Update README.md
Browse files
README.md
CHANGED
|
@@ -1,8 +1,9 @@
|
|
|
|
|
| 1 |
license: apache-2.0
|
| 2 |
datasets:
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
language:
|
| 7 |
- es
|
| 8 |
- en
|
|
@@ -14,6 +15,7 @@ tags:
|
|
| 14 |
- lora
|
| 15 |
- gguf
|
| 16 |
- multilingual
|
|
|
|
| 17 |
|
| 18 |
eva-mistral-turdus-7b-spanish
|
| 19 |
Mistral 7b-based model fine-tuned in Spanish to add high quality Spanish text generation.
|
|
@@ -30,6 +32,7 @@ Quantized using llama.cpp in int4 Q4_0 and int8 Q8_0
|
|
| 30 |
|
| 31 |
Usage: any framework that uses GGUF format. For native llama.cpp I recommend int4 model quality is not much different from int8 and is almost 2x speed.
|
| 32 |
|
|
|
|
| 33 |
./main -m $MODEL -c 512 -b 1024 -n 256 --keep 48 \
|
| 34 |
--repeat_penalty 1.0 --color -i \
|
| 35 |
-r "Enrique:" -f prompts/eva.txt
|
|
@@ -46,4 +49,4 @@ llama_print_timings: sample time = 2,15 ms / 81 runs ( 0,03 m
|
|
| 46 |
llama_print_timings: prompt eval time = 2786,32 ms / 50 tokens ( 55,73 ms per token, 17,94 tokens per second)
|
| 47 |
llama_print_timings: eval time = 10806,26 ms / 80 runs ( 135,08 ms per token, 7,40 tokens per second)
|
| 48 |
llama_print_timings: total time = 49858,03 ms / 130 tokens
|
| 49 |
-
|
|
|
|
| 1 |
+
---
|
| 2 |
license: apache-2.0
|
| 3 |
datasets:
|
| 4 |
+
- ecastera/wiki_fisica
|
| 5 |
+
- ecastera/filosofia-es
|
| 6 |
+
- jtatman/espanol_dolly_alpaca_format_combined
|
| 7 |
language:
|
| 8 |
- es
|
| 9 |
- en
|
|
|
|
| 15 |
- lora
|
| 16 |
- gguf
|
| 17 |
- multilingual
|
| 18 |
+
---
|
| 19 |
|
| 20 |
eva-mistral-turdus-7b-spanish
|
| 21 |
Mistral 7b-based model fine-tuned in Spanish to add high quality Spanish text generation.
|
|
|
|
| 32 |
|
| 33 |
Usage: any framework that uses GGUF format. For native llama.cpp I recommend int4 model quality is not much different from int8 and is almost 2x speed.
|
| 34 |
|
| 35 |
+
```
|
| 36 |
./main -m $MODEL -c 512 -b 1024 -n 256 --keep 48 \
|
| 37 |
--repeat_penalty 1.0 --color -i \
|
| 38 |
-r "Enrique:" -f prompts/eva.txt
|
|
|
|
| 49 |
llama_print_timings: prompt eval time = 2786,32 ms / 50 tokens ( 55,73 ms per token, 17,94 tokens per second)
|
| 50 |
llama_print_timings: eval time = 10806,26 ms / 80 runs ( 135,08 ms per token, 7,40 tokens per second)
|
| 51 |
llama_print_timings: total time = 49858,03 ms / 130 tokens
|
| 52 |
+
```
|