# Phi-2: causal language modeling with Transformers

This page is the Transformers contract of the Phi-2 MIC and also a usage example. [doc-builder](https://github.com/huggingface/doc-builder/blob/main/docs/runnable-code-blocks.md) renders the `runnable` blocks below as documentation, and its pytest plugin runs them as written against `microsoft/phi-2`. The MIC runner runs the same blocks against the slim fixture and the checkpoint in [`integration.json`](../../integration.json). Lines marked `# doc-builder: hide` only matter there: they swap in the pinned checkpoint or variant and print the result record.

## Score a batch of prompts

```py runnable:phi2
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, set_seed
import json  # doc-builder: hide
import os  # doc-builder: hide

model_id = "microsoft/phi-2"
model_id = os.environ.get("MIC_MODEL", model_id)  # doc-builder: hide
variant = json.loads(os.environ.get("MIC_DEVICE_KWARGS", "{}"))  # doc-builder: hide
variant = json.loads(os.environ.get("MIC_MODEL_KWARGS", "{}"))  # doc-builder: hide
device = "cuda" if torch.cuda.is_available() else "cpu"
set_seed(0)  # seeds torch, NumPy, and Python: loading and dropout draw random numbers

tokenizer = AutoTokenizer.from_pretrained(model_id)
# phi-2 ships in float16; load float32 so the training step below needs no loss scaling
model = AutoModelForCausalLM.from_pretrained(model_id, dtype=torch.float32).to(device)
if variant:  # doc-builder: hide
    del model  # doc-builder: hide
    model = AutoModelForCausalLM.from_pretrained(model_id, **{"dtype": torch.float32, **variant}).to(device)  # doc-builder: hide

# phi-2's tokenizer has no pad token; left padding keeps the last position a real token
if tokenizer.pad_token is None:
    tokenizer.pad_token = tokenizer.eos_token
tokenizer.padding_side = "left"

prompts = [
    "def print_prime(n):\n    \"\"\"\n    Print all primes between 1 and n\n    \"\"\"",
    "The capital of France is",
]
inputs = tokenizer(prompts, return_tensors="pt", padding=True).to(device)

with torch.no_grad():
    logits = model(**inputs).logits
# probe the last position (left padding puts the real last token there)
probe = logits[:, -1, :]
probe_ids = probe.topk(3, dim=-1).indices
print(tokenizer.batch_decode(probe_ids))
```

## One training step

```py runnable:phi2:2
labels = inputs["input_ids"].clone()
labels[inputs["attention_mask"] == 0] = -100

model.train()
optimizer = torch.optim.SGD([p for p in model.parameters() if p.requires_grad], lr=0.01)
loss = model(**inputs, labels=labels).loss
loss.backward()
before = {n: p.detach().cpu().clone() for n, p in model.named_parameters() if p.requires_grad}  # doc-builder: hide
optimizer.step()
print(f"loss: {loss.item():.4f}")

if os.environ.get("MIC_RUNNER"):  # doc-builder: hide
    named = dict(model.named_parameters())
    checked = ("model.layers.0.self_attn.q_proj.weight", "model.layers.0.mlp.fc1.weight")
    result = {
        "runtime": {
            "device": model.device.type,
            "dtype": str(model.dtype).removeprefix("torch."),
            "attn_implementation": model.config._attn_implementation,
        },
        "logits": {"probe": probe.float().tolist()},
        "generated_ids": probe_ids.tolist(),
        "training": {
            "loss": loss.item(),
            "gradients": {n: bool(named[n].grad is not None and torch.count_nonzero(named[n].grad) > 0 and torch.isfinite(named[n].grad).all()) for n in checked},
            "updated": any(not torch.equal(p.detach().cpu(), before[n]) for n, p in model.named_parameters() if n in before),
        },
    }
    print(json.dumps(result))
```