Comments (2)
@feiyu87yeah: Let me look into this.
The Natural Instructions v2 submixture compiles a very wide range of tasks, a few of which are language translation/identification, and cover tens of languages. It looks like some of these may have encoding issues. This comprises a small minority of examples in the overall mixture, but thank you for bringing it to our attention -- I'll take a look.
from flan.
There are likely problems in niv2_zs_noopt_train.jsonl.gz
I do not know where the problem is. Might be from the FLAN v2 codebase, or from how way niv2_zs_noopt_train.jsonl.gz is prepared.
For this particular case, what I see on my side has the right language.
The raw string is:
"Statement: \u0d08\u0d38\u0d4d\u0d31\u0d4d\u0d31\u0d47\u0d7a \u0d15\u0d3e\u0d31\u0d4d\u0d31\u0d3f\u0d7d \u0d08\u0d17\u0d4d\u0d30\u0d31\u0d4d\u0d31\u0d4d (B. i. coromandus) - 1783 - \u0d7d \u0d2a\u0d40\u0d31\u0d4d\u0d31\u0d7c \u0d2c\u0d4a\u0d21\u0d47\u0d7c\u0d1f\u0d4d\u0d1f\u0d4d \u0d06\u0d23\u0d4d \u0d06\u0d26\u0d4d\u0d2f\u0d2e\u0d3e\u0d2f\u0d3f \u0d35\u0d3f\u0d35\u0d30\u0d23\u0d02 \u0d28\u0d32\u0d4d\u0d15\u0d3f\u0d2f\u0d24\u0d4d. \u0d0f\u0d37\u0d4d\u0d2f\u0d2f\u0d3f\u0d32\u0d46\u0d2f\u0d41\u0d02 \u0d06\u0d2b\u0d4d\u0d30\u0d3f\u0d15\u0d4d\u0d15\u0d2f\u0d3f\u0d32\u0d46\u0d2f\u0d41\u0d02 \u0d38\u0d19\u0d4d\u0d15\u0d30\u0d35\u0d7c\u0d17\u0d4d\u0d17\u0d02 \u0d06\u0d23\u0d4d. \u0d24\u0d46\u0d15\u0d4d\u0d15\u0d41\u0d15\u0d3f\u0d34\u0d15\u0d4d\u0d15\u0d7b \u0d0f\u0d37\u0d4d\u0d2f\u0d2f\u0d3f\u0d7d \u0d28\u0d3f\u0d28\u0d4d\u0d28\u0d4d , \u0d1c\u0d2a\u0d4d\u0d2a\u0d3e\u0d28\u0d3f\u0d32\u0d47\u0d2f\u0d4d\u0d15\u0d4d\u0d15\u0d41\u0d02 \u0d38\u0d1e\u0d4d\u0d1a\u0d30\u0d3f\u0d15\u0d4d\u0d15\u0d41\u0d28\u0d4d\u0d28\u0d41. \u0d06\u0d38\u0d4d\u0d1f\u0d4d\u0d30\u0d47\u0d32\u0d3f\u0d2f\u0d2f\u0d3f\u0d32\u0d47\u0d2f\u0d4d\u0d15\u0d4d\u0d15\u0d41\u0d02, \u0d28\u0d4d\u0d2f\u0d42\u0d38\u0d4d \u0d32\u0d3e\u0d28\u0d4d\u0d31\u0d3f\u0d32\u0d47\u0d2f\u0d4d\u0d15\u0d4d\u0d15\u0d41\u0d02 \u0d2d\u0d3e\u0d17\u0d3f\u0d15\u0d2e\u0d3e\u0d2f\u0d3f \u0d26\u0d47\u0d36\u0d3e\u0d1f\u0d28\u0d02 \u0d28\u0d1f\u0d24\u0d4d\u0d24\u0d41\u0d28\u0d4d\u0d28\u0d41.\n\n Option A: \u0d06\u0d38\u0d4d\u0d1f\u0d4d\u0d30\u0d47\u0d32\u0d3f\u0d2f\u0d2f\u0d3f\u0d32\u0d46\n\n Option B: \u0d2a\u0d3e\u0d15\u0d3f\u0d38\u0d4d\u0d24\u0d3e\u0d28\u0d3f\u0d32\u0d47\u0d2f\u0d4d\u0d15\u0d4d\u0d15\u0d41\u0d02\n\n Option C: \u0d15\u0d4b\u0d2e\u0d4b\u0d31\u0d4b\u0d38\u0d4d\n\n Option D: \u0d05\u0d7c\u0d1c\u0d28\u0d4d\u0d31\u0d40\u0d28\u0d2f\u0d3f\u0d32\u0d41\u0d02"
If printing it out:
Statement: ഈസ്റ്റേൺ കാറ്റിൽ ഈഗ്രറ്റ് (B. i. coromandus) - 1783 - ൽ പീറ്റർ ബൊഡേർട്ട് ആണ് ആദ്യമായി വിവരണം നല്കിയത്. ഏഷ്യയിലെയും ആഫ്രിക്കയിലെയും സങ്കരവർഗ്ഗം ആണ്. തെക്കുകിഴക്കൻ ഏഷ്യയിൽ നിന്ന് , ജപ്പാനിലേയ്ക്കും സഞ്ചരിക്കുന്നു. ആസ്ട്രേലിയയിലേയ്ക്കും, ന്യൂസ് ലാന്റിലേയ്ക്കും ഭാഗികമായി ദേശാടനം നടത്തുന്നു.
Option A: ആസ്ട്രേലിയയിലെ
Option B: പാകിസ്താനിലേയ്ക്കും
Option C: കോമോറോസ്
Option D: അർജന്റീനയിലും
from flan.
Related Issues (20)
- Datafile AESLC-master.zip meet non_matching_checksum_error
- Broken Dataset Links in FLAN Repository HOT 1
- Question about data source splits
- Trying to access `splits['train']` but `splits` is empty
- Reproducing the flan_v2 results of T5-xl HOT 3
- Code for mixing Enrico sets? HOT 2
- Thank you for your efforts. This is a great job!I have a problem. Why is there only 70 tasks in this FLAN2021_submix_original, and it seems that there are over 140 in the paper? HOT 1
- Could you share the training loss to improve reproducibility? HOT 4
- [Question] Training data formatting for other custom models
- Questions about the Flan Held-Out Tasks (pubmed_qa) and experiment details in the flan collection paper
- Doubt about the TARGET parameter in the dataset
- Issue generating flan2021 submix HOT 1
- Does anyone know which 282 tasks are used? HOT 1
- AESLC Checksum error HOT 4
- No train split in `adversarial_qa_dbert_answer_the_following_q_template_0to10_no_opt_x_shot`
- some Enrico COT entries seem odd HOT 1
- How many samples were used to train Flan-T5?
- Finetuning is definitely better than zero-shot, this is inevitable. So how should I understand the title of this paper?
- Where can I find the test mixture and its templates to reproduce the results in the paper?
- Links to Enrico Shippole FLAN dataset are missing HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from flan.