tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

Update 4bit notes

Browse Source
This commit is contained in:
NanoCode012
2023-05-25 23:48:18 +09:00
parent f92245dbd6
commit a9e502ef45
1 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
README.md
View File

@@ -136,7 +136,7 @@ See sample configs in [configs](configs) folder or [examples](examples) for quic
- loading - loading
```yaml ```yaml
load_in_8bit: true load_in_4bit: true
load_in_8bit: true load_in_8bit: true
bf16: true # require >=ampere bf16: true # require >=ampere
fp16: true fp16: true
@@ -175,13 +175,15 @@ tokenizer_type: AutoTokenizer
# Trust remote code for untrusted source # Trust remote code for untrusted source
trust_remote_code: trust_remote_code:
# whether you are training a 4-bit quantized model # whether you are training a 4-bit GPTQ quantized model
load_4bit: true load_4bit: true
gptq_groupsize: 128 # group size gptq_groupsize: 128 # group size
gptq_model_v1: false # v1 or v2 gptq_model_v1: false # v1 or v2
# this will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer # this will attempt to quantize the model down to 8 bits and use adam 8 bit optimizer
load_in_8bit: true load_in_8bit: true
# use bitsandbytes 4 bit
load_in_4bit:
# Use CUDA bf16 # Use CUDA bf16
bf16: true # bool or 'full' for `bf16_full_eval`. require >=ampere bf16: true # bool or 'full' for `bf16_full_eval`. require >=ampere