Iker
/

TowerInstruct-7B-v0.2-EN2ES

@@ -1,199 +1,133 @@
 ---
 library_name: transformers
-tags: []
 ---
-# Model Card for Model ID
-<!-- Provide a quick summary of what the model is/does. -->
-## Model Details
-### Model Description
-<!-- Provide a longer summary of what this model is. -->
-This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
-- **Developed by:** [More Information Needed]
-- **Funded by [optional]:** [More Information Needed]
-- **Shared by [optional]:** [More Information Needed]
-- **Model type:** [More Information Needed]
-- **Language(s) (NLP):** [More Information Needed]
-- **License:** [More Information Needed]
-- **Finetuned from model [optional]:** [More Information Needed]
-### Model Sources [optional]
-<!-- Provide the basic links for the model. -->
-- **Repository:** [More Information Needed]
-- **Paper [optional]:** [More Information Needed]
-- **Demo [optional]:** [More Information Needed]
-## Uses
-<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
-### Direct Use
-<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
-[More Information Needed]
-### Downstream Use [optional]
-<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
-[More Information Needed]
-### Out-of-Scope Use
-<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
-[More Information Needed]
-## Bias, Risks, and Limitations
-<!-- This section is meant to convey both technical and sociotechnical limitations. -->
-[More Information Needed]
-### Recommendations
-<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
-Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
-## How to Get Started with the Model
-Use the code below to get started with the model.
-[More Information Needed]
-## Training Details
-### Training Data
-<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
-[More Information Needed]
-### Training Procedure
-<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
-#### Preprocessing [optional]
-[More Information Needed]
-#### Training Hyperparameters
-- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
-#### Speeds, Sizes, Times [optional]
-<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
-[More Information Needed]
-## Evaluation
-<!-- This section describes the evaluation protocols and provides the results. -->
-### Testing Data, Factors & Metrics
-#### Testing Data
-<!-- This should link to a Dataset Card if possible. -->
-[More Information Needed]
-#### Factors
-<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
-[More Information Needed]
-#### Metrics
-<!-- These are the evaluation metrics being used, ideally with a description of why. -->
-[More Information Needed]
-### Results
-[More Information Needed]
-#### Summary
-## Model Examination [optional]
-<!-- Relevant interpretability work for the model goes here -->
-[More Information Needed]
-## Environmental Impact
-<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
-Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
-- **Hardware Type:** [More Information Needed]
-- **Hours used:** [More Information Needed]
-- **Cloud Provider:** [More Information Needed]
-- **Compute Region:** [More Information Needed]
-- **Carbon Emitted:** [More Information Needed]
-## Technical Specifications [optional]
-### Model Architecture and Objective
-[More Information Needed]
-### Compute Infrastructure
-[More Information Needed]
-#### Hardware
-[More Information Needed]
-#### Software
-[More Information Needed]
-## Citation [optional]
-<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
-**BibTeX:**
-[More Information Needed]
-**APA:**
-[More Information Needed]
-## Glossary [optional]
-<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
-[More Information Needed]
-## More Information [optional]
-[More Information Needed]
-## Model Card Authors [optional]
-[More Information Needed]
-## Model Card Contact
-[More Information Needed]

 ---
 library_name: transformers
+base_model: Unbabel/TowerInstruct-7B-v0.2
+license: apache-2.0
+datasets:
+- Iker/InstructTranslation-EN-ES
+language:
+- en
+- es
+pipeline_tag: text-generation
+tags:
+  - translation
 ---
+This is a [TowerInstruct-7B](https://huggingface.co/Unbabel/TowerInstruct-7B-v0.2) model fine-tuned for translating instructions datasets from English into Spanish. This model has GPT4 translation quality, but you can run it on your own machine for free 🎉
+The model has been finetuned using ~1.500 prompts and answers from [teknium/OpenHermes-2.5](teknium/OpenHermes-2.5) translated to Spanish using GPT-4-0125-preview. The dataset is available here: https://huggingface.co/datasets/Iker/InstructTranslation-EN-ES/
+# Demo
+```python
+import torch
+from transformers import pipeline
+og = pipeline("text-generation", model="Unbabel/TowerInstruct-13B-v0.1", torch_dtype=torch.bfloat16, device_map=0)
+fn7 = pipeline("text-generation", model="Iker/TowerInstruct-7B-v0.2-EN2ES", torch_dtype=torch.bfloat16, device_map=1)
+fn = pipeline("text-generation", model="Iker/TowerInstruct-13B-v0.1-EN2ES", torch_dtype=torch.bfloat16, device_map=2)
+msg = """
+Let's use Bayes' theorem again to solve this problem:\n\nLet A represent the event that the man actually has the ability to predict dice rolls with 90% accuracy, and C represent the event of predicting correctly on the first attempt.\n\nWe want to find P(A|C), the probability that the man actually has the ability given that he predicted correctly on his first attempt.\n\nBayes' theorem states that P(A|C) = P(C|A) * P(A) / P(C)\n\nFirst, let's find P(C|A): the probability of predicting correctly on the first attempt if the man actually has the ability. Since he claims 90% accuracy, this probability is 0.9.\n\nNext, let's find P(A): the probability that someone actually has the ability to predict dice rolls with 90% accuracy. We are told this is 1%, so P(A) = 0.01.\n\nNow we need to find P(C): the overall probability of predicting correctly on the first attempt. This can be calculated as the sum of probabilities for each case: P(C) = P(C|A) * P(A) + P(C|¬A) * P(¬A), where ¬A represents not having the ability and P(¬A) = 1 - P(A) = 0.99.\n\nTo find P(C|¬A), the probability of predicting correctly on the first attempt without the ability, we use the fact that there's a 1/6 chance of guessing correctly by random chance: P(C|¬A) = 1/6.\n\nSo, P(C) = (0.9)*(0.01) + (1/6)*(0.99) = 0.009 + 0.165 = 0.174.\n\nFinally, we can calculate P(A|C) using Bayes' theorem:\n\nP(A|C) = P(C|A) * P(A) / P(C) = (0.9)*(0.01) / (0.174) ≈ 0.0517.\n\nTherefore, the probability that the man actually has the ability to predict dice rolls with 90% accuracy is approximately 5.17%.
+""".strip()
+messages = [
+    {"role": "user",
+     "content": f"Translate the following text from English into Spanish.\n{msg}\nSpanish:"},
+]
+prompt = og.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+outputs = og(prompt, max_new_tokens=1024, do_sample=False)
+print(outputs[0]["generated_text"])
+prompt = fn7.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+outputs = fn7(prompt, max_new_tokens=1024, do_sample=False)
+print(outputs[0]["generated_text"])
+prompt = fn.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+outputs = fn(prompt, max_new_tokens=1024, do_sample=False)
+print(outputs[0]["generated_text"])
+```
+### Unbabel/TowerInstruct-13B-v0.1
+This model fails with very large inputs, the answer is in Russian instead of Spanish 🥴
+```
+ Воспроизвем теорему Байеса для решения этой задачи:
+Предположим, что A означает событие, когда человек действительно обладает способностью предсказывать результаты броска кубика с то��ностью 90%, а C означает событие правильного предсказания на первой попытке.
+Мы хотим найти P(A|C), вероятность того, что у человека действительно есть способность, если он правильно предсказал на первой попытке.
+Теорема Байеса утверждает, что P(A|C) = P(C|A) * P(A) / P(C)
+Сначала определим P(C|A): вероятность правильного предсказания на первой попытке, если у человека действительно есть способность. Поскольку он утверждает, что его точность составляет 90%, эта вероятность равна 0,9.
+Далее определим P(A): вероятность того, что у кого-то действительно есть способность предсказывать результаты броска кубика с точностью 90%. Нам сказано, что она составляет 1%, поэтому P(A) = 0,01.
+Теперь нам нужно найти P(C): общую вероятность правильного предсказания на первой попытке. Это можно рассчитать как сумму вероятностей для каждого случая: P(C) = P(C|A) * P(A) + P(C|¬A) * P(¬A), где ¬A означает отсутствие способности, и P(¬A) = 1 - P(A) = 0,99.
+Чтобы найти P(C|¬A), вероятность правильного предсказания на первой попытке без способности, мы используем тот факт, что существует 1/6 шанс сделать правильный выбор случайно: P(C|¬A) = 1/6.
+Таким образом, P(C) = (0,9)*(0,01) / (0,174) ≈ 0,009 + 0,165 = 0,174.
+Наконец, мы можем рассчитать P(A|C) с помощью теоремы Байеса:
+P(A|C) = P(C|A) * P(A) / P(C) = (0,9)*(0,01) / (0,174) ≈ 0,0517.
+Таким образом, вероятность того, что у человека действительно есть способность предсказывать результаты броска кубика с точностью 90%, составляет примерно 5,17%.
+```
+### Iker/TowerInstruct-7B-v0.2-EN2ES
+```
+Vamos a usar de nuevo el teorema de Bayes para resolver este problema:
+A representa el evento de que el hombre realmente tenga la capacidad de predecir lanzamientos de dados con un 90% de precisión, y C representa el evento de predecir correctamente en el primer intento.
+Queremos encontrar P(A|C), la probabilidad de que el hombre realmente tenga la capacidad dado que predecía correctamente en su primer intento.
+El teorema de Bayes establece que P(A|C) = P(C|A) * P(A) / P(C)
+Primero, vamos a encontrar P(C|A): la probabilidad de predecir correctamente en el primer intento si el hombre realmente tiene la capacidad. Dado que afirma un 90% de precisión, esta probabilidad es 0.9.
+A continuación, vamos a encontrar P(A): la probabilidad de que alguien realmente tenga la capacidad de predecir lanzamientos de dados con un 90% de precisión. Nos dicen que esto es del 1%, así que P(A) = 0.01.
+Ahora necesitamos encontrar P(C): la probabilidad total de predecir correctamente en el primer intento. Esto se puede calcular como la suma de probabilidades para cada caso: P(C) = P(C|A) * P(A) + P(C|¬A) * P(¬A), donde ¬A representa no tener la capacidad y P(¬A) = 1 - P(A) = 0.99.
+Para encontrar P(C|¬A), la probabilidad de predecir correctamente en el primer intento sin la capacidad, usamos el hecho de que hay una probabilidad del 1/6 de adivinar correctamente por azar: P(C|¬A) = 1/6.
+Así, P(C) = (0.9)*(0.01) + (1/6)*(0.99) = 0.009 + 0.165 = 0.174.
+Finalmente, podemos calcular P(A|C) usando el teorema de Bayes:
+P(A|C) = P(C|A) * P(A) / P(C) = (0.9)*(0.01) / (0.174) ≈ 0.0517.
+Por lo tanto, la probabilidad de que el hombre realmente tenga la capacidad de predecir lanzamientos de dados con un 90% de precisión es aproximadamente del 5.17%.
+```
+### Iker/TowerInstruct-13B-v0.1-EN2ES
+```
+Vamos a usar de nuevo el teorema de Bayes para resolver este problema:
+ Sea A el evento de que el hombre realmente tenga la capacidad de predecir lanzamientos de dados con un 90% de precisión, y C el evento de predecir correctamente en el primer intento.
+ Queremos encontrar P(A|C), la probabilidad de que el hombre realmente tenga la capacidad dada que predijo correctamente en su primer intento.
+ El teorema de Bayes establece que P(A|C) = P(C|A) * P(A) / P(C)
+ Primero, vamos a encontrar P(C|A): la probabilidad de predecir correctamente en el primer intento si el hombre realmente tiene la capacidad. Dado que afirma un 90% de precisión, esta probabilidad es 0.9.
+ A continuación, vamos a encontrar P(A): la probabilidad de que alguien realmente tenga la capacidad de predecir lanzamientos de dados con un 90% de precisión. Se nos dice que este es 1%, así que P(A) = 0.01.
+ Ahora necesitamos encontrar P(C): la probabilidad general de predecir correctamente en el primer intento. Esto puede calcularse como la suma de probabilidades para cada caso: P(C) = P(C|A) * P(A) + P(C|¬A) * P(¬A), donde ¬A representa no tener la capacidad y P(¬A) = 1 - P(A) = 0.99.
+ Para encontrar P(C|¬A), la probabilidad de predecir correctamente en el primer intento sin la capacidad, utilizamos el hecho de que hay una probabilidad de 1/6 de adivinar correctamente por casualidad: P(C|¬A) = 1/6.
+ Así que, P(C) = (0.9)*(0.01) + (1/6)*(0.99) = 0.009 + 0.165 = 0.174.
+ Finalmente, podemos calcular P(A|C) usando el teorema de Bayes:
+ P(A|C) = P(C|A) * P(A) / P(C) = (0.9)*(0.01) / (0.174) ≈ 0.0517.
+ Por lo tanto, la probabilidad de que el hombre realmente tenga la capacidad de predecir lanzamientos de dados con un 90% de precisión es aproximadamente 5.17%.
+```