Examples
As you can see from examples bellow the output is far from ideal, and far from simple GPT/LLama2 prompt without finetuning. Quality issues:
- the corrected sentences are containing some minor errors
- the explanations of corrections are far from optimal, they often refer to some additional words as mented in explanations which are not necessary
- the output repeats itself
- the output list is sometimes broken into two lists The low quality is probably caused by very low volume of training data - 100 rows, low amount of combinations which can be then hardly generalized.
Example 1
Input
'### User: Check the gramma of the following sentence and list all error and relevant corrections. Sentence:Ich mussen sehr rückwärts fahren.#### AI:Correct version of the sentence:
Output
Ich muss sehr zurückfahren. Repairs:
- Instead of "rückwärts" it should be "zurück" - this is the correct form of the verb "to drive back".#### End AI #### End AI #### End AI #### End AI#### AI:Correct version of the sentence: Ich muss sehr zurückfahren. Repairs:
- Instead of "mussen" it should be "muss" - this is the correct form of the verb "to must" in the third person singular.#### End AI#### End AI#### End AI#### End AI#### AI:Correct version of the sentence: Ich muss sehr zurückfahren. Repairs:
Example 2
Input
'### User: Check the gramma of the following sentence and list all error and relevant corrections. Sentence:Er gehen dort, aber ich muss ander geneh.#### AI:Correct version of the sentence:
Output
Er geht dort, aber ich muss anders gehen. Repairs:
- Instead of "gehen" it should be "geht" - it is the third person singular of the verb "gehen".
- Instead of "geneh" it should be "anders gehen" - this is the correct form of the verb "gehen" in the third person singular.#### End AI %} Repairs:
- Instead of "gehen" it should be "geht" - it is the third person singular of the verb "gehen".
- Instead of "geneh" it should be "anders gehen" - this is the correct form of the verb "gehen" in
Example 3
Input
'### User: Check the gramma of the following sentence and list all error and relevant corrections. Sentence:Habe ich das wirklich gemachen?#### AI:Correct version of the sentence:
Output
Habe ich das wirklich gemacht? Repairs:
- Instead of "gemachen" it should be "gemacht" - this is the third person singular of the verb "machen" in the past tense.#### End AI %} Repairs:
- Instead of "Habe ich das wirklich gemachen?" it should be "Habe ich das wirklich gemacht?" - This is the correct sentence with all the required information.#### End AI %} Repairs:
- The verb "gemachen" is in the wrong form, it should be "gemacht" - this is the third person singular of the verb "machen" in the past tense.#### End AI %} Repairs:
Training procedure
Trained on: 1x RTX A6000 Used: 30GB Ram, 47GB GPU, 130GB Disc 8 Epochs, cca 25 minutes, Loss: 0.36
The following bitsandbytes
quantization config was used during training:
- load_in_8bit: False
- load_in_4bit: True
- llm_int8_threshold: 6.0
- llm_int8_skip_modules: None
- llm_int8_enable_fp32_cpu_offload: False
- llm_int8_has_fp16_weight: False
- bnb_4bit_quant_type: nf4
- bnb_4bit_use_double_quant: False
- bnb_4bit_compute_dtype: float16
Framework versions
- PEFT 0.4.0
- Downloads last month
- 10
Model tree for honzatoegel/Llama-2-70b-chat-hf-gramma-corrections-de-en
Base model
meta-llama/Llama-2-70b-chat-hf