|
232 |
+
temperature=0.9,
|
233 |
+
device="cuda"
|
234 |
+
)
|
235 |
+
print(f"Output: {output}")
|
236 |
+
|
237 |
+
# Example: Batch generation with different settings
|
238 |
+
# batch_generation_example(model, tokenizer, config)
|
239 |
+
|
240 |
+
# Example: Interactive generation
|
241 |
+
# interactive_generation(model, tokenizer, config)
|