michaelfeil
commited on
Commit
·
48bd069
1
Parent(s):
3b26078
Upload togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1 ctranslate fp16 weights
Browse files
README.md
CHANGED
@@ -34,18 +34,18 @@ inference:
|
|
34 |
max_new_tokens: 128
|
35 |
---
|
36 |
# # Fast-Inference with Ctranslate2
|
37 |
-
Speedup inference by 2x-
|
38 |
|
39 |
quantized version of [togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1](https://huggingface.co/togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1)
|
40 |
```bash
|
41 |
-
pip install hf-hub-ctranslate2>=2.0.6
|
42 |
```
|
43 |
Converted on 2023-05-19 using
|
44 |
```
|
45 |
-
ct2-transformers-converter --model togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1 --output_dir /home/
|
46 |
```
|
47 |
|
48 |
-
Checkpoint compatible to [ctranslate2](https://github.com/OpenNMT/CTranslate2) and [hf-hub-ctranslate2](https://github.com/michaelfeil/hf-hub-ctranslate2)
|
49 |
- `compute_type=int8_float16` for `device="cuda"`
|
50 |
- `compute_type=int8` for `device="cpu"`
|
51 |
|
@@ -63,7 +63,7 @@ model = GeneratorCT2fromHfHub(
|
|
63 |
tokenizer=AutoTokenizer.from_pretrained("togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1")
|
64 |
)
|
65 |
outputs = model.generate(
|
66 |
-
text=["How do you call a fast Flan-ingo?", "User: How are you doing?"],
|
67 |
)
|
68 |
print(outputs)
|
69 |
```
|
|
|
34 |
max_new_tokens: 128
|
35 |
---
|
36 |
# # Fast-Inference with Ctranslate2
|
37 |
+
Speedup inference while reducing memory by 2x-4x using int8 inference in C++ on CPU or GPU.
|
38 |
|
39 |
quantized version of [togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1](https://huggingface.co/togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1)
|
40 |
```bash
|
41 |
+
pip install hf-hub-ctranslate2>=2.0.6
|
42 |
```
|
43 |
Converted on 2023-05-19 using
|
44 |
```
|
45 |
+
ct2-transformers-converter --model togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1 --output_dir /home/feil_m/tmp-ct2fast-RedPajama-INCITE-Instruct-7B-v0.1 --force --copy_files tokenizer.json README.md tokenizer_config.json generation_config.json special_tokens_map.json .gitattributes --quantization float16
|
46 |
```
|
47 |
|
48 |
+
Checkpoint compatible to [ctranslate2>=3.13.0](https://github.com/OpenNMT/CTranslate2) and [hf-hub-ctranslate2>=2.0.6](https://github.com/michaelfeil/hf-hub-ctranslate2)
|
49 |
- `compute_type=int8_float16` for `device="cuda"`
|
50 |
- `compute_type=int8` for `device="cpu"`
|
51 |
|
|
|
63 |
tokenizer=AutoTokenizer.from_pretrained("togethercomputer/RedPajama-INCITE-Instruct-7B-v0.1")
|
64 |
)
|
65 |
outputs = model.generate(
|
66 |
+
text=["How do you call a fast Flan-ingo?", "User: How are you doing? Bot:"],
|
67 |
)
|
68 |
print(outputs)
|
69 |
```
|