Jayashree Sridhar
commited on
Commit
·
f8a9066
1
Parent(s):
94ae986
replaced mistral with TinyGPT2Model
Browse files- models/mistral_model.py +4 -5
models/mistral_model.py
CHANGED
|
@@ -24,9 +24,9 @@ class MistralModel:
|
|
| 24 |
|
| 25 |
def _initialize_model(self):
|
| 26 |
"""Initialize Mistral model with optimizations"""
|
| 27 |
-
print("Loading
|
| 28 |
|
| 29 |
-
model_id = "
|
| 30 |
|
| 31 |
# Load tokenizer
|
| 32 |
MistralModel._tokenizer = AutoTokenizer.from_pretrained(model_id, token=HUGGINGFACE_TOKEN,use_fast=False)
|
|
@@ -35,12 +35,11 @@ class MistralModel:
|
|
| 35 |
MistralModel._model = AutoModelForCausalLM.from_pretrained(
|
| 36 |
model_id,
|
| 37 |
token=HUGGINGFACE_TOKEN,
|
| 38 |
-
torch_dtype=torch.
|
| 39 |
-
device_map="auto",
|
| 40 |
load_in_8bit=True # Use 8-bit quantization for memory efficiency
|
| 41 |
)
|
| 42 |
|
| 43 |
-
print("
|
| 44 |
|
| 45 |
def generate(
|
| 46 |
self,
|
|
|
|
| 24 |
|
| 25 |
def _initialize_model(self):
|
| 26 |
"""Initialize Mistral model with optimizations"""
|
| 27 |
+
print("Loading TinyGPT2Model model...")
|
| 28 |
|
| 29 |
+
model_id = "sshleifer/tiny-gpt2"
|
| 30 |
|
| 31 |
# Load tokenizer
|
| 32 |
MistralModel._tokenizer = AutoTokenizer.from_pretrained(model_id, token=HUGGINGFACE_TOKEN,use_fast=False)
|
|
|
|
| 35 |
MistralModel._model = AutoModelForCausalLM.from_pretrained(
|
| 36 |
model_id,
|
| 37 |
token=HUGGINGFACE_TOKEN,
|
| 38 |
+
torch_dtype=torch.float32,
|
|
|
|
| 39 |
load_in_8bit=True # Use 8-bit quantization for memory efficiency
|
| 40 |
)
|
| 41 |
|
| 42 |
+
print("TinyGPT2Model loaded successfully!")
|
| 43 |
|
| 44 |
def generate(
|
| 45 |
self,
|