File size: 2,718 Bytes
ed6b901 15eb8ca ed6b901 |
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 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 |
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
from datasets import load_dataset
from tqdm import tqdm
import json
import csv
import os
import evaluate
# ==== CONFIG ====
MODEL_PATH = "../train/output/qlora-codellama-bugfix"
EVAL_FILE = "test_samples.jsonl"
OUTPUT_JSON = "./output/eval_results.json"
OUTPUT_CSV = "./output/eval_results.csv"
MAX_INPUT_LEN = 1024
MAX_NEW_TOKENS = 256
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
# ==== Ensure output folder exists ====
os.makedirs(os.path.dirname(OUTPUT_JSON), exist_ok=True)
# ==== Load model ====
print("🔄 Loading model...")
tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
model = AutoModelForCausalLM.from_pretrained(
MODEL_PATH,
torch_dtype=torch.bfloat16,
device_map="auto"
)
model.eval()
# ==== Load eval data ====
print("📂 Loading evaluation data...")
eval_data = load_dataset("json", data_files=EVAL_FILE, split="train")
# ==== Inference ====
results = []
print("⚙️ Running inference...")
for example in tqdm(eval_data):
prompt = example["prompt"]
reference = example["completion"]
inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=MAX_INPUT_LEN).to(DEVICE)
with torch.no_grad():
outputs = model.generate(
**inputs,
max_new_tokens=MAX_NEW_TOKENS,
do_sample=False,
num_beams=4,
pad_token_id=tokenizer.pad_token_id,
eos_token_id=tokenizer.eos_token_id
)
prediction = tokenizer.decode(outputs[0], skip_special_tokens=True)
results.append({
"prompt": prompt,
"reference": reference.strip(),
"prediction": prediction.strip()
})
# ==== Save results ====
with open(OUTPUT_JSON, "w", encoding="utf-8") as f:
json.dump(results, f, indent=2)
print(f"✅ Saved JSON to {OUTPUT_JSON}")
with open(OUTPUT_CSV, "w", encoding="utf-8", newline='') as f:
writer = csv.DictWriter(f, fieldnames=["prompt", "reference", "prediction"])
writer.writeheader()
writer.writerows(results)
print(f"✅ Saved CSV to {OUTPUT_CSV}")
# ==== Compute Metrics ====
print("📊 Computing BLEU and ROUGE...")
bleu = evaluate.load("bleu")
rouge = evaluate.load("rouge")
predictions = [r["prediction"] for r in results]
references = [r["reference"] for r in results]
bleu_score = bleu.compute(predictions=predictions, references=[[ref] for ref in references])
rouge_score = rouge.compute(predictions=predictions, references=references)
print("\n📈 Evaluation Results:")
print("BLEU:", bleu_score)
print("ROUGE:", json.dumps(rouge_score, indent=2))
|