Upload README.md

README.md

@@ -1,81 +1,85 @@
 ---
-license: apache-2.0  
-inference: false  
+license: cc-by-sa-4.0
+inference: false
 ---
 
-# slim-q-gen-phi-3
+# SLIM-EXTRACT
 
 <!-- Provide a quick summary of what the model is/does. -->
 
-slim-q-gen-phi-3 is a specialized function-calling model, finetuned on top of a phi-3-mini-4k base, to generate a python dictionary consisting of a single key "question".
+**slim-extract** implements a specialized function-calling customizable 'extract' capability that takes as an input a context passage, a customized key, and outputs a python dictionary with key that corresponds to the customized key, with a value consisting of a list of items extracted from the text corresponding to that key, e.g., 
 
-For most inference use cases, we would recommend using the quantized 'tool' version of this model - [slim-q-gen-phi-3-tool](https://huggingface.co/llmware/slim-q-gen-phi-3-tool)  
+&nbsp;&nbsp;&nbsp;&nbsp;`{'universities': ['Berkeley, Stanford, Yale, University of Florida, ...'] }`
 
-### Model Description
+This model is fine-tuned on top of [**llmware/bling-stable-lm-3b-4e1t-v0**](https://huggingface.co/llmware/bling-stable-lm-3b-4e1t-v0), which in turn, is a fine-tune of stabilityai/stablelm-3b-4elt.
 
-<!-- Provide a longer summary of what this model is. -->
+For fast inference use, we would recommend the 'quantized tool' version, e.g.,  [**'slim-extract-tool'**](https://huggingface.co/llmware/slim-extract-tool).  
 
-- **Developed by:** llmware
-- **Model type:** Dragon 
-- **Language(s) (NLP):** English
-- **License:** Apache 2.0
-- **Finetuned from model:** Microsoft Phi-3
 
+## Prompt format:
 
-## Bias, Risks, and Limitations
+`function = "extract"`  
+`params = "{custom key}"`  
+`prompt = "<human> " + {text} + "\n" + `  
+&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;&nbsp; &nbsp; &nbsp; &nbsp;`"<{function}> " + {params} + "</{function}>" + "\n<bot>:"`  
 
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
 
-Any model can provide inaccurate or incomplete information, and should be used in conjunction with appropriate safeguards and fact-checking mechanisms.
+<details>
+<summary>Transformers Script </summary>
 
+    model = AutoModelForCausalLM.from_pretrained("llmware/slim-extract")
+    tokenizer = AutoTokenizer.from_pretrained("llmware/slim-extract")
 
-## How to Get Started with the Model
+    function = "extract"
+    params = "company"
 
-The fastest way to get started with BLING is through direct import in transformers:
+    text = "Tesla stock declined yesterday 8% in premarket trading after a poorly-received event in San Francisco yesterday, in which the company indicated a likely shortfall in revenue."  
+    
+    prompt = "<human>: " + text + "\n" + f"<{function}> {params} </{function}>\n<bot>:"
 
-    from transformers import AutoTokenizer, AutoModelForCausalLM  
-    tokenizer = AutoTokenizer.from_pretrained("bling-phi-2-v0", trust_remote_code=True)  
-    model = AutoModelForCausalLM.from_pretrained("bling-phi-2-v0", trust_remote_code=True)  
+    inputs = tokenizer(prompt, return_tensors="pt")
+    start_of_input = len(inputs.input_ids[0])
 
-Please refer to the generation_test .py files in the Files repository, which includes 200 samples and script to test the model.  The **generation_test_llmware_script.py** includes built-in llmware capabilities for fact-checking, as well as easy integration with document parsing and actual retrieval to swap out the test set for RAG workflow consisting of business documents.  
-
-The dRAGon model was fine-tuned with a simple "\<human> and \<bot> wrapper", so to get the best results, wrap inference entries as:
-
-    full_prompt = "<human>: " + my_prompt + "\n" + "<bot>:"
-
-The BLING model was fine-tuned with closed-context samples, which assume generally that the prompt consists of two sub-parts:
-
-1.  Text Passage Context, and
-2.  Specific question or instruction based on the text passage
-
-To get the best results, package "my_prompt" as follows:
+    outputs = model.generate(
+        inputs.input_ids.to('cpu'),
+        eos_token_id=tokenizer.eos_token_id,
+        pad_token_id=tokenizer.eos_token_id,
+        do_sample=True,
+        temperature=0.3,
+        max_new_tokens=100
+    )
 
-    my_prompt = {{text_passage}} + "\n" + {{question/instruction}}
+    output_only = tokenizer.decode(outputs[0][start_of_input:], skip_special_tokens=True)
 
+    print("output only: ", output_only)  
 
-If you are using a HuggingFace generation script:
+    # here's the fun part
+    try:
+        output_only = ast.literal_eval(llm_string_output)
+        print("success - converted to python dictionary automatically")
+    except:
+        print("fail - could not convert to python dictionary automatically - ", llm_string_output)
+   
+   </details>  
+ 
+<details>  
 
-    # prepare prompt packaging used in fine-tuning process
-    new_prompt = "<human>: " + entries["context"] + "\n" + entries["query"] + "\n" + "<bot>:"
 
-    inputs = tokenizer(new_prompt, return_tensors="pt")  
-    start_of_output = len(inputs.input_ids[0])
 
-    #   temperature: set at 0.3 for consistency of output
-    #   max_new_tokens:  set at 100 - may prematurely stop a few of the summaries
+    
+<summary>Using as Function Call in LLMWare</summary>
 
-    outputs = model.generate(
-            inputs.input_ids.to(device),
-            eos_token_id=tokenizer.eos_token_id,
-            pad_token_id=tokenizer.eos_token_id,
-            do_sample=True,
-            temperature=0.3,
-            max_new_tokens=100,
-            )
+    from llmware.models import ModelCatalog
+    slim_model = ModelCatalog().load_model("llmware/slim-extract")
+    response = slim_model.function_call(text,params=["company"], function="extract")
 
-    output_only = tokenizer.decode(outputs[0][start_of_output:],skip_special_tokens=True)  
+    print("llmware - llm_response: ", response)
 
+</details>  
 
+    
 ## Model Card Contact
 
-Darren Oberst & llmware team
+Darren Oberst & llmware team  
+
+[Join us on Discord](https://discord.gg/MhZn5Nc39h)