File size: 3,583 Bytes
0c5425e | 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 | # DeepSeek-R1: causal language modeling with Transformers
This page is the Transformers contract of the DeepSeek-R1 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 `deepseek-ai/DeepSeek-R1`. 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.
## Predict the next token
```py runnable:causal_lm
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, set_seed
import json # doc-builder: hide
import os # doc-builder: hide
model_id = "deepseek-ai/DeepSeek-R1"
model_id = os.environ.get("MIC_MODEL", model_id) # 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, dropout, and data augmentation draw random numbers
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id).to(device)
if variant: # doc-builder: hide
del model # doc-builder: hide
model = AutoModelForCausalLM.from_pretrained(model_id, **variant).to(device) # doc-builder: hide
prompts = [
"The capital of France is",
"The capital of Germany is",
]
# Left padding keeps the last position the real next-token position for every row.
tokenizer.padding_side = "left"
inputs = tokenizer(prompts, return_tensors="pt", padding=True).to(device)
model.eval()
with torch.no_grad():
logits = model(**inputs).logits
next_token_ids = logits[:, -1, :].argmax(dim=-1)
print(tokenizer.batch_decode(next_token_ids[:, None]))
```
## One training step
```py runnable:causal_lm:2
# Causal LM labels: copy input_ids and mask padded positions.
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_a_proj.weight",
"model.layers.0.mlp.gate_proj.weight",
)
result = {
"runtime": {
"device": model.device.type,
"dtype": str(model.dtype).removeprefix("torch."),
"attn_implementation": model.config._attn_implementation,
},
"next_token_ids": next_token_ids.tolist(),
"probe_logits": logits[:, -1, :5].float().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))
```
|