length problems

app.py

@@ -25,8 +25,8 @@ HUB_TAGS = {
     "chat": ("conversational", "text-generation", "text2text-generation"),
     "txtqa": ("text-generation", "text2text-generation"),
     "otherqa": ("table-question-answering", "document-question-answering", "visual-question-answering"),
-    "asr": ("automatic-speech-recognition"),
-    "img2txt": ("image-to-text"),
+    "asr": ("automatic-speech-recognition",),
+    "img2txt": ("image-to-text",),
 }
 assert len(TASK_TYPES) == len(TASK_TYPES)
 assert all(tag in HUB_TAGS for tag in TASK_TYPES.values())
@@ -46,7 +46,7 @@ PROBLEMS = {
 INIT_MARKDOWN = """
  
 
-👈 Fill in as much data as you can...
+👈 Fill in as much information as you can...
 
  
 
@@ -68,9 +68,6 @@ DEMO_MARKDOWN = """
 """
 
 SUGGETIONS_HEADER = """
-🫥 NOTE: Clicking on links will crash the app. Copy them into your browser instead.
-See the Gradio issue [here](https://github.com/gradio-app/gradio/issues/3234). 🫥
-
 ✨ Here is a list of suggestions for you -- click to expand ✨
 """
 
@@ -80,19 +77,108 @@ TASK_MODEL_MISMATCH = """
 
 🤔 Why?  
 
-The selected model (`{model_name}`) isn't tagged as a good fit for the task you selected ("{task_type}").  
+The selected model (`{model_name}`) doesn't have a tag compatible with the task you selected ("{task_type}").
+Expected tags for this task are: {tags}  
 
 🤗 How?  
 
-Our recommendation is to go to our [tasks page](https://huggingface.co/tasks) and select one of our suggested
+Our recommendation is to go to our [tasks page](https://huggingface.co/tasks) and select one of the suggested
 models as a starting point.  
 
 😱 Caveats  
 
-The tags of a model are defined by the community and are not always accurate. If you think the model is incorrectly
+1. The tags of a model are defined by the community and are not always accurate. If you think the model is incorrectly
 tagged or missing a tag, please open an issue on the [model card](https://huggingface.co/{model_name}/tree/main).
 </details>
 """
+
+SET_MAX_NEW_TOKENS = """
+<details><summary>{count}. Control the maximum output length with `max_new_tokens`.</summary>
+&nbsp;
+
+🤔 Why? &nbsp;
+
+All text generation calls have a length-related stopping condition. Depending on the model and/or the tool you're
+using to generate text, the default value may be too small or too large. I'd recommend ALWAYS setting this option.
+&nbsp;
+
+🤗 How? &nbsp;
+
+Our text generation interfaces accept a `max_new_tokens` option. Set it to define the maximum number of tokens
+that can be generated. &nbsp;
+
+😱 Caveats &nbsp;
+
+1. Allowing a longer output doesn't necessarily mean that the model will generate longer outputs. By default,
+the model will stop generating when it generates a special `eos_token_id` token.
+2. You shouldn't set `max_new_tokens` to a value larger than the maximum sequence length of the model. If you need a
+longer output, consider using a model with a larger maximum sequence length.
+3. The longer the output, the longer it will take to generate.
+</details>
+"""
+
+SET_MIN_LENGTH = """
+<details><summary>{count}. Force a minimum output length with `min_new_tokens`.</summary>
+&nbsp;
+
+🤔 Why? &nbsp;
+
+Text generation stops when the model generates a special `eos_token_id`. If you prevent it from happening, the model is
+forced to continue generating. &nbsp;
+
+🤗 How? &nbsp;
+
+Our text generation interfaces accept a `min_new_tokens` argument. Set it to prevent `eos_token_id` from being
+generated until `min_new_tokens` tokens are generated. &nbsp;
+
+😱 Caveats &nbsp;
+
+1. The quality of the output may suffer if the model is forced to generate beyond its own original expectations.
+2. `min_new_tokens` must be smaller than than `max_new_tokens` (see related tip).
+</details>
+"""
+
+REMOVE_EOS_TOKEN = """
+<details><summary>{count}. Prevent the model of halting generation by removing `eos_token_id`.</summary>
+&nbsp;
+
+🤔 Why? &nbsp;
+
+Text generation stops when the model generates a special `eos_token_id`. If there is no `eos_token_id`, the model can't
+stop. &nbsp;
+
+
+🤗 How? &nbsp;
+
+Our text generation interfaces accept a `eos_token_id` argument. Set it to a null value (e.g., in Python,
+`eos_token_id=None`) to prevent generation to stop before it reaches other stopping conditions. &nbsp;
+
+😱 Caveats &nbsp;
+
+1. The quality of the output may suffer if the model is forced to generate beyond its own original expectations.
+</details>
+"""
+
+LIST_EOS_TOKEN = """
+<details><summary>{count}. Add a stop word through `eos_token_id`.</summary>
+&nbsp;
+
+🤔 Why? &nbsp;
+
+Text generation stops when the model generates a special `eos_token_id`. Actually, this attribute can be a list of
+tokens, which means you can define arbitrary stop words. &nbsp;
+
+
+🤗 How? &nbsp;
+
+Our text generation interfaces accept a `eos_token_id` argument. You can pass a list of tokens to make generation
+stop in the presence of any of those tokens. &nbsp;
+
+😱 Caveats &nbsp;
+
+1. When passing a list of tokens, you probably shouldn't forget to include the default `eos_token_id` there.
+</details>
+"""
 # =====================================================================================================================
 
 
@@ -122,13 +208,26 @@ def get_suggestions(task_type, model_name, problem_type):
     # Check if the model is valid for the task. If not, return straight away
     if not is_valid_task_for_model(model_name, task_type):
         counter += 1
-        suggestions += TASK_MODEL_MISMATCH.format(count=counter, model_name=model_name, task_type=task_type)
+        possible_tags = " ".join("`" + tag + "`" for tag in HUB_TAGS[TASK_TYPES[task_type]])
+        suggestions += TASK_MODEL_MISMATCH.format(
+            count=counter, model_name=model_name, task_type=task_type, tags=possible_tags
+        )
         return suggestions
 
     # Demo shortcut: only a few sections are working
-    if PROBLEMS.get(problem_type) not in ("", "length", "quality", "speed"):
+    if PROBLEMS.get(problem_type, "") not in ("", "length", "quality", "speed"):
         return DEMO_MARKDOWN
 
+    if PROBLEMS.get(problem_type, "") == "length":
+        counter += 1
+        suggestions += SET_MAX_NEW_TOKENS.format(count=counter)
+        counter += 1
+        suggestions += SET_MIN_LENGTH.format(count=counter)
+        counter += 1
+        suggestions += REMOVE_EOS_TOKEN.format(count=counter)
+        counter += 1
+        suggestions += LIST_EOS_TOKEN.format(count=counter)
+
     return suggestions
 # =====================================================================================================================