Requirements and Limits
Directory structure
Your PEFT addon must contain three files:
adapter_config.json
- The Hugging Face adapter configuration file.adapter_model.bin
oradapter_model.safetensors
- The saved addon file.fireworks.json
- A Fireworks configuration file.
It may also contain an optional README.md
file which will be ignored during the upload process.
Limits
The following limits are applied by default. Contact us if you need these limits relaxed.
Supported base models
Currently, the following base models are supported:
- accounts/fireworks/models/starcoder2-15b
- accounts/fireworks/models/starcoder2-3b
- accounts/fireworks/models/starcoder2-7b
- accounts/fireworks/models/deepseek-coder-1b-base
- accounts/fireworks/models/deepseek-coder-7b-base
- accounts/fireworks/models/deepseek-coder-v2-lite-base
- accounts/fireworks/models/llama-v2-34b-code
- accounts/fireworks/models/codegemma-2b
- accounts/fireworks/models/codegemma-7b
- accounts/fireworks/models/llama-v2-13b
- accounts/fireworks/models/llama-v2-13b-chat
- accounts/fireworks/models/llama-v2-70b
- accounts/fireworks/models/llama-v2-70b-chat
- accounts/fireworks/models/llama-v2-7b
- accounts/fireworks/models/llama-v2-7b-chat
- accounts/fireworks/models/llama-v3-8b-instruct-hf
- accounts/fireworks/models/llama-v3-70b-instruct-hf
- accounts/fireworks/models/llama-v3p1-8b-instruct
- accounts/fireworks/models/llama-v3p1-70b-instruct
- accounts/fireworks/models/llama-guard-2-8b
- accounts/fireworks/models/mistral-7b
- accounts/fireworks/models/mixtral-8x22b-hf
- accounts/fireworks/models/mixtral-8x22b-instruct-hf
- accounts/fireworks/models/mixtral-8x7b
- accounts/fireworks/models/mixtral-8x7b-instruct-hf
- accounts/fireworks/models/mistral-7b-instruct-v0p2
- accounts/fireworks/models/code-qwen-1p5-7b
- accounts/fireworks/models/qwen2-72b-instruct
- accounts/fireworks/models/stablecode-3b
- accounts/yi-01-ai/models/yi-large
Additional base models (including custom models) are supported for enterprise accounts.
The base model name is specified in fireworks.json
.
LoRA ranks
The LoRA rank must be an integer between 4 and 64, inclusive.
Supported target modules
Currently, the following target modules are supported:
- Llama and Mixtral models (all linear layers)
q_proj
k_proj
v_proj
o_proj
up_proj
/w1
down_proj
/w2
gate_proj
/w3
block_sparse_moe.gate
The target modules are specified in adapter_config.json
.
Was this page helpful?