File size: 1,370 Bytes
47618cd
25eeaa2
 
8b7971e
 
 
47618cd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3216427
840c47d
3216427
 
 
 
 
 
 
 
 
 
 
 
db1f5ea
3216427
db1f5ea
3216427
 
 
 
 
 
 
 
 
db1f5ea
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
---
language:
- en
license: apache-2.0
library_name: peft
base_model: databricks/dolly-v2-3b
---
## Training procedure


The following `bitsandbytes` quantization config was used during training:
- load_in_8bit: False
- load_in_4bit: True
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: nf4
- bnb_4bit_use_double_quant: True
- bnb_4bit_compute_dtype: bfloat16
### Framework versions


- PEFT 0.4.0.dev0

### Import Model
```python
import torch
from peft import PeftModel, PeftConfig
from transformers import AutoModelForCausalLM, AutoTokenizer

peft_model_id = "AhmedBou/databricks-dolly-v2-3b_for_clinical_terms_synonyms"
config = PeftConfig.from_pretrained(peft_model_id)
model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=True, device_map='auto')
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)

# Load the Lora model
model = PeftModel.from_pretrained(model, peft_model_id)
```

### Model Inference

```python
input_text = "participant safety -->: "
batch = tokenizer(input_text, return_tensors='pt')

with torch.cuda.amp.autocast():
    output_tokens = model.generate(**batch, max_new_tokens=50)

print('\n\n', tokenizer.decode(output_tokens[0], skip_special_tokens=True))
```