huggingface/text-generation-inference
5
0
Fork 0
mirror of https://github.com/huggingface/text-generation-inference.git synced 2025-09-11 12:24:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove some scaffolding.

Browse Source
This commit is contained in:
Nicolas Patry 2024-12-04 21:54:25 +01:00
parent a78b6fd1e8
commit ca8a115adc
No known key found for this signature in database
GPG Key ID: D2920555C90F704C
2 changed files with 1 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
launcher/src/main.rs
View File

@ -80,7 +80,7 @@ fn get_config(
}; };
let content = std::fs::read_to_string(filename)?; let content = std::fs::read_to_string(filename)?;
let config: RawConfig = serde_json::from_str(&content).expect("?"); let config: RawConfig = serde_json::from_str(&content)?;
let config: Config = config.into(); let config: Config = config.into();
Ok(config) Ok(config)

4
server/text_generation_server/models/flash_causal_lm.py
View File

@ -2112,10 +2112,6 @@ class FlashCausalLM(Model):
if prefill and prefill_logprobs: if prefill and prefill_logprobs:
# Get prefill logprobs with inplace softmax (avoid copying the `out` tensor (max_batch_prefill_tokens * vocab_size)) # Get prefill logprobs with inplace softmax (avoid copying the `out` tensor (max_batch_prefill_tokens * vocab_size))
free_memory = get_free_memory(self.device, MEMORY_FRACTION)
logger.info(f"Free memory {free_memory / 1e9}GB")
logmemory = out.nelement() * out.element_size()
logger.info(f"Log memory {logmemory / 1e9}GB")
torch.log_softmax(out, -1, out=out) torch.log_softmax(out, -1, out=out)
prefill_logprobs_tensor = out prefill_logprobs_tensor = out
prefill_logprobs = torch.gather( prefill_logprobs = torch.gather(