Add model card

README.md

@@ -0,0 +1,91 @@
+# SmolDeepSeek-V4 100M
+
+A small ~110M parameter language model implementing the **DeepSeek-V4 architecture**, fine-tuned for chat/instruction following. Trained from scratch — no weights from DeepSeek-V4 were used.
+
+- **Pretrained base model**: [cmpatino/smol-deepseek-v4-100m-pretrain](https://huggingface.co/cmpatino/smol-deepseek-v4-100m-pretrain)
+- **This model**: SFT on [HuggingFaceTB/smol-smoltalk](https://huggingface.co/datasets/HuggingFaceTB/smol-smoltalk)
+
+## Architecture
+
+This model implements key DeepSeek-V4 innovations at a miniature scale:
+
+| Component | Details |
+|---|---|
+| **Parameters** | ~110M total (41M embeddings, 69M non-embedding) |
+| **Hidden size** | 320 |
+| **Layers** | 8 |
+| **Attention heads** | 8 (1 KV head — MQA-style) |
+| **MLA** | Multi-head Latent Attention with q_lora_rank=160 |
+| **MoE** | 4 routed experts + 1 shared, top-2 routing |
+| **Hyper-Connections** | hc_mult=4, Sinkhorn routing (replacing residual connections) |
+| **MTP** | 1 next-token prediction layer |
+| **Vocab** | 129,280 (DeepSeek-V4 tokenizer) |
+| **Context** | 2,048 tokens |
+
+## Training
+
+### Stage 1: Pretraining
+- **Dataset**: [HuggingFaceFW/fineweb-edu](https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu)
+- **Steps**: 5,000 | **Tokens**: ~2.6B
+- **Batch**: 32 effective (8 × 4 GA) | **Seq length**: 2,048
+- **LR**: 6e-4, cosine, 3% warmup
+- **Precision**: bf16 mixed
+
+### Stage 2: SFT (this model)
+- **Dataset**: [HuggingFaceTB/smol-smoltalk](https://huggingface.co/datasets/HuggingFaceTB/smol-smoltalk) (460K conversations)
+- **Steps**: 3,000 | **Tokens**: ~72.7M
+- **Batch**: 32 effective (8 × 4 GA) | **Seq length**: 2,048
+- **LR**: 2e-5, cosine, 5% warmup
+- **Precision**: fp32
+
+### Metrics
+
+| Metric | Pretrained | SFT |
+|---|---|---|
+| **Eval loss** | — | 2.607 |
+| **Perplexity** (held-out) | 13.62 | 12.90 |
+| **Token accuracy** | 33.8% | 48.5% |
+
+## Usage
+
+```python
+import torch
+from safetensors.torch import load_file
+from transformers import AutoConfig, AutoModelForCausalLM, AutoTokenizer
+
+# Load model (recommended: manual load for reliability)
+config = AutoConfig.from_pretrained("cmpatino/smol-deepseek-v4-100m", trust_remote_code=True)
+model = AutoModelForCausalLM.from_config(config, trust_remote_code=True).float()
+
+# Download and load weights
+from huggingface_hub import hf_hub_download
+weights_path = hf_hub_download("cmpatino/smol-deepseek-v4-100m", "model.safetensors")
+state_dict = load_file(weights_path)
+model.load_state_dict(state_dict, strict=True)
+model = model.cuda().eval()
+
+tokenizer = AutoTokenizer.from_pretrained("cmpatino/smol-deepseek-v4-100m")
+
+# Chat
+messages = [{"role": "user", "content": "What are 3 benefits of exercise?"}]
+prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+input_ids = tokenizer.encode(prompt, return_tensors="pt").cuda()
+output = model.generate(input_ids, max_new_tokens=200, temperature=0.7, top_p=0.9,
+                        pad_token_id=tokenizer.eos_token_id)
+print(tokenizer.decode(output[0][input_ids.shape[1]:], skip_special_tokens=True))
+```
+
+## Limitations
+
+- **Tiny model**: 110M params with 129K vocabulary — most capacity goes to embeddings. Generations are often incoherent or factually wrong.
+- **Undertrained**: Only 5K pretrain + 3K SFT steps. Production models train for 100K+ steps on trillions of tokens.
+- **Educational purpose**: This model demonstrates the DeepSeek-V4 architecture at small scale. It is **not** suitable for any production use.
+- **Custom code**: Requires `trust_remote_code=True`.
+
+## Hardware
+
+Trained on 1× NVIDIA H100 80GB.
+
+## License
+
+Apache-2.0