Views
No views yet
mistralai/Mixtral-8x7B-Instruct-v0.1. This model was trained with SFT using Adaption's AutoScientist on the ai_medical_questions dataset.
1{
2 "job_id": "4f125dd7-bbbe-46a2-a4bf-ec81900a95b9",
3 "training_experiment_id": "33e0fe5f-0261-4db3-a7a6-bcf3e212a6a3",
4 "original_model_name": "mistralai/Mixtral-8x7B-Instruct-v0.1",
5 "trained_model_name": "adaption_ai_medical_questions",
6 "training_method": "sft",
7 "training_type": "lora",
8 "data_format": "chat",
9 "hyperparams": {
10 "lora": "true",
11 "lora_r": 64,
12 "n_epochs": 5,
13 "batch_size": "max",
14 "lora_alpha": 128,
15 "lora_dropout": 0,
16 "min_lr_ratio": 0.1,
17 "warmup_ratio": 0.05,
18 "weight_decay": 0.05,
19 "learning_rate": 0.0001,
20 "max_grad_norm": 1,
21 "train_on_inputs": "false",
22 "training_method": "sft",
23 "lr_scheduler_type": "cosine",
24 "scheduler_num_cycles": 0.5,
25 "lora_trainable_modules": "all-linear"
26 }
27}
| Domain | Win rate vs. base model |
|---|---|
| medical | 67% |
pip install torch transformers peft1import torch
2from transformers import AutoModelForCausalLM, AutoTokenizer
3from peft import PeftModel
4
5BASE = "mistralai/Mixtral-8x7B-Instruct-v0.1"
6ADAPTER = "<this-repo-id>"
7
8device = "cuda" if torch.cuda.is_available() else "cpu"
9dtype = torch.float32 if device == "cpu" else torch.bfloat16
10
11base = AutoModelForCausalLM.from_pretrained(BASE, dtype=dtype).to(device)
12model = PeftModel.from_pretrained(base, ADAPTER)
13# Optional: merge the LoRA weights into the base for faster inference
14model = model.merge_and_unload()
15model.eval()
16
17tokenizer = AutoTokenizer.from_pretrained(BASE)
18messages = [{"role": "user", "content": "Hello!"}]
19text = tokenizer.apply_chat_template(
20 messages, tokenize=False, add_generation_prompt=True)
21inputs = tokenizer(text, return_tensors="pt").to(device)
22
23with torch.inference_mode():
24 out = model.generate(**inputs, max_new_tokens=512)
25print(tokenizer.decode(out[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))