pharrow/TinyLlama-HTMLWeb-coder

1---2datasets:3- Tesslate/UIGEN-T24base_model:5- TinyLlama/TinyLlama-1.1B-Chat-v1.06---7---8    license: apache-2.09    tags:10      - tinyllama11      - causal-lm12      - merged-lora13    base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.014    merged_from:15      - snaplora-adapted16    ---17 18    # TinyLlama (Merged LoRA)19 20    This repository contains a TinyLlama model with LoRA weights merged into the base.21 22    - **Base model:** `TinyLlama/TinyLlama-1.1B-Chat-v1.0`  23    - **Adapter:** `snaplora-adapted`  24    - **Merge date:** 2025-09-14 23:12:26Z UTC25 26    ## Usage27 28    ```python29    from transformers import AutoModelForCausalLM, AutoTokenizer30    import torch31 32    model_id = "<this-repo-id>"33    tok = AutoTokenizer.from_pretrained(model_id, use_fast=True)34    model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")35 36    prompt = "Write a haiku about tiny models."37    inputs = tok(prompt, return_tensors="pt").to(model.device)38    with torch.no_grad():39        out = model.generate(**inputs, max_new_tokens=64)40    print(tok.decode(out[0], skip_special_tokens=True))41    ```42 43    ## Notes44 45    - The adapter was merged into the base weights using `peft.PeftModel.merge_and_unload()`.46    - Files are saved with `safetensors` when possible.