rishanthrajendhran commited on
Commit
e047d1e
·
verified ·
1 Parent(s): 1381383

Card: run without a GPU on Tinker

Browse files
Files changed (1) hide show
  1. README.md +16 -0
README.md CHANGED
@@ -57,6 +57,22 @@ To compare with idea-level detection on the same documents, `idealens run docs.j
57
  extracts their outlines and scores them with [IdeaLens](https://huggingface.co/rishanthrajendhran/IdeaLens). The rest
58
  of this section runs the model directly.
59
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
60
  ### Load the merged model (66 GB download)
61
 
62
  ```python
 
57
  extracts their outlines and scores them with [IdeaLens](https://huggingface.co/rishanthrajendhran/IdeaLens). The rest
58
  of this section runs the model directly.
59
 
60
+ ### Without a GPU: Tinker
61
+
62
+ ProseLens is also published on [Tinker](https://thinkingmachines.ai/tinker/), which runs it on its own servers, so any
63
+ laptop works. Sign up at [auth.thinkingmachines.ai/sign-up](https://auth.thinkingmachines.ai/sign-up), create an API key in
64
+ the Tinker console, and run:
65
+
66
+ ```bash
67
+ pip install "idealens[tinker]"
68
+ export TINKER_API_KEY=...
69
+ idealens score docs.jsonl -o scores.jsonl --model ProseLens --backend tinker
70
+ ```
71
+
72
+ Tinker bills your account per token: under $1 per 1,000 typical web documents at list price
73
+ (`--dry-run` gives an estimate). The checkpoint is `tinker://f21bf499-021e-5735-85f5-f5e1183aad78:train:0/sampler_weights/nemo_docs_v1m_docs_r64_ep1_final`; on our checks its scores
74
+ agree with the weights in this repo to about 0.002 in P(human).
75
+
76
  ### Load the merged model (66 GB download)
77
 
78
  ```python