# BLOOM outputs only few tokens

**URL:** </static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbloom-outputs-only-few-tokens%2F27306%3E%5C
**Category:** Inference Endpoints on the Hub\
**Created:** [December 6, 2022, 1:17am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbloom-outputs-only-few-tokens%2F27306 "2022-12-06T01:17:36Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![jjkirshbaum](https://avatars.discourse-cdn.com/v4/letter/j/74df32/32.png) [@jjkirshbaum](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fjjkirshbaum)%5C
**Post date:** [December 6, 2022, 1:17am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbloom-outputs-only-few-tokens%2F27306%2F1 "2022-12-06T01:17:36Z")

</div>

Hi, I am trying to inference on BLOOM using Inference API. It only produces a few tokens (max 1-3 sentences), even if I set ‘min\_length’ very high, for instance “min\_length”: 1024. How can I generate more tokens with BLOOM? Is there a limitation around this, or am I entering the parameters incorrectly?

---

<div class="post-metadata">

**Author:** ![jjkirshbaum](https://avatars.discourse-cdn.com/v4/letter/j/74df32/32.png) [@jjkirshbaum](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Fu%2Fjjkirshbaum)%5C
**Post date:** [December 6, 2022, 1:25am UTC](/static-proxy?url=https%3A%2F%2Fdiscuss.huggingface.co%2Ft%2Fbloom-outputs-only-few-tokens%2F27306%2F2 "2022-12-06T01:25:21Z")

</div>

This is what my entry looks like:

API\_URL = “[/static-proxy?url=https%3A%2F%2Fapi-inference.huggingface.co%2Fmodels%2Fbigscience%2Fbloom%5D(https%3A%2F%2Fapi-inference.huggingface.co%2Fmodels%2Fbigscience%2Fbloom)%E2%80%9D

HF\_TOKEN = ‘api\_org\_XXXXXXX’

headers = {“Authorization”: f"Bearer {HF\_TOKEN}"}

prompt\_text = “”“Text:  
Some prompt text is here because”“”

json\_ = {“inputs”: prompt\_text,  
“parameters”:  
{  
“top\_p”: 1.0,  
“temperature”: .72,  
“min\_length”: 1024,  
“return\_full\_text”: False  
}, “options”:  
{  
“use\_cache”: False,  
“wait\_for\_model”:True  
},}  
response = requests.post(API\_URL, headers=headers, json=json\_)  
output = response.json()