1.87B dense Indic base model (English + 22 Indian languages, custom 128K Indic tokenizer).
Checkpoint at
~step 10110
(~40-60B tokens), trained on ~83% Indic + English mix.
Per-language bits-per-byte (held-out): Hindi 0.96, Tamil 0.97, Bengali 1.03, Kannada 1.05.
Resume: build AZMXForCausalLM from config, load latest.pt state_dict, continue from its step
against the dataset shards. Target: 1.8T tokens (WSD schedule, 457763 steps).
Runs of Drew168 azmx-one-v1 on huggingface.co
2.7K
Total runs
0
24-hour runs
1.9K
3-day runs
2.6K
7-day runs
2.7K
30-day runs
More Information About azmx-one-v1 huggingface.co Model
azmx-one-v1 huggingface.co is an AI model on huggingface.co that provides azmx-one-v1's model effect (), which can be used instantly with this Drew168 azmx-one-v1 model. huggingface.co supports a free trial of the azmx-one-v1 model, and also provides paid use of the azmx-one-v1. Support call azmx-one-v1 model through api, including Node.js, Python, http.
azmx-one-v1 huggingface.co is an online trial and call api platform, which integrates azmx-one-v1's modeling effects, including api services, and provides a free online trial of azmx-one-v1, you can try azmx-one-v1 online for free by clicking the link below.
Drew168 azmx-one-v1 online free url in huggingface.co:
azmx-one-v1 is an open source model from GitHub that offers a free installation service, and any user can find azmx-one-v1 on GitHub to install. At the same time, huggingface.co provides the effect of azmx-one-v1 install, users can directly use azmx-one-v1 installed effect in huggingface.co for debugging and trial. It also supports api for free installation.