53 lines
2.1 KiB
Markdown
53 lines
2.1 KiB
Markdown
---
|
|
license: apache-2.0
|
|
pipeline_tag: text-generation
|
|
tags:
|
|
- finetuned
|
|
inference:
|
|
parameters:
|
|
temperature: 0.01
|
|
---
|
|
|
|
A Mistral7B Instruct (https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.1)
|
|
Finetune using QLoRA on the docs available in https://docs.modular.com/mojo/
|
|
|
|
The Mistral-7B-Instruct-v0.1 Large Language Model (LLM) is a instruct fine-tuned version of the [Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1) generative text model using a variety of publicly available conversation datasets.
|
|
|
|
## Instruction format
|
|
```python
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer, TextStreamer
|
|
import torch
|
|
|
|
device = "cuda" # the device to load the model onto
|
|
|
|
model_name = "mcysqrd/MODULARMOJO_Mistral_V1"
|
|
model = AutoModelForCausalLM.from_pretrained(model_name,
|
|
use_flash_attention_2=True,
|
|
max_memory={0: "24GB"},
|
|
device_map="auto",
|
|
trust_remote_code=True,
|
|
low_cpu_mem_usage=True,
|
|
return_dict=True,
|
|
torch_dtype=torch.bfloat16,
|
|
)
|
|
|
|
tokenizer = AutoTokenizer.from_pretrained(model_name,add_bos_token=True,trust_remote_code=True)
|
|
|
|
model.config.use_cache = True
|
|
def stream(user_prompt):
|
|
runtimeFlag = "cuda:0"
|
|
system_prompt = 'MODULAR_MOJO'
|
|
B_INST, E_INST = "[INST]", "[/INST]"
|
|
prompt = f"{system_prompt}{B_INST}{user_prompt.strip()}\n{E_INST}"
|
|
inputs = tokenizer([prompt], return_tensors="pt").to(runtimeFlag)
|
|
streamer = TextStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
|
|
_ = model.generate(**inputs, streamer=streamer, max_new_tokens=1600)
|
|
|
|
stream("""can you translate this python code to mojo to make more performant making T as struct?
|
|
class T():
|
|
self.init(v:float):
|
|
self.value=v
|
|
|
|
def sum_objects(a:T,b:T)->T:
|
|
return T(a.v+b.v)""")
|
|
``` |