Add to the readmes for stable-lm. (#1047)

README.md

@@ -62,6 +62,8 @@ We also provide a some command line based examples using state of the art models
 - [LLaMA and LLaMA-v2](./candle-examples/examples/llama/): general LLM.
 - [Falcon](./candle-examples/examples/falcon/): general LLM.
 - [Phi-v1.5](./candle-examples/examples/phi/): a 1.3b general LLM with performance on par with LLaMA-v2 7b.
+- [StableLM-3B-4E1T](./candle-examples/examples/stable-lm/): a 3b general LLM
+  pre-trained on 1T tokens of English and code datasets.
 - [Mistral7b-v0.1](./candle-examples/examples/mistral/): a 7b general LLM with
   performance larger than all publicly available 13b models as of 2023-09-28.
 - [StarCoder](./candle-examples/examples/bigcode/): LLM specialized to code generation.
@@ -152,6 +154,7 @@ If you have an addition to this list, please submit a pull request.
         - StarCoder.
         - Phi v1.5.
         - Mistral 7b v0.1.
+        - StableLM-3B-4E1T.
         - T5.
         - Bert.
     - Whisper (multi-lingual support).

candle-examples/examples/stable-lm/README.md

@@ -0,0 +1,25 @@
+# candle-stable-lm
+
+StableLM-3B-4E1T is a 3 billion parameter decoder-only language model
+pre-trained on 1 trillion tokens of diverse English and code datasets for 4
+epochs. See the [HuggingFace Hub Model
+Card](https://huggingface.co/stabilityai/stablelm-3b-4e1t).
+
+Note that this model is gated so you will have to request access on the Hub in
+order to be able to use it.
+
+## Running some example
+
+```bash
+$ cargo run --example stable-lm --release --features cuda -- --prompt 'What is the most efficient programming language in use?' --sample-len 150
+avx: true, neon: false, simd128: false, f16c: true
+temp: 0.00 repeat-penalty: 1.10 repeat-last-n: 64
+retrieved the files in 126.593µs
+loaded the model in 3.474148965s
+What is the most efficient programming language in use?
+The answer to this question depends on what you mean by "efficient". If you're talking about speed, then C++ and Java are probably your best bets. But if you're talking about ease of development, then Python is probably the way to go.
+Python is a high-level, interpreted language that is easy to learn and use. It has a large community of developers who are always working on new features and improvements.
+C++ is a low-level, compiled language that can be used for both desktop applications and web development. It's more difficult to learn than Python but offers greater control over the code.
+Java is another high-level language that is popular with programmers because it runs on many different platforms (including Android phones
+150 tokens generated (37.61 token/s)
+```