Refresh model card: dllm-hub style + arXiv 2604.26951 + PKU-YuanGroup/TIDE link + logo

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 tokenizer.json filter=lfs diff=lfs merge=lfs -text
+logo.gif filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,58 +1,104 @@
 ---
 license: apache-2.0
 library_name: transformers
+pipeline_tag: text-generation
+base_model: dllm-collection/Qwen3-0.6B-diffusion-bd3lm-v0.1
 tags:
-- diffusion-language-model
-- distillation
+- diffusion
 - dllm
-- qwen3
 - bd3lm
-base_model: dllm-collection/Qwen3-0.6B-diffusion-bd3lm-v0.1
+- distillation
 ---
 
+<center> <div style="text-align: center;"> <img src="logo.gif" width="300" />
+ </div> </center>
+
 # distill-WeDLM-TIDE_Shared
 
-This is the **TIDE-Shared (native, Pipeline B)** student checkpoint from
-*Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models*.
-It is a 0.6B Qwen3-BD3LM diffusion language model distilled in the **Shared-Tokenizer (WeDLM → BD3LM)** pipeline.
-Native variant for the shared-tokenizer pipeline; TIDAL + CompDemo.
+This model was introduced in the paper [Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models](https://huggingface.co/papers/2604.26951) — this is the **native (paper-best) variant** for its pipeline.
 
-- Code: https://github.com/Nobody-Zhang/dllm_release
-- Project page: https://pku-yuangroup.github.io/TIDE-Page/
-- Paper: https://arxiv.org/abs/XXXX.XXXXX
+`distill-WeDLM-TIDE_Shared` is a 0.6B diffusion language model distilled from WeDLM-8B-Instruct (8B dense) into the [`Qwen3-0.6B-diffusion-bd3lm-v0.1`](https://huggingface.co/dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1) student in the **Shared-Tokenizer (Pipeline B)** of the TIDE framework. Native variant for the shared-tokenizer pipeline; **TIDAL + CompDemo** over forward KL.
 
-## Loading
+## Model Overview
+
+- **Method**: TIDE — [Reverse CALM / TIDAL / CompDemo](https://arxiv.org/abs/2604.26951) (cross-architecture distillation for diffusion LMs)
+- **Framework**: [TIDE / dLLM](https://github.com/PKU-YuanGroup/TIDE)
+- **Student (initialization)**: [`Qwen3-0.6B-diffusion-bd3lm-v0.1`](https://huggingface.co/dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1) (BD3LM, block_size=32)
+- **Teacher**: [`tencent/WeDLM-8B-Instruct`](https://huggingface.co/tencent/WeDLM-8B-Instruct)
+- **Distillation mode**: `--distill_mode taid_aligned --use_comp_demo True`
+- **Datasets**: [tulu-3-sft-mixture](https://huggingface.co/datasets/allenai/tulu-3-sft-mixture), [smoltalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk), [opc-sft-stage1](https://huggingface.co/datasets/OpenCoder-LLM/opc-sft-stage1) and [opc-sft-stage2](https://huggingface.co/datasets/OpenCoder-LLM/opc-sft-stage2) — same composition as the [`Qwen3-0.6B-diffusion-bd3lm-v0.1`](https://huggingface.co/dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1) base. Pre-tokenized for this teacher in [`TIDE-dllm/distill_wedlm_sft`](https://huggingface.co/datasets/TIDE-dllm/distill_wedlm_sft).
+
+## Installation
+
+```shell
+pip install torch transformers accelerate
+```
 
-This checkpoint uses custom modeling code (`A2DQwen3LMHeadModel`) registered via the `auto_map` field in `config.json`. Pass `trust_remote_code=True`:
+## Quick Start
+
+> [!NOTE]
+> This checkpoint is fully compatible with the BD3LM `generate(...)` routine published with [`dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1`](https://huggingface.co/dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1) — only the model name changes.
 
 ```python
-from transformers import AutoConfig, AutoModelForMaskedLM, AutoTokenizer
+import torch
+from transformers import AutoModelForMaskedLM, AutoTokenizer
 
 repo = "TIDE-dllm/distill-WeDLM-TIDE_Shared"
-cfg = AutoConfig.from_pretrained(repo, trust_remote_code=True)
-tok = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
-model = AutoModelForMaskedLM.from_pretrained(repo, trust_remote_code=True)
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+model = AutoModelForMaskedLM.from_pretrained(
+    repo, dtype=torch.bfloat16, trust_remote_code=True,
+).to(device).eval()
+tokenizer = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
+
+prompts = [
+    [
+        {"role": "system", "content": "You are a helpful AI assistant."},
+        {"role": "user", "content": "Implement a DFS traversal in Python with clear inline comments."},
+    ],
+]
+encoded = [tokenizer.apply_chat_template(m, add_generation_prompt=True, tokenize=True, enable_thinking=False) for m in prompts]
+# ... use the same `generate()` function as in dllm-hub/Qwen3-0.6B-diffusion-bd3lm-v0.1.
 ```
 
-For sampling and evaluation, install the companion library:
+## Command-Line Interface
+
+For an interactive demo (visualised iterative denoising), use the script in the [TIDE / dLLM repo](https://github.com/PKU-YuanGroup/TIDE):
 
-```bash
-git clone https://github.com/Nobody-Zhang/dllm_release && cd dllm_release
-pip install -e .
+```shell
+python -u examples/a2d/bd3lm/chat.py \
+    --model_name_or_path TIDE-dllm/distill-WeDLM-TIDE_Shared \
+    --chat_template True --block_size 32 --remasking low_confidence \
+    --steps 256 --max_new_tokens 256
 ```
 
-See the GitHub README for full sampling, evaluation, and distillation instructions.
+## Reproducing this checkpoint
+
+```shell
+git clone https://github.com/PKU-YuanGroup/TIDE && cd TIDE
+pip install -e . && git submodule update --init --recursive
+pip install -e "lm-evaluation-harness[ifeval,math]" && pip install -e "tokenkit[full]"
+
+# Download the pre-tokenized SFT mixture for this teacher
+huggingface-cli download TIDE-dllm/distill_wedlm_sft --repo-type dataset \
+    --local-dir data/distill_wedlm_sft
+
+bash scripts/distill_wedlm.sh \
+    --data_path data/distill_wedlm_sft \
+    --distill_mode taid_aligned --use_comp_demo True \
+    --num_gpus 8
+```
 
 ## Citation
 
 ```bibtex
-@misc{tide2026,
-  title  = {Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models},
-  author = {Zhang, Gongbo and Wang, Wen and Tian, Ye and Yuan, Li},
-  year   = {2026},
-  eprint = {XXXX.XXXXX},
-  archivePrefix = {arXiv},
-  primaryClass  = {cs.CL},
-  url = {https://arxiv.org/abs/XXXX.XXXXX},
+@misc{zhang2026turningtidecrossarchitecturedistillation,
+      title={Turning the TIDE: Cross-Architecture Distillation for Diffusion Large Language Models},
+      author={Gongbo Zhang and Wen Wang and Ye Tian and Li Yuan},
+      year={2026},
+      eprint={2604.26951},
+      archivePrefix={arXiv},
+      primaryClass={cs.CL},
+      url={https://arxiv.org/abs/2604.26951},
 }
 ```