Update README.md

README.md

@@ -1,3 +1,16 @@
+LlamaDos is a model oriented to have conversations in Spanish. It results from a finetuning of the Llama2-7b model by Meta using various optimization techniques such as LoRa, quantization, gradient accumulation and much more.
+
+This has allowed the training to be performed on a single consumer graph (RTX 3090). More specifically, more than 250,000 conversational data were used and the training took approximately 140 hours.
+
+The training has been performed following the original data structure of the Llama2 paper, so it is recommended to follow the same structure for inference:
+
+<s>[INST] <<SYS>>
+{{ You are a helpful, respectful and honest conversational assistant. Have a conversation with the user in a natural way.  Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature. }}
+<</SYS>>
+
+{{ user_msg_1 }} [/INST] {{ model_answer_1 }} </s><s>[INST] {{ user_msg_2 }} [/INST]
+
+
 This work is funded by the Comunidad de Madrid through the call Research Grants for Young Investigators from Universidad Politécnica de Madrid (GENIUS:APOYO-JOVENES-21-TAXTYC-32-K61X37), and supported by the following projects: European Commission through Project ASTOUND (101071191–-HORIZON-EIC-2021-PATHFINDERCHALLENGES-01) and BEWORD (PID2021-126061OB-C43) funded by
 MCIN/AEI/10.13039/501100011033 and, as appropriate, by “ERDF A way of making Europe”, by the
 “European Union”.