Upload 2 files

README.md

@@ -1,3 +1,34 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+inference: false 
+base_model: llmware/slim-sql-1b-v0
+base_model_relation: quantized 
+tags: [green, p1, llmware-fx, ov, emerald]
+---
+
+# slim-sql-npu-ov  
+
+**slim-sql-npu-ov** is a small specialized function calling model that takes as input a table schema and a natural language query, and outputs a SQL statement that corresponds to the query, and can be run against a database table.  This is a very small text-to-sql model designed for reasonable accuracy on single tables and relatively straightforward queries, and for easy integration into multi-step processes.     
+
+This is an OpenVino int4 quantized version of slim-sql-1b-v0, providing a very fast, very small inference implementation, optimized for AI PCs using Intel NPU.    
+
+
+### Model Description
+
+- **Developed by:** llmware  
+- **Model type:** tinyllama  
+- **Parameters:** 1.1 billion
+- **Model Parent:** llmware/slim-sql-1b-v0
+- **Language(s) (NLP):** English  
+- **License:** Apache 2.0  
+- **Uses:** Text-to-SQL conversion      
+- **RAG Benchmark Accuracy Score:** NA  
+- **Quantization:** int4  
+  
+## Model Card Contact
+
+[llmware on github](https://www.github.com/llmware-ai/llmware)  
+
+[llmware on hf](https://www.huggingface.co/llmware)  
+
+[llmware website](https://www.llmware.ai)  

config.json

@@ -1,35 +1,105 @@
-{
-  "_name_or_path": "llmware/slim-sql-1b-v0",
-  "aib_version": "",
-  "architectures": [
-    "LlamaForCausalLM"
-  ],
-  "attention_bias": false,
-  "attention_dropout": 0.0,
-  "bos_token_id": 1,
-  "eos_token_id": 2,
-  "head_dim": 64,
-  "hidden_act": "silu",
-  "hidden_size": 2048,
-  "initializer_range": 0.02,
-  "intermediate_size": 5632,
-  "max_position_embeddings": 2048,
-  "mlp_bias": false,
-  "model_type": "llama",
-  "num_attention_heads": 32,
-  "num_hidden_layers": 22,
-  "num_key_value_heads": 4,
-  "pretraining_tp": 1,
-  "rms_norm_eps": 1e-05,
-  "rope_scaling": null,
-  "rope_theta": 10000.0,
-  "tie_word_embeddings": false,
-  "torch_dtype": "float32",
-  "trained": "custom training",
-  "training_comments": "shared_LLaMa-1.3B",
-  "training_dataset": "",
-  "training_timestamp": "Mon Jan  1 00:46:42 2024",
-  "transformers_version": "4.48.3",
-  "use_cache": true,
-  "vocab_size": 32000
-}
+{
+  "_name_or_path": "llmware/slim-sql-ov",
+  "aib_version": "",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 2048,
+  "initializer_range": 0.02,
+  "intermediate_size": 5632,
+  "max_position_embeddings": 2048,
+  "mlp_bias": false,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 22,
+  "num_key_value_heads": 4,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "tie_word_embeddings": false,
+  "trained": "custom training",
+  "training_dataset": "",
+  "transformers_version": "4.41.2",
+  "use_cache": true,
+  "vocab_size": 32000,
+  "prompt_wrapper": "human_bot",
+  "description": "slim-sql is a text-to-sql model.",
+  "prompt_format": "<human> {table_schema} \n {question} \n<bot>:",
+  "output_format": "{sql}",
+  "tokenizer_local": "tokenizer_tl.json",
+  "tokenizer_config": {"bos_id":[1], "bos_token":["<s>"], "eos_id":[2],"eos_token":["</s>"]},
+  "model_parent": "llmware/slim-sql-1b-v0",
+  "description": "Text-to-SQL model from llmware - finetuned on tiny-llama - 1.1 parameter base",
+  "quantization": "int4",
+  "model_family": "OVGenModel",
+  "parameters": 1.1,
+  "primary_keys": ["sql"],
+  "output_values": ["{{sql statement}}"],
+  "publisher": "llmware",
+  "release_date": "september 2024",
+  "function_call": "sql",
+  "test_params": ["sql"],
+   "test_set": [
+  {
+   "context": "CREATE TABLE customers (customer_name text, account_number integer,customer_level text, vip_customer text, annual_spend text, user_name text)",
+   "query": "Which customers are VIP customers?",
+   "answer": "SELECT * FROM customers WHERE vip_customer='yes'"
+  },
+  {
+   "context": "CREATE TABLE customers (customer_name text, account_number integer,customer_level text, vip_customer text, annual_spend text, user_name text)",
+   "query": "What is the annual spend for customer Rachel Michaels?",
+   "answer": "SELECT annual_spend FROM customers WHERE customer_name='Rachel Michaels'"
+  },
+  {
+   "context": "CREATE TABLE customers (customer_name text, account_number integer,customer_level text, vip_customer text, annual_spend text, user_name text)",
+   "query": "How many customers spend more than $1000 per year?",
+   "answer": "SELECT COUNT(*) FROM customers WHERE annual_spend > $1000"
+  },
+  {
+   "context": "CREATE TABLE customers (customer_name text, account_number integer,customer_level text, vip_customer text, annual_spend text, user_name text)",
+   "query": "Who are the customers with gold customer level?",
+   "answer": "SELECT customer_name FROM customers WHERE customer_level = 'gold'"
+  },
+  {
+   "context": "CREATE TABLE customers (customer_name text, account_number integer,customer_level text, vip_customer text, annual_spend text, user_name text)",
+   "query": "Which customer has account number 9382035?",
+   "answer": "SELECT * FROM customers WHERE account_number = 9382035"
+  },
+  {
+   "context": "CREATE TABLE customers (customer_name text, account_number integer,customer_level text, vip_customer text, annual_spend text, user_name text)",
+   "query": "What is the account number of customer Susanna Jones?",
+   "answer": "SELECT account_number FROM customers WHERE customer_name='Susanna Jones'"
+  },
+  {
+   "context": "CREATE TABLE library (library_name text, unique_doc_id integer, documents integer, blocks integer, images integer, pages integer, tables integer, account_name text)",
+   "query": "How many pages are in the human resources library?",
+   "answer": "SELECT pages FROM library WHERE library_name = 'human resources'"
+  },
+  {
+   "context": "CREATE TABLE library (library_name text, unique_doc_id integer, documents integer, blocks integer, images integer, pages integer, tables integer, account_name text)",
+   "query": "Which libraries have more than 1000 images?",
+   "answer": "SELECT * FROM library WHERE images > 1000"
+  },
+  {
+   "context": "CREATE TABLE library (library_name text, unique_doc_id integer, documents integer, blocks integer, images integer, pages integer, tables integer, account_name text)",
+   "query": "How many blocks are in the finance library?",
+   "answer": "SELECT blocks FROM library WHERE library_name = 'finance library'"
+  },
+  {
+   "context": "CREATE TABLE library (library_name text, unique_doc_id integer, documents integer, blocks integer, images integer, pages integer, tables integer, account_name text)",
+   "query": "What is a list of all of the libraries?",
+   "answer": "SELECT * FROM library"
+  },
+  {
+   "context": "CREATE TABLE library (library_name text, unique_doc_id integer, documents integer, blocks integer, images integer, pages integer, tables integer, account_name text)",
+   "query": "Which library has unique_doc_id of 8329?",
+   "answer": "SELECT * FROM library WHERE unique_doc_id = 8329"
+  }
+ ]
+}