Clarify DiBO prompt formatting

State that the inherited tokenizer chat template is not used by the DiBO training or evaluation path.

README.md

@@ -50,12 +50,12 @@ from transformers import AutoModel, AutoTokenizer
 repo_id = "zpointsun/DiBO-TFBind8"
 tokenizer = AutoTokenizer.from_pretrained(
     repo_id,
-    revision="v1.1.6",
+    revision="v1.1.7",
     trust_remote_code=True,
 )
 model = AutoModel.from_pretrained(
     repo_id,
-    revision="v1.1.6",
+    revision="v1.1.7",
     trust_remote_code=True,
     use_safetensors=True,
     torch_dtype="auto",
@@ -66,6 +66,11 @@ model.eval()
 The packaged tokenizer already includes the four DiBO delimiter tokens. Do not
 add them or resize embeddings again after loading this export.
 
+The tokenizer configuration retains LLaDA's `chat_template` metadata, but DiBO
+does not call `apply_chat_template` during training or evaluation. DiBO directly
+tokenizes its rendered unified prompt-response corpus with the delimiter tokens
+above; do not insert chat headers when reproducing the released evaluation path.
+
 ## B. Download and load the original checkpoint
 
 The original artifact uses the released DiBO loader, which initializes the
@@ -74,7 +79,7 @@ and strictly loads `checkpoint["model"]`.
 
 ```bash
 hf download zpointsun/DiBO-TFBind8 dibo_tfbind8_final.pt \
-  --revision v1.1.6 --local-dir checkpoints/dibo-tfbind8
+  --revision v1.1.7 --local-dir checkpoints/dibo-tfbind8
 ```
 
 ```python
@@ -87,7 +92,7 @@ assert LLADA_MODEL_REVISION == "08b83a6feb34df1a6011b80c3c00c7563e963b07"
 checkpoint_path = hf_hub_download(
     "zpointsun/DiBO-TFBind8",
     filename="dibo_tfbind8_final.pt",
-    revision="v1.1.6",
+    revision="v1.1.7",
 )
 model, tokenizer = load_model_and_tokenizer(DEFAULT_MODEL_ID, device="cuda")
 checkpoint = torch.load(checkpoint_path, map_location="cuda")
@@ -102,7 +107,7 @@ From a checkout of the released DiBO code and its oracle environment:
 ```bash
 # Standard Transformers export
 python eval.py --tasks TFBind8-Exact-v0 \
-  --model_name_or_path zpointsun/DiBO-TFBind8 --model_revision v1.1.6 \
+  --model_name_or_path zpointsun/DiBO-TFBind8 --model_revision v1.1.7 \
   --seeds <SEEDS> --max_attempts 1000
 
 # Canonical local .pt checkpoint

SHA256SUMS

@@ -1,19 +1,20 @@
-c5ad7abc64bf9d84fe407ffaa65015297c0b6873855c78c754ab36eb6fbd90d8  README.md
-ffaa3217d5f7d83a027fca9df0070c0082726c6770f44bb94f1caf931036539b  UPSTREAM_NOTICE.md
-ea63d0120e6dac7cfdb9ded5e2f4f9716092ee6f8e1b416b65aa60a8e66164f8  __init__.py
-b1d2f6933a003e283cf779b8c46b387bd674d99429db348fe7760063987eab60  artifact_manifest.json
-fa264b764d6e0efe12731bd63bed9725795e188458de4192d418759f3cc9407f  config.json
-c947a64d8b735affa69d5adeb416c83ce607c8925270884e4471ae11068e2bfe  configuration_llada.py
-0d8e17f7aa60baf4173d71a4ec8475d182a992812b59d29a728eec719d1b67bf  dibo_tfbind8_final.pt
-1afd19bb061c1f0ae8e664681b84c53a0f12e50daf7464f900e0bc4353b15670  export_validation.json
-14d78817248c7e2c9ffe477316ce17b0a51c12e8f5057ddf506ba175991a7a1c  generation_config.json
-b1d0d9d09be47f1c47dff91926fd7f3eda272dc2ecbdd1b1a68f4bf42984473c  model-00001-of-00004.safetensors
-dbe8101a92f660a2c2bd3d5d3dae6cc58f8258ccafc5d487d627f4aa7f48f7ad  model-00002-of-00004.safetensors
-62c12ec9f508f42811eaa1880dfcb7c4fbca1a48d8379e796d2e8f34f978abfe  model-00003-of-00004.safetensors
-0a6587d94453234ffa5d0e18cddfde93219c40754a2d5359e0ce1e095f714286  model-00004-of-00004.safetensors
-06db2635e42448b011de23b0b5b4ba22bc1f582ca5af2e745d48123adafc3e27  model.safetensors.index.json
-c1d3021bbf2d1291251ea5bc5033f6cbbcdc1e2452285463bae90e0747d8b898  modeling_dibo_llada.py
-98bac7e53fef0bb7ca01e3716c11a7f710d183e10dbb9783b88db9dbba2e3766  modeling_llada.py
-09e3c271fe6b5a3ab93f8a8f3b7239d198c6f5d45eb2eaf8399e4ec5af813dc2  special_tokens_map.json
-99174cd2c97e90835223e88ed9d7c704f457e725073ca74fcd495b83ba584b56  tokenizer.json
-d6f02013659be27a0d9e2f7fc304f8a95b86ad58c648e9af6bee038feb91cb03  tokenizer_config.json
+model.safetensors.index.json  06db2635e42448b011de23b0b5b4ba22bc1f582ca5af2e745d48123adafc3e27
+special_tokens_map.json  09e3c271fe6b5a3ab93f8a8f3b7239d198c6f5d45eb2eaf8399e4ec5af813dc2
+model-00004-of-00004.safetensors  0a6587d94453234ffa5d0e18cddfde93219c40754a2d5359e0ce1e095f714286
+dibo_tfbind8_final.pt  0d8e17f7aa60baf4173d71a4ec8475d182a992812b59d29a728eec719d1b67bf
+generation_config.json  14d78817248c7e2c9ffe477316ce17b0a51c12e8f5057ddf506ba175991a7a1c
+export_validation.json  1afd19bb061c1f0ae8e664681b84c53a0f12e50daf7464f900e0bc4353b15670
+model-00003-of-00004.safetensors  62c12ec9f508f42811eaa1880dfcb7c4fbca1a48d8379e796d2e8f34f978abfe
+modeling_llada.py  98bac7e53fef0bb7ca01e3716c11a7f710d183e10dbb9783b88db9dbba2e3766
+tokenizer.json  99174cd2c97e90835223e88ed9d7c704f457e725073ca74fcd495b83ba584b56
+7c12b1e90339820f7945f8ea52b08bb1c12e498065227722e3599a99cdbe56dd  README.md
+model-00001-of-00004.safetensors  b1d0d9d09be47f1c47dff91926fd7f3eda272dc2ecbdd1b1a68f4bf42984473c
+artifact_manifest.json  b1d2f6933a003e283cf779b8c46b387bd674d99429db348fe7760063987eab60
+modeling_dibo_llada.py  c1d3021bbf2d1291251ea5bc5033f6cbbcdc1e2452285463bae90e0747d8b898
+README.md  c5ad7abc64bf9d84fe407ffaa65015297c0b6873855c78c754ab36eb6fbd90d8
+configuration_llada.py  c947a64d8b735affa69d5adeb416c83ce607c8925270884e4471ae11068e2bfe
+tokenizer_config.json  d6f02013659be27a0d9e2f7fc304f8a95b86ad58c648e9af6bee038feb91cb03
+model-00002-of-00004.safetensors  dbe8101a92f660a2c2bd3d5d3dae6cc58f8258ccafc5d487d627f4aa7f48f7ad
+__init__.py  ea63d0120e6dac7cfdb9ded5e2f4f9716092ee6f8e1b416b65aa60a8e66164f8
+config.json  fa264b764d6e0efe12731bd63bed9725795e188458de4192d418759f3cc9407f
+UPSTREAM_NOTICE.md  ffaa3217d5f7d83a027fca9df0070c0082726c6770f44bb94f1caf931036539b