building at github.com/ItshMoh
Joined September 2023
- Tweets 252
- Following 447
- Followers 48
- Likes 582
63 Photos and videos
13h
Here is the link to try it - huggingface.co/spaces/build-…
I have deployed it on the hugging face spaces. I was building this for the build-small-hackathon by @huggingface and @Gradio .
13h
Done with the finetuning of the Qwen2.5-Coder-7B-Instruct. I finetuned it for generating excalidraw images. I generated around 1300 samples of excali DSL and then trained the model to produce DSL which is later to be converted into excali json by the converter.
3
657
13h
Done with the finetuning of the Qwen2.5-Coder-7B-Instruct. I finetuned it for generating excalidraw images. I generated around 1300 samples of excali DSL and then trained the model to produce DSL which is later to be converted into excali json by the converter.
1
683
Jun 14
Deepseek v4-flash is too cheap for its quality. More than 6 M token and just 1$. It is just crazy good for most of the use cases and the price makes it more lit.
1
20
Jun 14
Generating dataset is a nightmare.
It had stopped after genarating 560 samples. The new samples generated are all invalid according to my converter. This can't be hallucinations.
1
4
Jun 14
Here comes the data generation part.
Using deepseek -v4- flash for generating approx 1300 samples.
These samples are JSON DSL for the excalidraw. Then that DSL will be converted into excalidraw JSON by the converter that i have designed.
Jun 13
Trying to finetune qwen model to generate the excalidraw images better than claude.
Currently generating dataset and writing the validators.
I will probably finetune the 7b coder one as it is good in producing structured outputs.
1
27
Jun 13
Trying to finetune qwen model to generate the excalidraw images better than claude.
Currently generating dataset and writing the validators.
I will probably finetune the 7b coder one as it is good in producing structured outputs.
1
48
Jun 7
Dataset generation is the most frustrating part in the finetuning of model. Good dataset generation will consume your whole mental energy.
1
5
Jun 1
If anyone wants to run their resume. Here is the link github.com/interviewstreet/h…
Jun 1
Ran my resume against the Hacker rank ATS . It gives me 106/100. I got shocked but later on debug I found that it gives bonus also which is added to the total score so max cap becomes 120. I got 16 bonus and scored full in everything except the open source where I got 25/35.
2
19
Jun 1
Ran my resume against the Hacker rank ATS . It gives me 106/100. I got shocked but later on debug I found that it gives bonus also which is added to the total score so max cap becomes 120. I got 16 bonus and scored full in everything except the open source where I got 25/35.
2
32
May 23
Have you ever used chrome://tracing? I have used it today for the first time, for visualising the onnx runtime profiler json for piper model. This tracing helps to generate flamechat from json if given in a specific format(catapult format) .
1
12
May 23
Studying about audio data and their features. Then I tried to generate mel spectrogram and then inverting it with the griffin-lim to get the waveform.
github.com/ItshMoh/rubic
21
May 21
I usually forgot some of the important tasks that i have to do in a day, like filling form of the thesis exam etc. Some small non professional task. I made a reminder agent for myself , which message me on telegram for these chores. I named it Doraemon.
1
22
May 20
I was not aware this type of import actually had a name. I thought people just forgot to import in the begining and did it inside the function like a lazy person. ironically it is called lazy import.
1
15
Apr 20
Now I will be reading this paper to make a presentation for my Computer Vision Course. I will try to write a blog about it.
2
26
Mohan retweeted
Apr 19
“Codex Hackathons have such great builder energy”,
and Bangalore - you 10x-ed it. 🔥
To all the builders, thank you for joining us for a day of jamming with Codex.
6
12
138
7,210
Apr 19
Going back to Varanasi. It's been an amazing journey in Bangalore.
Endsem exam is coming. I need to ask notes from maglu
1
22