| --- |
| language: ["es", "en"] |
| license: apache-2.0 |
| tags: |
| - bittensor |
| - subnet-20 |
| - bitagent |
| - finney |
| - tao |
| - tool-calling |
| - bfcl |
| - reasoning |
| - agent |
| base_model: Salesforce/xLAM-7b-r |
| pipeline_tag: text-generation |
| model-index: |
| - name: antonio-bfcl-toolmodel |
| results: |
| - task: |
| type: text-generation |
| name: Generative reasoning and tool-calling |
| metrics: |
| - type: accuracy |
| value: 0.0 |
| --- |
| |
| # 馃 Antonio BFCL Toolmodel |
|
|
| Este modelo forma parte del ecosistema **BitAgent (Subnet-20)** de Bittensor, dise帽ado para tareas de *tool-calling*, razonamiento l贸gico estructurado y generaci贸n de texto contextual. |
| Optimizado para comunicaci贸n eficiente entre agentes dentro del protocolo Finney. |
|
|
| --- |
|
|
| ## 馃殌 Descripci贸n t茅cnica |
|
|
| **antonio-bfcl-toolmodel** est谩 basado en un modelo open-source tipo `xLAM-7b-r`, ajustado para: |
|
|
| - 馃搳 *Razonamiento simb贸lico y factual multiling眉e* |
| - 馃З *Tool-calling autom谩tico* (formato JSON conforme a los prompts de Subnet-20) |
| - 馃攧 *Respuestas deterministas* con `temperature=0.1` y `top_p=0.9` |
| - 鈿欙笍 *Compatibilidad total con el pipeline de BitAgent Miner (v1.0.51)* |
| - 馃寪 *Idiomas soportados*: Espa帽ol 馃嚜馃嚫 e Ingl茅s 馃嚞馃嚙 |
|
|
| --- |
|
|
| ## 馃З Integraci贸n con Subnet-20 |
|
|
| Los validadores pueden invocar este modelo a trav茅s de los protocolos: |
|
|
| - `QueryTask` |
| - `QueryResult` |
| - `IsAlive` |
| - `GetHFModelName` |
| - `SetHFModelName` |
|
|
| El modelo responde mediante `bittensor.dendrite` y cumple con la especificaci贸n **BitAgent v1.0.51**. |
|
|
| --- |
|
|
| ## 馃 Ejemplo de inferencia local |
|
|
| ```python |
| from transformers import AutoTokenizer, AutoModelForCausalLM |
| |
| model_name = "Tonit23/antonio-bfcl-toolmodel" |
| tokenizer = AutoTokenizer.from_pretrained(model_name) |
| model = AutoModelForCausalLM.from_pretrained(model_name) |
| |
| prompt = "Resuelve esta operaci贸n: 12 + 37 = " |
| inputs = tokenizer(prompt, return_tensors="pt") |
| outputs = model.generate(**inputs, max_new_tokens=32) |
| print(tokenizer.decode(outputs[0], skip_special_tokens=True)) |
| |